EDGE Method for Identification and Prediction of Dynamic Object Trajectories for a Mobile Robot Based on YOLO26n

Abstract

This paper proposes an EDGE method for the identification and trajectory prediction of dynamic objects for a mobile robot based on YOLO26n, which integrates local object detection, estimation of object positions and motion parameters, short-term trajectory prediction, and assessment of potential collision risk. The numerical simulation results confirmed the operability of the proposed method: the successful detection rate reached 96.68%, the position estimation RMSE after filtering was 0.0470m, and the trajectory prediction RMSE for a 2.0s prediction horizon was 0.3193m. The obtained results demonstrate the potential of the proposed method for local analysis of dynamic environments and early detection of potentially hazardous interactions between a mobile robot and moving objects