Konferens: FAST-zero'15: 3rd International symposium on future active safety technology toward zero traffic accidents, 2015, Gothenburg

A new method for generating a separate two-dimensional free space grid map for ADAS based on data from radar sensors is presented in this paper. We introduce a free space model based on an inverse sensor model to compute the Gaussian-based free space probability for each cell of the free space grid map. A Bayesian approach is used for free space probability estimation independently from the occupancy probability, which enables increased amount of information for environment description. For this purpose, two free space grid maps are generated: The instantaneous free space map is generated in each measuring cycle and the accumulated free space map is generated once and updated in each measuring cycle. We describe how the free space grid maps are generated and updated by new observations. In contrast to other approaches, the detection accuracy is taken into account in the free space model. Finally we present the experimental results obtained from real world environments.