Gaming meets Parking: Testing for automated parking functions

Parking assistance functions such as Park Assist Plus already support customers today when maneuvering into and out of a garage or parking space, based on sensor data. "However, the trend is evolving towards using map-based data for parking functions, similar to driving functions”, explains Christian Feist, Product Manager for Map-Based Parking at CARIAD. This can enable assistance on the way to the parking spot: the vehicle moves through the parking lot and parks automatically at the end.

In the concept development for next-generation automated parking functions, Christian and his team encountered a problem: Existing editors to create virtual environments for simulation training were designed for highways and roads and therefore did not meet the requirements of parking environments. They needed a different solution to train their models. To create a new editor to match the needs for parking functions, the team had the idea to approach a gaming studio. Simulation games are quite common in the gaming industry, especially for the mobility sector: From farming simulators to truck simulators to flight simulators – gaming studios simulate nearly everything. Why shouldn’t they be able to simulate parking facilities, too?  

Speed bumps, boundary posts and fences can be installed in the editor, or the traffic routing can be defined: single-lane, two-lane or with a roundabout. "There are so-called garage regulations in many countries, but within the limits of these regulations, many things are still possible. If we wanted to test all of that in the real world, we would spend a long time searching for these places first."

Once the parking space has been created in the editor, it is exported to a digital map format. This map format is compatible with the Navigation Data Standard (NDS) and can be processed by a prototype parking function. But there are also pedestrians in multi-storey car parks, vehicles face narrow corners, many slopes and curbs and have to drive slowly, as well as forward and backward or brake immediately as soon as an obstacle appears. Different situations like these are modeled in scenario editors, which also use the map data exported from the Parking Garage Editor. "Theoretically, we can run a function in millions of scenarios in a wide variety of parking facilities overnight and test its performance," says Christian.