Research on Mobile Robot Localization Method Based on Adaptive Motion Model and Double-Threshold Relocation Strategy

To address the problems of particle degradation, localization failure, and sensor noise in mobile robot localization algorithms, robust Monte Carlo localization (RMCL) is proposed in this article. It is an enhanced approach combining an adaptive odometry motion model and adaptive relocation detection. The approach incorporates velocity factors and environmental variation factors to optimize the particle filter model. Then, the result is used as the initial value, and a precise result is obtained based on scan matching. The accurately matched result is then reintroduced into the particle filter as particles. Finally, the decision for relocalization is made based on the results of precise matching and the particle distribution. The simulated experiments were conducted to evaluate the performance of our algorithm. The real-world experiments are also conducted employing a submillimeter high-precision instrument. The experimental results demonstrate that compared with the mainstream 2-D laser-based localization methods, the accuracy of our algorithm is improved by more than 29.97%, and the failure rate of positioning in the real environment is below 10%. Thus, our algorithm substantially enhances accuracy for complex indoor positioning.