Evaluation of sensor technologies and algorithms to detect vulnerable road users (VRUs) by utilizing sensor data of a passenger car and transmission of appropriate Collective Perception Messages (CPM) using the Vehicle-To-Everything (V2X) technology

Agricultural tractors (ATs) on public roads pose elevated safety risks due to their size, mass, and limited sensing capabilities. Although AT-involved crashes are relatively rare, their severity is disproportionately high. This thesis investigates a multi-sensor fusion approach leveraging cameras, LiDAR, and radar on intelligent connected vehicles to enhance early detection of vulnerable road users (VRUs) and other traffic participants. By integrating these sensors, the aim is to improve situational awareness and reduce collision risks. Evaluations indicate that while the fused system can reliably detect vehicles in simple scenarios, it underperforms in complex conditions and fails to meet VRU detection expectations. Furthermore, the current setup operates offline and processes only one fused object at a time, constraining real-world applicability. Future work will focus on deploying the algorithm in operational environments, refining test scenarios, incorporating 3D vision via stereo or depth cameras, and expanding the system’s capacity to simultaneously track multiple objects to achieve more robust and comprehensive safety enhancements.