In combination with free space estimation, a new algorithm from Baselabs (Chemnitz, Germany) enables vehicles to automatically approach parking spaces and park independently. This functionality matches with the integrated detection of dynamic and static objects in the Dynamic Grid.
Using automated parking functions in the public domain imposes strict environmental perception requirements. Static objects like curbs and parked vehicles need to be detected. However, detecting dynamic objects like pedestrians, bicycles, and other vehicles is even more critical. In addition to these pre-known objects, the systems must also be able to reliably detect pre-unknown moving objects in order to avoid accidents. The Dynamic Grid detects and tracks objects of arbitrary shapes without requiring extensive training. In addition, it determines the direction of movement and the velocity of dynamic objects. The sensor data from radars, lidars, or cameras with semantic segmentation is processed in a single integrated step in real-time. Thus, it avoids inconsistencies in the environmental model that typically occur with the classic combination of two distinct methods. The algorithm runs on automotive CPUs in real-time and is implemented according to ISO26262, Baselabs says.
“The Dynamic Grid has three key advantages that apply to automated parking as well: a high detection rate, a low false alarm rate, and it runs on production CPUs”, says Norman Mattern, Head of Product Development at Baselabs.
“With the Dynamic Grid as a software library, our customers gain very fast and cost-effective access to a sensor fusion technology that is suitable for series production for tomorrow’s driving and parking functions,” comments Robin Schubert, Managing Director of Baselabs.