Automotive -

Autonomous Valet Parking

The SLAM challenge

Autonomous valet parking (AVP) is set to deliver great benefits to the driver, but also operators of parking infrastructures: less time spent parking, fewer scratches, better use of space and potential value-add services such as an automated drive-through carwash while the car is in the lot or garage.

Yet, parking garages are quite different to the road scenarios that many autonomous driving and ADAS systems are built for. There are no HD Maps of parking garages as they are often entirely indoor with no distinct features and they can even span multiple stories.

How ArtiSLAM can be applied

Artisense's direct visual SLAM approach can cope with the repeated infrastructure within parking garages, unlike SLAM methods that rely primarily on features. The same information from the vehicle's camera that is used for localization can also be used to create or update the map of the garage. Users would only need to drive the car once before it is able to drive itself autonomously. The generated map can also be shared with other vehicles. Combining relocalization in the map with direct visual inertial odometry (VIO) allows reliable tracking across multiple levels even without distinct features.

How Clients Benefit

• Global pose indoors: Map creation with GNSS fused give clients the possibility of obtaining a position in global coordinates even within the parking garage. At the same time, the map acts as an extension of the HD maps in global coordinates.
• Cope with repetitive sceneries: Even in tough, similar-looking environments, ArtiSLAM's direct camera-based SLAM is robust enough to ensure clients have the accurate localization that is needed to find the right parking spot.