An Adaptive Path Planning Method for Indoor and Outdoor Integrated Navigation
Yun Hong, Xiaodong Zhang, Xiaonan Wang, Hongye Wang, Hexiao Wang, Haonan Wang · 2025
This paper addresses the demand for indoor-outdoor integrated navigation and proposes a set of adaptive path planning methods. Firstly, by fusing visual and laser data, high-precision outdoor model reconstruction and indoor map creation are accomplished respectively through RealityCapture software and the FAST-LIVO2 algorithm, and an indoor map data model suitable for navigation is constructed in accordance with national standards. Secondly, a seamless positioning technology based on multi-source sensor fusion is proposed. By combining GNSS (RTK), IMU, UWB and geomagnetic-assisted tightly-coupled schemes, the problem of positioning continuity during the transition between indoor and outdoor environments is effectively solved. The positioning robustness is enhanced by optimizing the deployment of UWB anchor points and introducing the RANSAC algorithm. In terms of path planning, a hierarchical framework of “global A* + local dynamic window approach (DWA) + reinforcement learning fine-tuning” is adopted, combined with semantic cost functions and social force models, to achieve dynamic obstacle avoidance and scene-adaptive optimization. Experiments show that this method significantly reduces the collision rate in complex scenarios and can adapt to special scenarios such as peak hours on campus. The research in this paper provides technical support and practical references for indoor-outdoor integrated navigation.