Filtering and prediction techniques in radar tracking of vehicles for assisted driving

An advanced system of driving assistance is considered, which is based on a network of radar (and, in some situations, infrared) sensors alongside the motorway. The system is capable of tracking each single vehicle in the coverage area individually and to transmit the picture of vehicle positions and velocities to a control center in real time. The latter processes the received data, in order to detect possible dangerous situations or anomalies in the traffic flow and to correlate them to specific vehicles (slower or stopping vehicles ahead, queues, and the like), and it retransmits the appropriate pieces of information to receiving devices on-board the vehicles, through a dedicated digital radio channel. In this way, drivers can be advised by the system on the road situation in front of them, and become aware of danger conditions in real time, even in the case of scarce or almost null visibility. The paper presents the characteristics of the tracking and prediction mechanisms used for updating the traffic picture at the control center, as well as to reduce the flow of information that is fed back to the on-board devices.