Vision-Based Trajectory Planning for Cable-Constrained Mobile Platforms Using an Improved Dynamic Window Approach

To mitigate the risk of cable winding failures in cable-reeling vehicles (CRVs) used in mining environments, this paper presents a stereo vision sensor-driven perception–planning framework. Its core innovation is a cable-morphology-constrained trajectory planning method based on an improved Dynamic Window Approach (DWA). A binocular vision sensor is deployed to acquire synchronized image and point cloud data. A Rectangular Region Segmentation (RRS) algorithm is developed to process sensor-derived binary masks, achieving robust cable morphology estimation at 0.018 s per frame, with stable performance across up to 1200 segmentation regions. By integrating neural network-based perception with the improved DWA, proximal cable morphology constraints are incorporated into local trajectory evaluation to coordinate vehicle motion and cable geometry during obstacle avoidance. To address the inherent limitations of depth sensors in long-range measurement, a proximal cable morphology extraction strategy is proposed to provide more reliable cable geometry information for trajectory planning. Experimental results demonstrate stable system performance at 4–6 FPS under a movement speed of 0.2 m/s, with accurate heading maintenance under near-range cable constraints. These findings demonstrate the feasibility of integrating real-time cable morphology perception with cable-constrained trajectory planning for autonomous navigation of cable-powered mining equipment.