Autonomous Human Tracking using Vision-LiDAR Fusion with Model Predictive Control-Based Trajectory Following

In this paper, a golf cart with improved autonomous driving capabilities has been used to construct a real-time human tracking application. The proposed tracking method is based on a human detection module that combines camera and LiDAR data to improve environmental perception capabilities using the YOLOv7 algorithm with sensor fusion. Following maximum accuracy human detection, the instantaneous position of the detected person is determined as the target, and optimal path following toward this target is achieved using the Model Predictive Control (MPC) algorithm. The MPC algorithm predict the required steering angle in real-time to reach the detected person by considering the vehicle’s current state and dynamic constraints, while the speed value is calculated according to the remaining distance to the target. This method allows the car to track the detected person without requiring any external intervention. MPC-based tracking mechanisms in autonomous vehicle technologies are proven to be effective by the system created in this study, which exhibits reliable and secure human tracking in dynamic situations.