Submitted by brad on Fri, 2014-12-19 13:39.
Yesterday’s note on Here’s maps brought up the question of the wisdom of map-based driving. While I addressed this a bit earlier let me add a bit more detail.
A common first intuition is that because people are able to drive just fine on a road they have never seen before that this is how robots will do it. They are bothered that present designs instead create a super-detailed map of the road by having human driven cars scan the road with sensors in advance. After all, the geometry of the road can change due to construction; what happens then?
They hope for a car that, like a human, can build its model of the road in real time while driving the road for the first time. That would be nice, of course, and gives you a car that can drive most roads right away, without needing to map them. But it’s a much harder problem to solve, and unlikely to ever be solved perfectly. Car companies are building very simple systems which can follow the lines on a freeway under human supervision without need for a map. But real city streets are a different story.
The first thing to realize is that any system which could build the correct model as you drive is a system that could build a map with no human oversight, so the situations are related. But building a map in advance is always going to have several very large advantages:
- You build the map from not just one scan of the road, but several, and done in different lanes and directions. As a result, you get 3-D scans of everything from different angles, and can build a superior model of the world.
- Using multiple scans lets you learn about things that are stationary but move one day to the next, like parked cars.
- You can process the data using a cloud supercomputer in as much time, memory and data storage as you want. Your computer is effectively thousands of times more capable.
- Humans can review the map built by the software if there’s anything it is uncertain about (or even if there is nothing) at their leisure.
- Humans can also test the result of the automatic and guided mapping to assure accuracy with one extra drive down the road.
In turn there are disadvantages
- At times, such as construction, the road will have changed from when it was mapped
- This process costs effort, and so the vehicle either does not drive off the map, or only handles a more limited set of simpler roads off the map.
The advantages are so great that even if you did have a system which could handle itself without a map, it is still always going to be able to do better
with a map. Even with a great independent system you would want to make an effort to map the most popular roads and the most complex roads, up to the limit of your budget. The cost is an issue, but the cost of mapping roads is nothing compared to the cost of building or maintaining them. It’s a few times driving down the road, and some medium-skilled labour.
The road has changed
Let’s get to the big issue — the map is wrong, usually because construction has changed it.
First of all, we must understand that the sensors always disagree with the map, because the sensors are showing all the other cars and pedestrians etc. Any car has to be able to perceive these and drive so as not to hit them. If a traffic cone, “road closed” sign or flagman appears in the road, a car is not going to just plow into them because they are not on the map! The car already knows where not to go, the question is where it should go when the lanes have changed.
Even vehicles not rated to drive any road without a map can probably still do basic navigation and stay within their lane markers without a map. For the 10,000 miles of driving you do in a year, you need a car that does that 99.99999% of the time (for which you want a map) but it may be acceptable to have a car that’s only 99.9% able to do that for the occasional mile of restriped road. Indeed, when there are other, human-driven cars on the road, a very good strategy is just to follow them — follow one in front, and watch cars to the side. If the car has a clear path following new lane markers or other cars, it can do so.
Google, for example, has shown videos of their vehicle detecting traffic cones and changing lanes to obey the cones. That’s today — it is only going to get better at this.
But not all the time. There will be times when the lanes are unclear (sometimes the old lanes are still visible or the new ones are not well marked.) If there are no other cars to follow, there are also no other cars to hit, and no other traffic to block.
Still, there will be times when the car is not sure of where to go, and will need help. Of course, if there is a passenger in the car, as there would be most of the time, that passenger can help. They don’t need to be a licenced driver, they just need to be somebody who can point on the screen and tell the car which of the possible paths it is considering is the right one. Or guide it with something like a joystick — not physically driving but just guiding the car as to where to go, where to turn.
If the car is empty, and has a network connection, it can send a picture, 3-D scan and low-res video to a remote help station, where a person can draw a path for the car to go for its next 100 meters, and keep doing that. Not steering the car but helping it solve the problem of “where is my lane?” The car will be cautious and stop or pull over for any situation where it is not sure of where to go, and the human just helps it get over that, and confirms where it is safe to go.
If the car is unmanned and has no network connection of any kind, and can’t figure out the road, then it will pull over, or worst case, stop and wait for a human to come and help. Is that acceptable? Turns out it probably is, due to one big factor:
This only applies to the first car to encounter an unplanned, unreported construction zone
We all drive construction zones every day. But it’s much more rare that we are the first car to drive the construction zone as they are setting it up. And most of the rules I describe above are only for the first connected car to encounter a surprise change to the road. In other words, it’s not going to happen very often. Once a car encounters a surprise change to the road, it will report the problem with the map. Immediately all other cars will know about the zone.
If that first car is able to navigate the new zone, it will be scanning it with sensors, and uploading that data, where a crew can quickly build a corrected map. Within a few minutes, the map and the road will no longer differ. And that first car will be able to navigate the new zone 99.999% of the time — either because it has a human on board, remote human help or it’s a simple enough change that the car is able to drive it with an incorrect map.
In addition, the construction zone has to be a surprise. That means that, in spite of regulations, the construction crews did not log plans for it in the appropriate databases. Today that happens fairly often, but over time it’s going to happen less. In fact, there are plans to have transponders on construction equipment and even traffic cones that make it impossible to create a new construction zone without it showing up in the databases. Setting up a road change has a lot of strongly enforced safety rules, and I predict we’ll see “Get out your smartphone and make sure the zone is in the database before you create it” as one of them, especially since that’s so easy to do.
(You have probably also seen that tools like Waze, driven by ordinary human driver smartphones, are already mapping all the construction zones when they pop up.)
If a complex zone is present and unmapped, unmanned cars just won’t route through there until the map is updated. The more important the zone, the more quickly it will get updated. If need be, a mapping worker will go out in a car before work even begins. If a plan was filed, we’ll also know the plan for the zone, and whether cars can handle it with an old map or not.
Most of the time, though, a human passenger will be there to guide the car through the zone. Not to steer — there may not be a steering wheel — but to guide. The car will go slowly and stay safe.
Once a car is through, it will send the scans up to the mapping center, and all future cars will have a map to guide them until the crew changes the road again without logging it. I believe that doing so should be made against safety regulations, and be quite rare.
So look at those numbers. I will hope it’s reasonable to expect that 99% of construction zones will be logged in road authority databases before they begin. Of the 1% that aren’t, there will be a first robocar to encounter the zone. 90% of the time that car will have a passenger able to help. For the 10% unmanned cars, I predict a data network will be available 99% of the time. (Some would argue 100% of the time because unmanned cars will just not go where there is not a data connection, and we may also get new data services like Google’s Loon, or Facebook’s drone program to assure coverage everywhere.)
So now we are looking at one construction zone in 100,000 where there was no warning, there is no human, and there is no data. But we’ve rated are car as able to handle handle off-map driving 99.9% of the time. For the other .1%, it decides it can’t see a clear path, and pulls over. When it doesn’t report back in on the other side of the data dead zone, a service vehicle is dispatched and fixes the problem.
So now in one in 100,000,000 construction zones, we have a car deciding to pull over. Perhaps for half of those, it can’t figure out how to pull over, and it stops in the lane. Not great — but this is one in 200 million construction zones. In other words, it happens with much less frequency than accidents or stalled cars. And there is even a solution. If a construction worker flashes an ID card at the car’s camera when it’s in a confused state, the car can then follow that worker to a place to stop. In fact, since the confused state is so rare, there is probably not even a need for an ID card. Just walk up, make a “follow me” gesture and walk the car where it needs to go.
Tweak these numbers as you like. Perhaps you think there will be far more construction zones not logged in databases. Perhaps you think the car’s ability to drive a changed zone will only be 50%. Perhaps you think there will still be lots of unmanned cars running in wireless dead zones in 2020. Even so the number of cars that stop and give up will still be far fewer than the number of cars that block roads today due to accidents and mechanical problems. In other words, no big whoop.
It’s important to realize that unmanned cars are not in a hurry. They can avoid zones they are not comfortable with. If they can’t get through at all, the taxi company sending the car can just send another from a different direction in almost all cases.
It’s also important to realize that cars in an uncertain situation are also not in a big hurry. They will slow until they can be sure they are safe and able to handle the road. Slow, it turns out, is easy. Slow and heavy traffic (ie. a traffic jam) is actually also very easy — you don’t even need to see the lines on the road to handle that one; you usually can’t.
Once again this is only for the first car to encounter the surprise zone. Much more common will be a car that is the first to encounter a planned zone. This car will always have a competent passenger, because the service will not direct an unmanned car into an unknown construction zone where there is no data. This passenger will get plenty of warning, and their car may well pull over so there is no transition from full-auto to semi-auto while the car is moving. Then this person will guide the car through the zone at reduced speed. Probably just with a joystick, though possibly there will handlebars that can pop out or plug in if true semi-manual driving is needed.
New road signs
Road signs are a different problem. Already there are very decent systems for recognizing road signs captured by the camera — systems that actually do better at it than human beings. But sometimes there are road signs with text, and the system may recognize them, but not understand them. Here again we may call upon human beings, either in the vehicle, or available via a data connection. Once again, this is only for the first unmanned car to encounter the new road sign.
I will propose something stronger, though. I believe there should be a government mandated database of all road signs. Further, I believe the law should say that no road sign has legal effect until it is entered in the database. Ie. if you put up a sign with a new speed limit, it is not a violation of the limit to ignore the sign until the sign is in the database. At least not for robots. Once again, all this needs is that the crews putting in the signs have smartphones so they can plonk the sign on the map and enter what it is.
We may never need this, though, because the ability of computers to read signs is getting very good. It may be faster to just make it even better than to wait for a law that mandates the database. With a 3-D map, you will never miss a brand new sign, but you might get confused by a changed sign — you will know it changed but may need to ask for help to understand it if it is non-standard. There are already laws that standardize road signs, but only to a limited extent. Even so, the number of sign styles in any given country is still a very manageable number.
Random road events
Sometimes driving geometry changes not due to construction, but due to accidents and the environment. Trees get knocked down. Roads flood. Power lines may fall. The trees will be readily seen, and for the first car to come to a fallen tree, the procedure will be similar, though in a low traffic area the vehicles will be programmed to go around them, as they are for stalled cars and slow moving vehicles. Flooding and power lines are more challenging because they are harder to see. Flooding, of course, does not happen by surprise. That there is flooding in a region will be well known so cars will be on the lookout for it. Human guides will again be key.
A plane is not a bird
Aircraft do not fly by flapping their wings, and robocars will not see the world as people do nor drive as they do. When they have accurate maps, it gives us much more confidence in their safety, particularly the ability to pick the right path reliably at speed. But they have a number of tools open to them for driving a road that doesn’t match the map precisely without needing to have the ability to drive unmapped roads 99.999999% of the time. That’s a human level ability and they don’t need it.
Submitted by brad on Thu, 2014-12-18 14:14.
I see new articles on robocars in the press every day now, though most don’t say a lot new. Here, however, are some of the recent meaningful stories from the last month or two while I’ve been on the road. There are other sites, like the LinkedIn self-driving car group and others, if you want to see all the stories.
Winners chosen in UK competition
Four cities in the UK have been chosen for testing and development of robcars using the £10 million funding contest. As expected, Milton Keynes was chosen along with Coventry, and also Greenwich and Bristol. The BBC has more.
Chinese competition has another round
Many don’t know it, but China has been running its own “DARPA Grand Challenge” style race for 6 years now. The entrants are mostly academic, and not super far along, but the rest of the world stopped having contests long ago, much to its detriment. I was recently in Beijing giving a talk about robocars for guests of Baidu — my venue was none other than the Forbidden City — and the Chinese energy is very high. Many, however, thought that an announcement that Baidu would provide map data for BMW car research meant that Baidu was doing a project the way Google is. It isn’t, at least for now.
LA Mayor wants the cars
I’ve seen lots of calls from cities and regions that robocars come there first. In the fall, the mayor of Los Angeles made such a call. What makes this interesting is that LA is indeed a good early target city, with nice wide and simple roads, lots of freeways, and relatively well-behaved drivers compared to the rest of the world. And it’s in California, which is where a lot of the best development is happening, although that’s all in the SF Bay Area.
Concept designs for CES and beyond
More interesting concept cars are arising, as designers realize what they can do when freed of having a driver’s seat that faces forward and has all the controls, and as electric drivetrains allow you to move around where the drivetrain goes. Our friends at the design firm IDEO came up with some concepts that are probably not realistic but illustrate worthwhile principles. In particular, their vision of the delivery robot is quite at odds with mine. I see delivery robots as being very small, just suitcase sized boxes on wheels, except for the few that are built for very large cargo like furniture and industrial deliveries. Delivery robots will come to you on your schedule, not on the delivery company’s schedule. There will be larger robots with compartments when you can service a group of people who live together, but there is a limit to how many you can serve and still deliver at exactly the right time that people expect.
Everybody is also interested to see what Daimler will unveil at the Consumer Electronics Show. They showed off an interior with face-to-face seating and everybody wearing a VR headset, and have been testing a car under wraps.
It’s interesting to think about the VR headset. A lot of people would get sick if jostled in a car while wearing a VR headset. However, it might be possible to have the VR headset deliberately bounce the environment it’s showing you, so that it looks like you’re riding a car in that environment that’s bumping just the way you are. Or even walking.
Here (Nokia/Navteq) builds a big library of HD maps
Robocars work better if they get a really detailed map of their environment to drive with. Google’s project is heavily based on maps, and they have mapped out all the roads they test near Google HQ. Nokia’s “Here” division has decided to enter this in a big way. Nokia calls its projects “HD Maps,” which is a good name because you want to make it clear that these are quite unlike the navigation maps we are used to from Google, Here and other companies. These maps track every lane and path a car could take on the road, but also every lane marker, every curb, every tree — anything that might be seen by the cameras and 3D sensors.
Nokia makes the remarkable claim to have produced 1.2 million miles of HD Maps in 30 countries in the last 15 months. That’s remarkable because Google declared that one of their unsolved problems was that the cost of producing maps, and they were working to bring that cost down. Either Nokia/Here has made great strides in reducing that cost, or their HD Maps are not quite at the level of accuracy and detail that might be needed.
Nonetheless, the cost of the mapping will come down. In fact, many people express surprise when they learn that the cars rely so heavily on maps, as they expect a vehicle that, like a human being, can easily drive on a road they’ve never seen before, with no map. Humans can do that, but a car that could do that is also a car that could build the sort of map we’re talking about, in real time. Making the map ahead of time has several advantages, and is easier to do than doing it in real time. Perhaps some day that real-time map builder (what roboticists call Simultaneous localization and mapping) will arise, but for now, pre-mapping is the way to go.
510 Systems story told (sort of.)
There was recently press about the kept-quiet acquisition by Google of 510 Systems. I was at Google at the time, and it involves friends of mine, so I will have to say there are some significant errors in the story, but it’s interesting to see it come out. It wasn’t really that secret. What Anthony did with PriBot was hardly secret — he was on multiple TV shows for his work — and that he was at Google working at first on Streetview and later on the car was also far from secret. But it wasn’t announced so nobody picked up on it.
Submitted by brad on Tue, 2014-12-16 01:07.
Uber is spreading fast, and running into protests from the industries it threatens, and in many places, the law has responded and banned, fined or restricted the service. I’m curious what its battles might teach us about the future battles of robocars.
Taxi service has a history of very heavy regulation, including government control of fares, and quota/monopolies on the number of cabs. Often these regulations apply mostly to “official taxis” which are the only vehicles allowed to pick up somebody hailing a cab on the street, but they can also apply to “car services” which you phone for a pick-up. In addition, there’s lots of regulation at airports, including requirements to pay extra fees or get a special licence to pick people up, or even drop them off at the airport.
Why we have Taxi regulation and monopolies
The heavy regulation had a few justifications:
- When hailing a cab, you can’t do competitive shopping very easily. You take the first cab to come along. As such there is not a traditional market.
- Cab oversupply can cause congestion
- Cab oversupply can drive the cost of a taxi so low the drivers don’t make a living wage.
- We want to assure public safety for the passengers, and driving safety for the drivers.
- A means, in some places, to raise tax revenue, especially taxing tourists.
Most of these needs are eliminated when you summon from an app on your phone. You can choose from several competing companies, and even among their drivers, with no market failure. Cabs don’t cruise looking for fares so they won’t cause much congestion. Drivers and companies can have reputations and safety records that you can look up, as well as safety certifications. The only remaining public interest is the question of a living wage.
Taxi regulations sometimes get stranger. In New York (the world’s #1 taxi city) you must have one of the 12,000 “medallions” to operate a taxi. These medallions over time grew to cost well north of $1 million each, and were owned by cab companies and rich investors. Ordinary cabbies just rented the medallions by the hour. To avoid this, San Francisco made rules insisting a large fraction of the cabs be owned by their drivers, and that no contractual relationship could exist between the driver and any taxi company.
This created the situation which led to Uber. In San Francisco, the “no contract” rule meant if you phoned a dispatcher for a cab, they had no legal power to make it happen. They could just pass along your desire to the cabbie. If the driver saw somebody else with their arm up on the way to get you, well, a bird in the hand is worth two in the bush, and 50% of the time you called for a cab, nobody showed up!
Uber came into that situation using limos, and if you summoned one you were sure to get one, even if it was more expensive than a cab. Today, that’s only part of the value around the world but crazy regulations prompted its birth.
The legal battles (mostly for Uber)
I’m going to call all these services (Uber, Lyft, Sidecar and to some extent Hail-O) “Online Ride” services. read more »
Submitted by brad on Fri, 2014-12-12 09:24.
Dave Barry once wrote that there is a federal law that no two people on a plane can pay the same price for their seat. Airlines use complex systems to manage ticket prices, constantly changing them based on expected demand and competition, and with over a dozen fare classes with different rules.
When it comes to the rules, a usual principle is that only the more expensive tickets give you the flexibility to change your plans. For any reasonable price, you will have change and cancellation fees, and for the lowest cost tickets, changes are next to impossible. This is compounded by the fact that changes usually require paying the difference to the current price, but the current price in the few days before a flight is the very expensive flexible price. Missing a flight or deciding to move a fight a day can be hugely expensive.
The flexible tickets are ridiculously expensive as well, often 2x or even 3x the inflexible cost. In general, unless you change your plans a lot, you are still better off buying the cheap inflexible tickets and then eating the high cost on the relatively rare times you make changes. (Many airlines do offer cheap “same day” changes, particularly to status flyers.)
Flexible tickets can command this price because they are of greatest use to business passengers. We fly more on short notice, and need to make sudden changes, while people on vacation generally do have a fixed schedule. Airlines know business customers will pay more, and so they search for things that only business passengers want, and charge heavily for them.
Sell me a ticket where I have to be flexible
For leisure travel, here’s an alternative. Sell me a ticket that allows reasonable and low-cost changes when seats are available. Make it not a big deal to let me leave when I want to. To make this ticket cheap, but a big burden on me — the airline can also delay my flight.
What this would mean is that up to some amount of time, like 24 hours before the flight, the airline can email me and say, “Sorry, that flight is selling out, we’ve moving you to another flight.” The other flight would be within a time window — the longer the window, the cheaper the ticket. 24 to 48 hours would usually be enough.
The typical business passenger is not going to tolerate this. In business, time is money and losing a day just isn’t an option.
Some leisure passengers would not tolerate it either. If you have other bookings that are hard to change, like sold-out hotels, or a cruise, you don’t want to miss them. (Though in the world of flight cancellations you have to prepare for this sometimes.) But many hotels and other things are pretty flexible.
Most could handle such a rule going home, unless they are going home and must get to work the next day. For retired people, and the many people who work flexible schedules (consultants, writers and many other self-employed) it is not a big issue to get home a day or two late. And for many of these people it’s also not a big issue to arrive at the destination a day late, and certainly not a few hours late. In addition, many people taking an extended trip to multiple cities would be perfectly fine with the idea that they might spend an extra day in Rome and a day less in London, or vice versa. (On shorter trips with several flights a day, the delay might well be only a few hours.)
You could also offer the airline the power to make you leave earlier, but they would have to give you more notice on most legs.
This is great for the airline. They get the power to move people off full planes to replace them with high revenue customers at no cost, and put them on planes that are less full, where the seats are almost free. (If both planes are full, they would not move you.) Today they do this by asking for volunteers and paying them with vouchers, or on some occasions doing a forced bumping.
This is like standby, in a way, but less uncertain than that. A bit more like the way employees fly free on their off-hours.
There is one class of business passenger who might tolerate this, namely those making a visit to a branch office. They might be able to continue work for another day at the branch rather than go home if they don’t have meetings scheduled. I don’t think there would be a lot of this, unless you could also do it for business class tickets.
As part of the deal, the airline would also offer you a guaranteed low rate on an airport hotel for your extra day. They already have negotiated rates and spaces. With advance notice, though, you will probably be able to stay at your own hotel unless you travel at a sold-out time. These fares might make more sense in shoulder seasons, where hotel changes are easy.
As a passenger
As a reminder, you do all this to save money on a flexible ticket. You get a ticket where you can leave whenever you want without a large change fee. For a certain class of voyager (the retired in particular) this is the sort of ticket they want. Of course, seats have to be available, you can’t switch to a sold-out flight, and seat selection may be limited if you do things on short notice. But it need not always be on short notice.
The notice from the airline could even be long, too. Their computers are estimating the load all the time, and they might send you a request to move even a week or month in advance. For a higher cost, you might lengthen the window so you need a week’s notice if you are going to be moved (and they might then move you forward or backward.)
Submitted by brad on Tue, 2014-12-09 13:25.
When I talk about robocars, I often get quite opposite reactions:
- Americans, in particular, will never give up car ownership! You can pry the bent steering wheel from my cold, dead hands.
- I can’t see why anybody would own a car if there were fast robotaxi service!
- Surely human drivers will be banned from the roads before too long.
I predict neither extreme will be true. I predict the market will offer all options to the public, and several options will be very popular. I am not even sure which will be the most popular.
- Many people will stick to buying and driving classic, manually driven cars. The newer versions of these cars will have fancy ADAS systems that make them much harder to crash, and their accident levels will be lower.
- Many will buy a robocar for their near-exclusive use. It will park near where it drops them off and always be ready. It will keep their stuff in the trunk.
- People who live and work in an area with robotaxi service will give up car ownership, and hire for all their needs, using a wide variety of vehicles.
- Some people will purchase a robocar mostly for their use, but will hire it out when they know they are not likely to use it, allowing them to own a better car. They will make rarer use of robotaxi services to cover specialty trips or those times when they hired it out and ended up needing it. Their stuff will stay in a special locker in the car.
In addition, people will mix these models. Families that own 2 or more cars will switch to owning fewer cars and hiring for extra use and special uses. For example, if you own a 2 person car, you would summon a larger taxi when 3 or more are together. In particular, parents may find that they don’t want to buy a car for their teen-ager, but would rather just subsidize their robotaxi travel. Parents will want to do this and get logs of where their children travel, and of course teens will resist that, causing a conflict. read more »
Submitted by brad on Thu, 2014-12-04 09:12.
In August, I attended the World Science Fiction Convention (WorldCon) in London. I did it while in Coeur D’Alene, Idaho by means of a remote Telepresence Robot(*). The WorldCon is half conference, half party, and I was fully involved — telepresent there for around 10 hours a day for 3 days, attending sessions, asking questions, going to parties. Back in Idaho I was speaking at a local robotics conference, but I also attended a meeting back at the office using an identical device while I was there.
After doing this, I have written up a detailed account of what it’s like to attend a conference and social event using these devices, how fun it is now, and what it means for the future.
You can read Attending the World Science Fiction convention on the other side of the world by remote telepresence robot
For those of you in the TL;DR crowd, the upshot is that it works. No, it’s not as good as being there in person. But it is a substantial fraction of the way there, and it’s going to get better. I truly feel I attended that convention, but I didn’t have spend the money and time required to travel to London, and I was able to do other things in Idaho and California at the same time.
When you see at new technology that seems not quite there yet, you have to decide — is this going to get better and explode, or is it going to fizzle. I’m voting for the improvement argument. It won’t replace being there all of the time, but it will replace being there some of the time, and thus have big effects on travel — particularly air travel — and socialization. There are also interesting consequences
for the disabled, for the use of remote labour and many other things.
(*)As the maker will point out, this is not technically a robot, just a remote controlled machine. Robots have sensors and make some of their own decisions on how they move.
Submitted by brad on Thu, 2014-11-27 14:32.
Recently I tried Facebook/Oculus Rift Crescent Bay prototype. It has more resolution (I will guess 1280 x 1600 per eye or similar) and runs at 90 frames/second. It also has better head tracking, so you can walk around a small space with some realism — but only a very small space. Still, it was much more impressive than the DK2 and a sign of where things are going. I could still see a faint screen door, they were annoyed that I could see it.
We still have a lot of resolution gain left to go. The human eye sees about a minute of arc, which means about 5,000 pixels for a 90 degree field of view. Since we have some ability for sub-pixel resolution, it might be suggested that 10,000 pixels of width is needed to reproduce the world. But that’s not that many Moore’s law generations from where we are today. The graphics rendering problem is harder, though with high frame rates, if you can track the eyes, you need only render full resolution where the fovea of the eye is. This actually gives a boost to onto-the-eye systems like a contact lens projector or the rumoured Magic Leap technology which may project with lasers onto the retina, as they need actually render far fewer pixels. (Get really clever, and realize the optic nerve only has about 600,000 neurons, and in theory you can get full real-world resolution with half a megapixel if you do it right.)
Walking around Rome, I realized something else — we are now digitizing our world, at least the popular outdoor spaces, at a very high resolution. That’s because millions of tourists are taking billions of pictures every day of everything from every angle, in every lighting. Software of the future will be able to produce very accurate 3D representations of all these spaces, both with real data and reasonably interpolated data. They will use our photographs today and the better photographs tomorrow to produce a highly accurate version of our world today.
This means that anybody in the future will be able to take a highly realistic walk around the early 21st century version of almost everything. Even many interiors will be captured in smaller numbers of photos. Only things that are normally covered or hidden will not be recorded, but in most cases it should be possible to figure out what was there. This will be trivial for fairly permanent things, like the ruins in Rome, but even possible for things that changed from day to day in our highly photographed world. A bit of AI will be able to turn the people in photos into 3-D animated models that can move within these VRs.
It will also be possible to extend this VR back into the past. The 20th century, before the advent of the digital camera, was not nearly so photographed, but it was still photographed quite a lot. For persistent things, the combination of modern (and future) recordings with older, less frequent and lower resolution recordings should still allow the creation of a fairly accurate model. The further back in time we go, the more interpolation and eventually artistic interpretation you will need, but very realistic seeming experiences will be possible. Even some of the 19th century should be doable, at least in some areas.
This is a good thing, because as I have written, the world’s tourist destinations are unable to bear the brunt of the rising middle class. As the Chinese, Indians and other nations get richer and begin to tour the world, their greater numbers will overcrowd those destinations even more than the waves of Americans, Germans and Japanese that already mobbed them in the 20th century. Indeed, with walking chairs (successors of the BigDog Robot) every spot will be accessible to everybody of any level of physical ability.
VR offers one answer to this. In VR, people will visit such places and get the views and the sounds — and perhaps even the smells. They will get a view captured at the perfect time in the perfect light, perhaps while the location is closed for digitization and thus empty of crowds. It might be, in many ways, a superior experience. That experience might satisfy people, though some might find themselves more driven to visit the real thing.
In the future, everybody will have had a chance to visit all the world’s great sites in VR while they are young. In fact, doing so might take no more than a few weekends, changing the nature of tourism greatly. This doesn’t alter the demand for the other half of tourism — true experience of the culture, eating the food, interacting with the locals and making friends. But so much commercial tourism — people being herded in tour groups to major sites and museums, then eating at tour-group restaurants — can be replaced.
I expect VR to reproduce the sights and sounds and a few other things. Special rooms could also reproduce winds and even some movement (for example, the feeling of being on a ship.) Right now, walking is harder to reproduce. With the OR Crescent Bay you could only walk 2-3 feet, but one could imagine warehouse size spaces or even outdoor stadia where large amounts of real walking might be possible if the simulated surface is also flat. Simulating walking over rough surfaces and stairs offers real challenges. I have tried systems where you walk inside a sphere but they don’t yet quite do it for me. I’ve also seen a system where you are held in place and move your feet in slippery socks on a smooth surface. Fun, but not quite there. Your body knows when it is staying in one place, at least for now. Touching other things in a realistic way would require a very involved robotic system — not impossible, but quite difficult.
Also interesting will be immersive augmented reality. There are a few ways I know of that people are developing
- With a VR headset, bring in the real world with cameras, modify it and present that view to the screens, so they are seeing the world through the headset. This provides a complete image, but the real world is reduced significantly in quality, at least for now, and latency must be extremely low.
- With a semi-transparent screen, show the augmentation with the real world behind it. This is very difficult outdoors, and you can’t really stop bright items from the background mixing with your augmentation. Focus depth is an issue here (and is with most other systems.) In some plans, the screens have LCDs that can go opaque to block the background where an augmentation is being placed.
- CastAR has you place retroreflective cloth in your environment, and it can present objects on that cloth. They do not blend with the existing reality, but replace it where the cloth is.
- Projecting into the eye with lasers from glasses, or on a contact lens can be brighter than the outside world, but again you can’t really paint over the bright objects in your environment.
Getting back to Rome, my goal would be to create an augmented reality that let you walk around ancient Rome, seeing the buildings as they were. The people around you would be converted to Romans, and the modern roads and buildings would be turned into areas you can’t enter (since we don’t want to see the cars, and turning them into fast chariots would look silly.) There have been attempts to create a virtual walk through ancient Rome, but being able to do it in the real location would be very cool.
Submitted by brad on Mon, 2014-10-27 10:52.
There’s been a lot of press recently about an article in Slate by
Lee Gomes which paints a pessimistic picture of the future of
robocars, and particularly Google’s project. The Slate article
is a follow-on to a similar article in MIT Tech Review
Gomes and others seem to feel that they and the public were led to believe that
current projects were almost finished and ready to be delivered any day, and
they are disappointed to learn that these vehicles are still research projects
and prototypes. In a classic expression of the Gartner Hype Cycle there are
now predictions that the technology is very far away.
Both predictions are probably wrong. Fully functional robocars that can drive almost
everywhere are not coming this decade, but nor are they many decades away.
But more to the point, less-functional robocars are probably coming this decade — much sooner than
these articles expect, and these vehicles are much more useful and commercially
viable than people may expect.
There are many challenges facing developers, and those challenges will
keep them busy refining products for a long time to come. Most of those
challenges either already have a path to solution, or constrain a future vehicle
only in modest ways that still allow it to be viable. Some of the problems
are in the “unsolved” class. It is harder to predict when those
solutions will come, of course, but at the same time one should remember that
many of the systems in today’s research vehicles were in this class just a
few years ago. Tackling hard problems is just what these teams are good at
doing. This doesn’t guarantee success, but neither does it require you bet
And very few of the problems seem to be in the “unsolvable without human-smart AI” class,
at least none that bar highly useful operation.
Gomes’ articles have been the major trigger of press, so I will go over those issues
in detail here first. Later, I will produce an article that has even
more challenges than listed, and what people hope to do about them. Still, the critiques are written almost as though they
expected Google and others, rather than make announcements like “Look at the new milestone we are pleased to
have accomplished” to instead say, “Let’s tell you all the things we haven’t done yet.”
Gomes begins by comparing the car to the Apple Newton, but forgets that
9 years after the Newton fizzled we had the success of the Palm Pilot, and
10 years after that Apple came back with the world-changing iPhone. Today, the pace of
change is much faster than in the 80s.
Here are the primary concerns raised:
Maps are too important, and too costly
Google’s car, and others, rely on a clever technique that revolutionized
the DARPA challenges. Each road is driven manually a few times, and the
scans are then processed to build a super-detailed “ultramap” of all the
static features of the road. This is a big win because big server
computers get to process the scans in as much time as they need,
and see everything from different angles. Then humans can review and
correct the maps and they can be tested. That’s hard to beat, and you
will always drive better if you have such a map than if you don’t.
Any car that could drive without a map would effectively be a car that’s
able to make an adequate map automatically. As things get closer to that,
making maps will become cheaper and cheaper.
Naturally, if the road differs from the map, due to construction or
other changes, the vehicle has to notice this. That turns out to be
fairly easy. Harder is assuring it can drive safely in this situation.
That’s still a much easier problem than being able to drive safely
everywhere without a map, and in the worst case, the problem of the
changed road can be “solved” by just the ability to come to a safe stop.
You don’t want to do that super often, but it remains the fail-safe
out. If there is a human in the car, they can guide the vehicle in this.
Even if the vehicle can’t figure out where to go to be safe, the human
can. Even a remote human able to look at transmitted pictures can help
the car with that — not live steering, but strategic guidance.
This problem only happens to the first car to encounter the surprise construction.
If that car is still able to navigate (perhaps with human help,) the map can be quickly rebuilt, and
if the car had to stop, all unmanned cars can learn to avoid the zone.
They are unmanned, and thus probably not in a hurry.
The cost of maps
In the interests of safety, a lot of work is put into today’s maps. It’s a cost
that somebody like Google or Mercedes can afford if they need to, (after all, Google’s already
scanned every road
in many countries multiple times) but it would be high for smaller players. read more »
Submitted by brad on Sun, 2014-10-26 14:22.
A recent newspaper column where people complained about carpool cheats got me thinking — could cheating actually be a solution to some carpool problems?
For many years, the wisdom was that carpool lanes were helping traffic and the environment, but that wisdom has been changing, and it is now seen that the lanes actually hurt (at least the traffic) in many cases. As such, the new approach is to build “managed lanes” and in particular the High-Occupancy-Toll (HOT) lanes which let solo drivers pay to use the lane. In addition, low emission cars and motorcycles usually get to use the lanes solo.
Why does this help? It turns out that a typical configuration of 3 solo lanes and one carpool lane is performing badly when the carpool lane is well under capacity. The ideal road would have all 4 lanes running just under 100% capacity (which is around 2,000 cars per hour, or 8,000 for the whole road.) At rush hour, however, the lanes often collapse in congestion to stop and go, which can drop as low as 1,300 vehicles/hour.
Carpool approaches suggest that if you have one carpool lane running at less than capacity (and thus congestion free and highly attractive) that you will make people choose to carpool. Each carpool takes a car or two off the road, which is a win for congestion (and the environment.)
Consider one carpool situation, where the carpool lane is running free at 50% of capacity, and the other 3 lanes are at 100% of capacity. You’re now moving 7,000 vehicles/hour instead of 8,000, but that would be OK if it’s because you took more than 1,000
vehicles off the road.
Unfortunately that’s not even remotely true. The vast majority of the carpools on the road are natural carpools that would have happened anyway. Couples or families travelling together. “Kidpools” where in almost all cases no car was taken off the road. The permitted solo drivers in low emission vehicles and motorcycles don’t remove cars, but are greener. The number of “induced” carpools — carpools that were created because of the attractive travel time offered by the carpool lane — is quite low. Perhaps as low as 10%, but likely not more than 20%. HOV-3 lanes may have more induced carpools.
To make it worse, consider a carpool lane at 70% usage (good) but the 3 other lanes in congestion, and now getting 1,500 vehicles per hour. We’ve dropped our road to just 5,900 cars per hour. And at 20% induced carpools we only took 280 cars off the road, for a total of 6,180 instead of our ideal of 8,000. There is a zone of congestion where moving another 500 cars from the solo lanes to the carpool lane would relieve the congestion in the solos, and we would get closer to our 8,000.
That’s what HOT lanes are about. By charging a fee, they move solo drivers who are willing to pay to use the underutilized carpool lane, and we remove them from the solos, increasing their throughput as well. It’s a win-win-win. HOT lanes adjust the price — if the carpool lane is starting to fill up, the price jacks up. The goal is to keep the carpool lane enough below 100% capacity that it flows smoothly, which is good for flow and also what makes it attractive in the first place to make those induced carpools.
With HOT, you can have 1,000 carpoolers and 900 paying solos and also 200 induced carpools so the lane is now delivering the equivalent of 2,100 vehicles/hour and everybody wins. Letting efficient solos use the lane doesn’t involve money, but subsidizes efficient vehicles.
Without HOT, the bizarre conclusion is that cheaters are helping move traffic along. Cheaters only cheat when the carpool lane is going really well — ie. underutilized — and the solo lanes are getting congested. Cheaters take some load off the solo lanes and make use of the wasted capacity. They will not cheat if the carpool lane is not beating the solo lanes by a nice margin. If the carpool lane gets overloaded, they are going to leave it — why risk the ticket?
I should note that I have never, ever deliberately cheated in the carpool lane. (Like most, once or twice I have forgotten what time it was for a minute or two.) I am not trying to justify cheating, and in fact one concern is that some cheaters will read this and imagine they are doing a service. Cheaters are helping the system, but in a completely unfair and inappropriate way.
One reason we don’t have more HOT lanes, now that people realize that they are better, is that it costs a lot of money to put them in. Part of that money is for infrastructure — gantries, transponders, signs with prices, enforcement teams, operations teams. The biggest cost comes from the fact that generally people like to make HOT lanes truly separate from the main lanes, with a double line, and entry/exit only allowed at certain points. That means restriping or even new construction.
Many of the world’s transit systems work on an honour system. You have to buy a ticket, but nothing checks this. Instead, if you are caught on board without a ticket, you pay a fat fine. The fine is often calculated to balance the enforcement level, so that a regular cheater will be caught enough that it’s more expensive to cheat than to buy tickets. But often not a lot more expensive, as it turns out.
What if HOT lanes were the same way? Go ahead and cheat! Install random enforcement stations with cameras, and enforce enough so that any regular “cheater” gets fines which are calculated to collect as much or more money than the tolls.
The obvious flaw here is that this only works for the regular cheater. It’s too random, and an occasional lane user (or tourist) would be taking a big gamble, without enough use to balance it out. So we can add payment by cell phone to even things out.
Before leaving, or after arriving, tell your phone or browser you will be using or did use the lane. (The reason to do it in advance is
you will get a better price.) Your phone can show you the price, and some road signs will display it as well. This gives you a token which includes the time and your licence plate. If you get a fine notice, you can nullify it by providing the token.
(If you don’t care about privacy, you could register the licence plate directly. But I do care about privacy.)
This works with minimal new infrastructure. And payment via phone would be set to be cheaper than the average payment you would pay through random fines, so most people would do it. And all this happens with minimal new infrastructure, as long as you don’t need to reconfigure the lanes.
Enforcement can involve cameras, which may or may not be recording. You need enough of them so that people don’t just briefly switch out of the carpool lane just before coming to a camera, so this has some infrastructure cost. The camera would record the photo of the front seats of your car, and your plate. In isolate carpool lanes this does work better.
This is aimed at places where 2 is a carpool. It means something controversial. Carpoolers must share the front seat. And that means no kidpooling with children small enough to be required to ride in the back seat. Some people will hate that (parents) and some will love it (those who feel that kidpooling is unfair because it almost never causes an induced carpool.) This controversy can be some what mitigated by offering a discount to people who declare they are kidpooling (or better, multi-family kidpooling) with occasional checks.
It’s also an issue for Taxis, Uber and people with chauffeurs. Forcing the latter to pay won’t bother many people. Taxis can be given special status. Ad-hoc taxis, like Uber, can be told, “hey, just make the ride in the front if you want a free entry.” Is that such a big burden? If so, alternate systems can be set up, including requesting a token over the smartphone which can be compared to audited records of fares.
The camera stations could also photograph in through the sides of vehicles. Tinted side windows would not get to be carpools. This is harder than just doing the front, and harder to hide. And there would still be occasional live human observers, to the extent that cost allows.
To avoid risk of people wanting to use phones while driving, we simply allow you to buy a retroactive token within a day of your trip. (You don’t learn about your fine for a couple of days.) You could do that on the web, on a smartphone, by text (retroactive only) or even at any convenience store or gas station that has a payment machine. (This idea is not new. A decade ago I drove a toll road in Melbourne which lets you buy a toll pass at a gas station after you drive the road.)
Or, of course, just pay the fines if they are not that much more expensive, on average than buying tokens.
Even carpoolers could register that they carpooled, in case a problem comes up. Users will want to register an e-mail address or app address with the system under their plate to get notices of fines. If you don’t, notices would come by postal mail. If somebody else registers your plate and you don’t, it might delay notice of fines but you would fix this after the first one.
If the typical toll is $3, and the fine is $300, you probably would get a fine notice you need to nullify perhaps every 75 uses on average. This makes paying cheaper. The smartphone app would also notice when you travel the route and remind you.
To protect privacy, the system would not remember tokens it issues, and it would erase all images once it was confirmed the car was legit (carpool, allowed vehicle or had a token.) Only the images of non-carpools who did not respond with their token would be retained for issuing fines to their car.
There can be problems with photo enforcement if it is dark (as it is during winter for portions of rush hour) or in places where the sun is at just the wrong angle. The latter can be fixed because we know just where the sun will be. The former is more challenging. Cameras would need to be placed in line with suitable street lights, and have larger lenses. During the day used cell phones in rainproof cases with tiny solar panels could do the job at low cost.
Submitted by brad on Wed, 2014-10-22 13:26.
In late August, I visited Singapore to give an address at a special conference announcing a government sponsored collaboration involving their Ministry of Transport, the Land Transport Authority and A-STAR, the government funded national R&D centre. I got a chance to meet the minister and sit down with officials and talk about their plans, and 6 months earlier I got the chance to visit A-Star and also the car project at the National University of Singapore. At the conference, there were demos of vehicles, including one from Singapore Technologies, which primarily does military contracting.
Things are moving fast there, and this week, the NUS team announced they will be doing a live public demo of their autonomous golf carts and they have made much progress. They will be running the carts over a course with 10 stops in the Singapore Chinese and Japanese Gardens. The public will be able to book rides online, and then come and summon and direct the vehicles with their phones. The vehicles will have a touch tablet where the steering wheel will go. Rides will be free. Earlier, they demonstrated not just detecting pedestrians but driving around them (if they stay still) but I don’t know if this project includes that.
This is not the first such public demo - the CityMobil2 demonstration in Sardinia ran in August, on a stretch of beachfront road blocked to cars but open to bicycles, service vehicles and pedestrians. This project slowed itself to unacceptably slow speeds and offered a linear route.
The Singapore project will also mix with pedestrians, but the area is closed to cars and bicycles. There will be two safety officers on bicycles riding behind the golf carts, able to shut them down if any problem presents, and speed will also be limited.
Singapore is interesting because they have a long history of transportation innovation, and good reason for it. As a city-state, it’s almost all urban, and transportation is a real problem. That’s why congestion charging was first developed in Singapore, along with other innovations. Every vehicle in Singapore has a transponder, and they use them not just for congestion tolling, but to pay for parking seamlessly in almost all parking lots and a few other tricks.
In spite of this history of innovation, Singapore is also trending conservative — this might dampen truly fast innovation, but this joint project is a good start. Though I advised them that private projects will be able to move faster than public sector ones, in my view.
The NUS project is a collaboration with MIT, involving professor Emilio Frazzoli. Their press release has more details, including maps showing the route is non-linear but the speed is slow.
Submitted by brad on Tue, 2014-10-21 14:33.
Some recent announcements have caused lots of press stir, and I have not written much about them, both because of my busy travel schedule, but also because there is less news that we might imagine.
Tesla is certainly an important company to watch. As the first successful start-up car company in the USA, they are showing they know how to do things differently, taking advantage of the fact that they don’t have a baked in knowledge of “how a car company works” the way other companies do. Tesla’s announcements of plans for more self-driving are important. Unfortunately, the announcements around the new dual-motor Model S involve offerings quite similar to what can be found already in cars from Mercedes, Audi and a few others. Namely advanced ADAS and the combination of lane-keeping and adaptive cruise control to provide a hands-off cruise control where you must keep your eyes on the road.
One notable feature demonstrated by Tesla is automatic lane change, which you trigger by hitting a turn signal. That’s a good interface, but it must be made clear to people that they still have the duty to check that it’s safe to change lanes. It’s not that easy for a robocar’s sensors, especially the limited sensor package in the Telsa, to see a car coming up fast behind you in the next lane. On some highways relative speeds can get pretty high. You’re not likely to be hit by such cars, but in some cases that’s because they will probably brake for you, not because you did a fully safe lane change.
Much more interesting are Elon Musk’s predictions of a real self-driving car in 5 to 6 years. He means one where you can read a book, or even, as he suggests, go to sleep. Going to sleep is one of the greatest challenges, almost as hard as operating unmanned or carrying a drunk or disabled person. You won’t likely do that just with cameras — but 5 to 6 years is a good amount of time for a company like Tesla.
Another unusual thing about Tesla is that while they are talking about robocars a lot, they have also built one of the finest driver’s cars ever made. The Model S is great fun to drive, and has what I call a “telepathic” interface sometimes — the motors have so much torque that you can almost think about where you want to go and the vehicle makes it happen. (Other examples of telepathic interfaces include touch-typing and a stickshift.) In some ways it is the last car that people might want to automate. But it’s also a luxury vehicle, and that makes self-driving desirable too.
Another recent announcement creating buzz is Audi’s self-driving race car on a test track in Germany. Audi has done racing demos several times now. They are both important but also unimportant. It definitely makes sense to study how to control a car in extreme, high performance situations. To understand the physics of the tires so fully that you can compete in racing will teach lessons of use in danger situations (like accidents) or certain types of bad weather.
At the same time, real-world driving is not like racing, and nobody is going to be doing race-like driving on ordinary streets in their robocar. 99.9999% of driving consists of “staying in your lane” and some other basic maneuvers and so racing is fun and sexy but not actually very high on the priority list. (Not that teams don’t deserve to spend some of their time on a bit of fun and glory.) The real work of building robocars involves putting them through all the real-world road situations you can put them through, both real and in some cases simulated on a track or in a computer.
Google first showed its system to many people by having it race figure-8s on the roof parking lot at the TeD conference. The car followed a course through a group of cones at pretty decent speed and wowed the crowd with the tight turns. What most of the crowd didn’t know was that the cones were only there for show, largely. The car was guiding itself from its map of all the other physical things in the parking lot — line markers, pavement defects and more. The car is able to localize itself fine from those things. The cones just showed the public that it really was following the planned course. At the same time, making a car do that is something that was accomplished decades ago, and is used routinely to run “dummy cars” on car company test tracks.
A real demo turns out to be very boring, because that’s how being driven should be. I’m not saying it’s bad in any way to work on racing problems. The only error would be forgetting that the real-world driving problems are higher priority and success in them is less dramatic but more impressive in the technical sense.
This doesn’t mean we won’t see more impressive demos soon. Many people have shown off automatic braking. Eventually we will see demos of how vehicles respond in danger situations — accidents, pedestrians crossing into the road and the like. A tiny part of driving but naturally one we care about. And we will want them to understand the physics of what the tires and vehicle are capable of so that they perform well, but not so they can find the most efficient driving line on the track.
There was some debate about having a new self-driving car contest like the DARPA grand challenges, and a popular idea was man vs. machine, including racing. That would have been exciting. We asked ourselves whether a robot might have an advantage because it would have no fear of dying. (It might have some “fear” of smashing its owners very expensive car.) Turns out this happens on the racetrack fairly often with new drivers who try to get an edge by driving like they have no fear, that they will win all games of chicken. When this happens, the other drivers get together to teach that new driver a lesson. A lesson about cooperating and reciprocation in passing and drafting. So the robots would need to be programmed with that as well, or their owners would find a lot of expensive crashes and few victories.
Submitted by brad on Mon, 2014-10-13 20:59.
Here’s an interview with me in the latest Wall Street Journal on the subject of robocars and seniors.
This has always been a tricky question. Seniors are not early adopters, so the normal instinct would be to expect them to fear a new technology as dramatic as this one. Look at the market for simplified cell phones aimed at seniors who can’t imagine why they want a smartphone. Not all are like this, but enough are to raise the question.
Sometimes this barrier is broken. Pictures of grandchildren in e-mail brought grandparents online, as did video calls with them. Necessity overcomes the fear of change.
As people get older, they start losing driving ability. They die more often in accidents, eventually surpassing the rates of reckless teens, because they are more fragile, and they make mistakes that cause other people to hit them. Many seniors report troubles with vision at night, and they stop driving at night. In some cases, they get their licences taken away by the state — though the AARP and others fight this so it’s rare — or their kids take away their keys when things get really dangerous. And the kids become a taxi service for their parents.
The boomer generation, which took over the suburbs and exurbs have nice houses with minimal transit. Some find themselves leaving that home because they can’t drive any more and they will become a shut-in if they don’t do something.
The robocar offers answers to many of these problems. Safe transportation for those with disabilities. (Eventually even mild dementia.) Inexpensive taxi transportation anywhere, including those low-transit suburbs. And a chance to video chat with the grandchildren while on the way.
It’s no surprise that retirement communities are discussed as an early deployment zone for robocars. In those communities, you have a controlled street environment — often with heavy use of NEVs/golf carts already. You have people losing the ability to drive who have limited mobility needs. If they can get to basic shopping and a few other locations (including transit hubs to travel further) they can do pretty well.
Until the robocar came along, we were all doomed to lose the freedom cars gave us. This is no longer going to happen.
Submitted by brad on Tue, 2014-09-23 12:07.
I’ve been on the road a lot, talking in places like Singapore, Shenzen and Hong Kong, and visiting Indonesia which is a driving chaos eye-opener. In a bit over 10 hours I will speak at Swiss Re’s conference on robocars and insurance in Zurich. While the start will be my standard talk, in the latter section we will have some new discussion of liability and insurance.
A live stream of the event should be available at http://swissre.adobeconnect.com/theautonomouscar/ I talk at 8:45am Central European Summer Time.
A lot of news while I’ve been on the road — driving permits in California, new projects and the Singapore effort I was there at the announcement of. And lots of non-news that got people very excited like the “revelation” that Google’s car doesn’t drive in snow (nobody thought it could) or on all roads (nobody even suggested this) or that it was forced to add a steering wheel for testing (this was always planned, Google participating in the hearings writing those laws.) And lots of car company announcements from the ITS world congress (a conference that 2 years ago barely acknowledged the presence of self-driving cars.)
More to come later.
Submitted by brad on Fri, 2014-08-22 18:59.
There’s another video presentation by me that I did while visiting Big Think in NYC.
This one is on The NSA, Snowden and the “tradeoff” of Privacy and Security.
Earlier, I did a 10 minute piece on Robocars for Big Think that won’t be news to regular readers here but was reasonably popular.
Submitted by brad on Wed, 2014-07-30 13:01.
A whole raft of recent robocar news.
UK to modify laws for full testing, large grants for R&D
The UK announced that robocar testing will be legalized in January, similar to actions by many US states, but the first major country to do so. Of particular interest is the promise that fully autonomous vehicles, like Google’s no-steering-wheel vehicle, will have regulations governing their testing. Because the US states that wrote regulations did so before seeing Google’s vehicle, their laws still have open questions about how to test faster versions of it.
Combined with this are large research grant programs, on top of the £10M prize project to be awarded to a city for a testing project, and the planned project in Milton Keynes.
Jerusalem’s MobilEye going public in largest Israeli IPO
The leader in doing automated driver assist using cameras is Jerusalem’s MobilEye. This week they’re going public, to a valuation near $5B and raising over $600 million. MobilEye makes custom ASICs full of machine vision processing tools, and uses those to make camera systems to recognize things on the road. They have announced and demonstrated their own basic supervised self-driving car with this. Their camera, which is cheaper than the radar used in most fancy ADAS systems (but also works with radar for better results) is found in many high-end vehicles. They are a supplier to Tesla, and it is suggested that MobilEye will play a serious role in Tesla’s own self-driving plans.
As I have written, I don’t believe cameras are even close to sufficient for a fully autonomous vehicle which can run unmanned, though they can be a good complement to radar and especially LIDAR. LIDAR prices will soon drop to the low $thousands, and people taking the risk of deploying the first robocars would be unwise to not use LIDAR to improve their safety just to save a few thousand for early adopters.
Chinese search engine Baidu has robocar (and bicycle) project
Baidu is the big boy in Chinese search — sadly a big beneficiary of Google’s wise and moral decision not to be collaborators on massive internet censorship in China — and now it’s emulating Google in a big way by opening its own self-driving car project.
Various stories suggest a vehicle which involves regular handoff between a driver and the car’s systems, something Google decided was too risky. Not many other details are known.
Also rumoured is a project with bicycles. Unknown if that’s something like the “bikebot” concept I wrote about 6 years ago, where a small robot would clamp to a bike and use its wheels to deliver the bicycle on demand.
Why another search engine company? Well, one reason Google was able to work quickly is that it is the world’s #1 mapping company, and mapping plays a large role in the design of robocars. Baidu says it is their expertise in big data and AI that’s driving them to do this.
Velodyne has a new LIDAR
The Velodyne 64 plane LIDAR, which is seen spinning on top of Google’s cars and most of the other serious research cars, is made in small volumes and costs a great deal of money — $75,000. David Hall, who runs Velodyne, has regularly said that in volume it would cost well under $1,000, but we’re not there yet. He has released a new LIDAR with just 16 planes. The price, while not finalized, will be much higher than $1K but much lower than $75K (or even the $30K for the 32 plane version found on Ford’s test vehicle and some others.)
As a disclaimer, I should note I have joined the advisory board of Quanergy, which is making 8 plane LIDARs at a much lower price than these units.
Nissan goes back and forth on dates
Conflicting reports have come from Nissan on their dates for deployment. At first, it seemed they had predicted fairly autonomous cars by 2020. A later announcement by CEO Carlos Ghosn suggested it might be even earlier. But new reports suggest the product will be less far along, and need more human supervision to operate.
FBI gets all scaremongering
Many years ago, I wrote about the danger that autonomous robots could be loaded with explosives and sent to an address to wreak havoc. That is a concern, but what I wrote was that the greater danger could be the fear of that phenomenon. After all, car accidents kill more people every month in the USA than died at the World Trade Center 13 years ago, and far surpass war and terrorism as forms of violent death and injury in most nations for most of modern history. Nonetheless, an internal FBI document, released through a leak, has them pushing this idea along with the more bizarre idea that such cars would let criminals multitask more and not have to drive their own getaway cars. read more »
Submitted by brad on Wed, 2014-07-23 15:32.
I have many more comments pending on my observations from the recent AUVSI/TRB Automated Vehicles Symposium, but for today I would like to put forward an observation I made about two broad schools of thought on the path of the technology and the timeline for adoption. I will call these the aggressive and conservative schools. The aggressive school is represented by Google, Induct (and its successors) and many academic teams, the conservative school involves car companies, most urban planners and various others.
The conservative (automotive) view sees this technology as a set of wheels that has a computer.
The aggressive (digital) school sees this as a computer that has a set of wheels.
The conservative view sees this as an automotive technology, and most of them are very used to thinking about automotive technology. For the aggressive school, where I belong, this is a computer technology, and will be developed — and change the world — at the much faster pace that computer technologies do.
Neither school is probably entirely right, of course. It won’t go as gung-ho as a smartphone, suddenly in every pocket within a few years of release, being discarded when just 2 years old even though it still performs exactly as designed. Nor will it advance at the speed of automotive technology, a world where electric cars are finally getting some traction a century after being introduced.
The conservative school embraces the 4 NHTSA Levels or 5 SAE levels of technology, and expects these levels to be a path of progress. Car companies are starting to sell “level 2” and working on “level 3” and declaring level 4 or 5 to be far in the future. Google is going directly to SAE level 4.
The two cultures do agree that the curve of deployment is not nearly-instant like a smartphone. It will take some time until robocars are a significant fraction of the cars on the road. What they disagree on is how quickly that has a big effect on society. In sessions I attended, the feeling that the early 2020s would see only a modest fraction of cars being self-driving meant to the conservatives that they would not have that much effect on the world.
In one session, it was asked how many people had cars with automatic cruise control (ACC.) Very few hands went up, and this is no surprise — the uptake of ACC is quite low, and almost all of it is part of a “technology package” on the cars that offer it. This led people to believe that if ACC, now over a decade old, could barely get deployed, we should not expect rapid deployment of more complete self-driving. And this may indeed be a warning for those selling super-cruise style products which combine ACC and lanekeeping under driver supervision, which is the level 2 most car companies are working on.
To counter this, I asked a room how many had ridden in Uber or its competitors. Almost every hand went up this time — again no surprise. In spite of the fact that Uber’s cars represent an insignificant fraction of the deployed car fleet. In the aggressive view, robocars are more a service than a product, and as we can see, a robocar-like service can start affecting everybody with very low deployment and only a limited service area.
This dichotomy is somewhat reflected in the difference between SAE’s Level 4 and NHTSA’s. SAE Level 4 means full driving (including unmanned) but in a limited service area or under other limited parameters. This is what Google has said they will make, this is what you see planned for services in campuses and retirement communities. This is where it begins, and grows one region at a time. NHTSA’s levels falsely convey the idea that you slowly move to fully automated mode and immediately do it over a wide service area. Real cars will vary as to what level of supervision they need (the levels) over different times, streets and speeds, existing at all the levels at different times.
Follow the conservative model and you can say that society will not see much change until 2030 — some even talk about 2040. I believe that is an error.
Another correlated difference of opinion lies around infrastructure. Those in the aggressive computer-based camp wish to avoid the need to change the physical infrastructure. Instead of making the roads smart, make the individual cars smart. The more automotive camp has also often spoken of physical changes as being more important, and also believes there is strong value in putting digital “vehicle to vehicle” radios in even non-robocars. The computer camp is much more fond of “virtual infrastructure” like the detailed ultra-maps used by Google and many other projects.
It would be unfair to claim that the two schools are fully stratified. There are researchers who bridge the camps. There are people who see both sides very well. There are “computer” folks working at car companies, and car industry folks on the aggressive teams.
The two approaches will also clash when it comes to deciding how to measure the safety of the products and how they should be regulated, which will be a much larger battle. More on that later.
Submitted by brad on Mon, 2014-07-14 13:59.
It’s a big week for Robocar conferences.
In Berkeley, yesterday I attended and spoke at the “Robotics: Science and Systems” conference which had a workshop on autonomous vehicles. That runs to Wednesday, but overlapping and near SF Airport is the Automated Vehicles Symposium — a merger of the TRB (Transportation Research Board) and AUVSI conferences on the same topic. 500 are expected to attend.
Yesterday’s workshop was pretty good, with even a bit of controversy.
- Ed Olson on more of the lessons from aviation on handoff between automation and manual operation. This keeps coming up a a real barrier to some of the vehicle designs that have humans share the chores with the system.
- Jesse Levinson of Stanford’s team showed some very impressive work in automatic calibration of sensors, and even fusion of LIDAR and camera data, aligning them in real time in spite of movement and latency. This work will make sensors faster, more reliable and make fusion accurate enough to improve perception.
- David Hall, who runs Velodyne, spoke on the history of their sensors, and his plans for more. He repeated his prediction that in large quantities his sensor could cost only $300. (I’m a bit skeptical of that, but it could cost much, much less than it does today.) David made the surprising statement that he thinks we should make dedicated roads for the vehicles. (Surprising not just because I disagree, but because you could even get by without much LIDAR on such roads.)
- Marco Panove of Stanford showed research they did on Taxi models from New York and Singapore. The economics look very good. Dan Fagnant also presented related research assuming an on-demand semi shared system with pickup stations in every TAZ. It showed minimal vacant miles but also minimal successful rideshare. The former makes sense when it’s TAZ to TAZ (TAZs are around a square mile) but I would have thought there would be more rideshare. The conclusion is that VMT go up due to empty miles, but that rideshare can partially compensate, though not as much as some might hope.
- Ken Laberteaux of Toyota showed his research on the changing demographics of driving and suburbs. Conclusion: We are not moving back into the city, suburbanization is continuing. Finding good schools continues to drive people out unless they can afford private school are are childless.
The event had a 3-hour lunch break, where most went to watch some sporting event from Brazil. The Germans at the conference came back happier.
Some good technical talks presented worthwhile research
- Sheng Zhao and a team from UC Riverside showed a method to get cm accuracy in position and even in pose (orientation) from cheap GPS receivers, by using improved math on phase-matching GPS. This could also be combined with cheap IMUs. Most projects today use very expensive IMUs and GPSs, not the cheap ones you find in your cell phone. This work may lead to being able to get reliable data from low cost parts.
- Matthew Cornick and a team from Lincoln Lab at MIT showed very interesting work on using ground penetrating radar to localize. With GPR, you get a map of what’s below the road — you see rocks and material patterns down several feet. These vary enough, like the cracks and lines on a road, and so you can map them, and then find your position in that map — even if the road is covered in snow. While the radar units are today bulky, this offers the potential for operations in snow.
- A team from Toyota showed new algorithms to speed up the creation of the super-detailed maps needed for robocars. Their algorithms are good at figuring out how many lanes there are and when they start and stop. This could make it much cheaper to build the ultramaps needed in an automatic way, with less human supervision.
The legal and policy sessions got more heated.
- Bryant Walker Smith laid out some new proposals for how to regulate and govern torts about the vehicles.
- Eric Feron of Georgia Tech made proposals for how to do full software verification. Today formally proving and analysing code for correctness takes 0.6 hours per line of code — it’s not practical for the 50 million line (or more) software systems in cars today. Jonathan argues it can be made cheaper, and should be done. Note that fully half the cost of developing the 787 aircraft was software verification!
The final session, on policy included:
- Jane Lappin on how DoT is promoting research.
- Steve Shladover on how we’re all way to optimistic on timelines, and that coming up with tests to demonstrate superior safety to humans is very far away, since humans run 65,000 hours between injury accidents.
- Myself on why regulation should keep a light touch, and we should not worry too much about the Trolley Problem — which came up a couple of times.
- Raj Rajkumar of CMU on the success they have had showing the CMU/GM car to members of congress.
Now on to the AVS tomorrow.
Submitted by brad on Tue, 2014-06-24 09:45.
Some recent press and talks:
Earlier in June I sat down with “Big Think” for an interview they have titled “Robocars 101” explaining some of the issues around the cars.
I also did a short interview on NPR’s “All Things Considered” not long after Google’s new car was announced. What you might find interesting is how I did it. I was at a friend’s house in Copenhagen and went into a quiet room where they called me on my cell phone. However, I also started a simple audio recorder app on my phone. When we were done, I shared the mp3 of a better sample from the same microphone with them, which they mixed in.
As a result, the interview sounds almost like it was done in-studio instead of over an international cell phone call.
Videos of my talks at Next Berlin at at Dutch Media Future Week 2014 are also up. And a shortened talk at Ontario Centers for Excellence Discovery 2014 in Toronto May 12. There we had the Governor General of Canada as our opening act. :-) That’s just 3 of the 11 events I was at on that trip.
Completely off the Robocar track is a short interview with CNBC where I advise people to invest in Bitcoin related technology, not in bitcoins.
Submitted by brad on Sun, 2014-06-22 20:51.
So far it’s been big players like Google and car companies with plans in the self-driving space. Today, a small San Francisco start-up named Cruise, founded by Kyle Vogt (a founder of the web video site Justin.tv) announces their plans to make a retrofit kit that will adapt existing cars to do basic highway cruise, which is to say, staying in a lane and keeping pace behind other cars while under a driver’s supervision.
I’ve been following Cruise since its inception. This offering has many similarities to the plans of major car companies, but there are a few key differences:
- This is a startup, which can be more nimble than the large companies, and having no reputation to risk, can be bolder.
- They plan to make this as a retrofit kit for a moderate set of existing cars, rather than custom designing it to one car.
They’re so dedicated to the retrofit idea that the Audi A4 they are initially modifying does not even have drive-by-wire brakes like the commonly used hybrid cars. Their kit puts sensors on the roof, and puts a physical actuator on the brake and another physical actuator on the steering wheel — they don’t make use of the car’s own steering motor. They want a kit that can be applied to almost any car the market tells them to target.
They won’t do every car, though. All vendors have a strong incentive to only support cars they have given some solid testing to, so most plans don’t involve retrofit at all, and of course Google has now announced their plans to design a car from scratch. Early adopters may be keen on retrofit.
I rode in the car last week during a demo at Alemeda air station, a runway familiar to viewers of Mythbusters. There they set up a course of small orange cones, which are much easier to see than ordinary lane markings, so it’s hard to judge how well the car does on lane markings. It still has rough edges, to be sure, but they don’t plan to sell until next year. In the trial, due to insurance rules, it kept under 40mph, though it handled that speed fine, though drifted a bit in wider parts of the “lane.”
On top is an aerodynamic case around a sensor pack which is based on stereo cameras and radar from Delphi. Inside is just a single button in the center arm console to enable and disable cruise mode. You take the car to the lane and push the button.
All stuff we’ve seen before, and not as far along, but the one key difference — being a nimble startup — may make all the difference. Only early adopters will pay the $10,000 for a product where you must (at least for now) still watch the road, but that may be all that is needed.