3D-SportsNavNet: An Innovative Path Planning and Navigation System for Sports Supply Robots in Complex Dynamic Environments

Abstract

As sports equipment robots are increasingly applied in modern sports and outdoor activities, existing technologies face challenges in adaptability and real-time response when dealing with dynamic environments. To address these issues, this paper proposes 3D-SportsNavNet, an innovative path planning model for complex dynamic environments. The model integrates three key modules: multimodal environment perception and reconstruction, adaptive dynamic path planning, and intelligent navigation optimization. The main contributions include: (1) a novel multimodal fusion framework integrating RGB-D cameras and LiDAR with DCNNs and PointNet for real-time 3D reconstruction, (2) an adaptive planning strategy combining Deep Q-Learning and Proximal Policy Optimization for dynamic obstacle avoidance with 30Hz update frequency, and (3) a self-supervised learning mechanism enabling continuous optimization without extensive labeled data. Experimental validation across three diverse scenarios demonstrates that 3D-SportsNavNet achieves 93.7% path planning success rate, reduces collision incidents by over 50%, and decreases energy consumption by 9-13% compared to baseline methods (DWA, RRT, traditional DRL). The model provides an effective solution for sports equipment robots operating in complex dynamic environments.