Germany-based Baselabs, a sensor fusion specialist, has released Dynamic Grid, an algorithm that generates a consistent environmental model from high-resolution raw sensor data. The company claims that this helps accelerate the development of data fusion systems for autonomous driving capabilities, especially in difficult urban environments. By skipping time-consuming algorithm training, car developers can develop driver assistance systems such as parking and congestion pilots with better performance than traditional tracking and grid methods.
Urban autonomous driving capabilities present very high requirements for the environmental model used. On the sensor side, the industry is moving towards using high resolution sensors to get the data they need at a sufficient level of detail. Traditional algorithmic techniques for sensor fusion reach their limits in these situations. Dynamic grids, on the other hand, can process high-resolution sensor data from devices such as radar and laser scanners at the raw data level. It is also possible to use a camera with semantic segmentation. As a result, the system provides a self-consistent environmental model that accurately and robustly detects dynamic and static objects in the vehicle environment. In addition, the algorithm estimates free space to identify driving areas or parking spaces. This algorithm runs in real time on the CPU of the car and is implemented according to ISO 26262.
https://www.automotivetestingtechnologyinternational.com/news/new-sensor-fusion-algorithm-for-urban-driving-scenarios.html New sensor fusion algorithm for urban driving scenarios