Cooperative Intention Detection using Machine Learning

Cyclists will play an essential role in future traffic. It is crucial to detect cyclists and predict their intentions to avoid accidents and achieve a highly efficient traffic flow. Vehicles equipped with sensors, data processing systems, and communication abilities acquire and maintain a local model of their surrounding traffic environment, e.g., crossing cyclists. Cooperating and interacting vehicles, roadside units, and cyclists equipped with smart devices and other body-worn sensors exchange information. They form a multi-modal sensor system with the goal to reliably detect cyclists and their intentions under consideration of real-time requirements and uncertainties. The resulting model allows extending the perceptual horizon of individual road users beyond their sensory capabilities, enabling improved and more reliable cyclist intention detection. In this thesis, a holistic approach for detecting the intentions of cyclists using cooperative methods is presented.