Mercedes "Automated Valet Parking" disappoints


Promised for years, you can finally do automated valet parking if you have a 2021 Mercedes S class and park in one garage at Stuttgart airport. First demonstrated at Stanford in 2009, this feature is long overdue, and this implementation is quite disappointing, doing little more than save the driver a few minutes of walking.

I go into the details of what robotic parking could and should do, even today in a new article on at:

Mercedes "Automated Valet Parking" disappoints


We cannot agree that Autonomous Valet Parking was in the "solved problems" column ten years ago. There is a drastic difference between an academic demo and production ready technology. The academic demo must work 1% of the time, however in production, 99% is not enough.

There is also a question of cost, demos from years ago use lidars, sensors that are prohibitively expensive, and, as the narrator in the Stanford demo video admits, have the back of the car lined with computers. Even cheaper camera based solutions available today, still command high increased costs in sensors and computational resources, which are only acceptable for top end vehicles with larger profit margins. While true that Waymo and other self-driving car companies have solutions in a very advanced stage, this does not automatically present a viable business model to accelerate the technology.

The legal side must also be taken into consideration, things are only now starting to get clearer, but laws regarding this technology still vary drastically in each country. Even in the confines of Stuttgart's airport, there are circulation and safety rules, not to mention deals in place with parking providers and taxi companies that must be adhered to.

To conclude, we agree with the author that the greater the technology, the greater the advantages. The difference between Mercedes/Bosch demo and the more useful drop off at the curb, is mainly due to regulation and current infrastructure. What is stopping widespread use is the relatively high cost of a reliable solution, which is what Parkopedia is aiming to resolve. Parkopedia is creating high-resolution indoor parking maps that can be used by autonomous vehicles without a lidar to localise and navigate indoors without a GPS signal and do so at a fraction of the cost of Bosch’s solution which requires a significant capital investment for every parking facility. You can find a demonstration of autonomous parking using Parkopedia’s indoor maps on and

I don't put it in the solved column in 2009! However, it's possible to do a lot more than the Mercedes does. It's possible to need no special hardware in the garage, just a map is enough, to show where the valet section is and how to get there, maybe a server that counts the cars going in and out to tell you it's full.

The law is much less of a barrier on private land. There it's mostly a question of safety and liability.

My disappointment is they could have released a product that actually did something useful, like giving you parking for 1/2 price or increasing the capacity of a parking garage (important at an airport this weekend.) Instead they did what you might be correct to call a demonstration, rather than a product as they are calling it. But a demonstration was possible some time ago (not 2009.)

Add new comment