Research on Autonomous Navigation and Dynamic Path Planning Control System for Robots in Unknown Environments

Abstract

This paper presents an integrated autonomous navigation system for indoor mobile robots operating in unknown environments, combining adaptive PID and model predictive control (MPC). The proposed system incorporates three key innovations: a multi-sensor fusion framework integrating LiDAR, vision, and IMU data for robust environmental perception; an improved FastSLAM 2.0 algorithm achieving 3.5 cm mapping accuracy; and a hybrid dynamic path planning method combining enhanced $D^{*}$ Lite with time-varying RRT. The system architecture features a four-wheel differential robot model with adaptive control parameters that dynamically adjust based on environmental complexity. Extensive experiments across various indoor environments demonstrate superior performance, achieving a 2.83 cm average tracking error, 96.5% navigation success rate, and 41.2% faster response compared to traditional control methods. The system maintains 92.1% success rate in high-density environments (0.3 people $/ \mathbf{m}^{2}$) and demonstrates a 23.7% overall performance improvement over existing commercial solutions, particularly excelling in obstacle-dense scenarios with 36.5% reduced replanning frequency.