Improving the Accuracy of Path Tracking for Mobile Robots: Optimizing the Combination of RRT Algorithm and Pure Pursuit Control

摘要

Against the backdrop of Industry 4.0 transformation, this study addresses the core shortcomings of the Rapidly-exploring Random Tree (RRT) algorithm in high-dimensional multi-degree-of-freedom space planning—specifically, excessive path tortuosity and redundant nodes—by proposing an optimization strategy based on a “jump connection” mechanism. This strategy systematically eliminates redundant nodes, achieving an average path reduction rate of 18.4% (21.1% in standard 100px×100px maps and 27.2% in larger 200px×200px maps) while maintaining planning efficiency comparable to traditional RRT. A key innovation is the step-size-efficiency adaptive model: increasing the step size from 1.0px to 10.0px enhances planning efficiency by 740 times, reducing time from 934.14 s to 1.26 s, with optimized path lengths converging to small-step solutions(178. 7px-183.1 px). Path tracking verification demonstrates significant accuracy improvements: Pure Pursuit control reduces the error rate from 0.229% to 0.097% (a 57.6% decrease) at a preview distance of 2.0px, primarily due to eliminating 72% of abrupt curvature changes. Critically, PID control also shows consistent error rate reduction (e.g., from 0.210% to 0.090% at kp=3.0), attributed to minimized curvature variations and nonlinear disturbances. Comparative analysis reveals that the optimized RRT outperforms RRT* in time efficiency (e.g., 1.4s vs. 49.8s for similar paths) despite marginally longer path lengths. Implemented on an NVIDIA RTX 3090 platform with 90GB memory, the solution achieves sub-second responses (1.26s), offering 15-20% cycle time improvement potential for industrial applications such as automotive welding and precision assembly. However, limitations persist in dynamic obstacle environments and 3D space adaptation, necessitating future work on multi-algorithm integration and real-time obstacle avoidance heuristics.