Human-following control of an autonomous agricultural vehicle using an adaptive navigation point time elastic band method

The application of autonomous navigation technology in agriculture is becoming increasingly widespread. However, the diversity of agricultural environments continues to pose challenges to achieving full autonomy. Using autonomous vehicles to assist with material transportation can reduce human labor demands and significantly improve operational efficiency. Currently, most agricultural vehicles operate in a fixed mode, simply following a person. However, this approach may introduce potential safety risks and limit the effectiveness of human-vehicle interaction. Moreover, generating a rational following path and ensuring strong obstacle-passing capability are critical requirements during material transportation. To address these challenges, we propose an Adaptive Navigation Point Time Elastic Band (ANP-TEB) method that enables adaptive mode switching between following and accompanying behaviors. Owing to its rolling-window velocity vector calculation, as well as its stopping and elastic following strategies, the proposed method outperforms the conventional TEB and other baseline methods in both simulation and field experiments. Person-detection results indicate that the re-matching module reduces the root mean square error of angle by 6.785° and distance by 0.115 m in simulation, and by 4.109° and 0.089 m, respectively, in real-world tests, demonstrating a clear improvement in agreement with the ground-truth data. Simulation results further indicate that the ANP-TEB method generates more rational trajectories during turning maneuvers, achieving an accompanying deviation of 0.137 m and a following deviation of 0.187 m during straight-line motion. In field experiments, the ANP-TEB method demonstrated effective mode-switching capability and was the only approach to successfully complete the material transportation task. It produced a rational trajectory with an average following distance of 2.005 m, which is 0.565 m shorter than that achieved by the conventional TEB method. Overall, these results demonstrate that the proposed ANP-TEB method enables adaptive mode switching, improves trajectory rationality, and enhances obstacle-passing capability, making it well-suited for human-following material transportation tasks in complex agricultural environments.