Clustering of Automotive Radar Data in Multi-Sensor Networks

Future advanced driver assistant systems not only will use information of multiple sensor types, but also employ multiple sensors of the same kind. This will be especially true in the case of automotive radar sensors, which when combined into a radar sensor network, allow for extended features like 360 degree coverage around the ego vehicle or extracting additional information of the other vehicles in the scenario.

For these extended features to be employed, it is neccessary to correctly group together radar target points belonging to a single extended object, like a vehicle, by employing sophisticated clustering algorithms. In this thesis, new clustering apporaches for the usage in a radar sensor network are to be developed, which not only consider the range, angle and velocity information of the single target points, but also takes into account the configuration of the sensor in the network as well as the motion of extended objects. The clustering algorithm will then be validated with real-life measurement data.