Development of a state estimator for legged robots with robust environmental adaptability

Abstract

The legged robots have recently attracted much attention because they can be operated in harsh environments where it is difficult for mobile robots to travel. In order to use such legged robots in various environments, it is essential to estimate the robot's state accurately. However, it is difficult for previous approaches to estimate the state accurately in environments where the robot is prone to slip, where lighting changes are severe, and where vision information is limited since the existing algorithms usually do not consider these harsh environments. Therefore, this study proposes a state estimation framework for stable and accurate state estimation of the legged robots in various environments. In particular, this study proposed vision-based algorithms to overcome the slippery environment and the environment where vision information is insufficient. The performance of each algorithm was proven through various experiments.