Enhanced Localization for Bipedal Humanoid Robots in Dynamic and Noisy Environments

Abstract

Localization is a critical capability for bipedal humanoid robots operating in dynamic environments. This paper proposes a enhanced localization algorithm designed for humanoid robots performing tasks in dynamic and noisy environments, such as soccer fields. The proposed localization algorithm employs an enhanced particle filter, incorporating adaptive particle distribution based on reliability and advanced keypoint recognition techniques, to address issues of sparse landmarks, environmental disturbances, and sensor noise from bipedal motion with slip. The proposed algorithm has been evaluated in RoboCup-inspired scenarios with varying levels of disturbances, including obstacles, dynamic tasks, and visual obstructions. Experimental results demonstrate the algorithm's enhancement, achieving high localization accuracy and stability under diverse conditions.