ALEL-AMCL: an artificial landmark enhanced localization of mobile robotic system based on AMCL algorithm

摘要

Abstract In factory environments, stable and high-precision localization capabilities are crucial for the automated guided vehicles. Adaptive Monte Carlo localization (AMCL) is widely used in 2D laser navigation due to its excellent performance in addressing global localization and position tracking issues in localization. However, AMCL faces challenges in factory areas where laser degeneracy and dynamic changes occur, the particles can diverge due to observational uncertainty, leading to severe localization errors. We propose a method artificial landmark enhanced localization AMCL (ALEL-AMCL) that involves pre-positioning reflector posts only in degraded and dynamically changing scenes. In the observation update step of particle weight calculation, the weight of the artificial landmarks observations is incorporated, reducing the impact of laser degeneracy environments and changing scenes on robot localization. Multiple global localization experiments and end-to-end pose tracking tests were conducted in areas with laser degeneracy and dynamic changes. The results show that the ALEL-AMCL algorithm maintains localization errors around 0.015 m and 0.5 ∘ . In large-scale environment stability tests, ALEL-AMCL runs stably even in areas without artificial landmarks, showing no drift like that seen with extended Kalman filter. Compared to the AMCL algorithm, it reduces the average absolute pose error in the translational direction by 60.49% and the root mean squared error in the translational direction decreased by 78.89%.