Research on Object Detection and Tracking Methods for aLow-Speed Mobile Platform

Enhancing the positioning stability and accuracy of autonomous following systems poses a significant challenge, particularly in dynamic indoor environments susceptible to occlusion and interference. This paper proposes an innovative approach that integrates Ultra-Wideband (UWB) technology with computer vision-based gait analysis to overcome these limitations. First, a low-power, high-update-rate UWB positioning network is established based on an optimized Double-Sided Two-Way Ranging (DS-TWR) protocol. To compensate for UWB's deficiencies under Non-Line-of-Sight (NLOS) conditions, a visual gait recognition process utilizing the GaitPart framework is introduced for target identification and relative motion estimation. Subsequently, an Extended Kalman Filter (EKF) is developed to seamlessly fuse absolute UWB measurements with gait-based relative kinematic information, thereby generating precise and robust estimates of the leader's trajectory. This estimated path is tracked by a differentially driven mobile platform via a Model Predictive Controller (MPC). Experimental results demonstrate that the tracking deviation for most trajectory points remains within 50 mm, with a maximum observed deviation of 115 mm during turns, confirming its strong robustness and practical utility in real-world intelligent vehicle applications.