BIG-STEP: Better Initialized State Estimator for Legged Robots with Fast and Robust Ground Segmentation

Abstract

Legged robots are crucial in various applications, from search and rescue operations to exploration missions in challenging terrain. Accurate estimation of the robot’s state is paramount for achieving precise and reliable navigation. However, estimating the state of legged robots presents unique challenges due to inherent uncertainties, dynamics, and environmental interactions. We propose a novel state estimator for legged robots that leverages the ground plane to mitigate errors in the z-component of the state estimate especially. By exploiting the information provided by the effectively estimated ground plane, which serves as a reliable reference, the proposed estimator compensates for errors and enhances the accuracy of the estimated state. To validate the effectiveness of the proposed state estimator, real-world experiments are conducted on a legged robot platform. The results demonstrate significant improvements in state estimation accuracy, particularly in the z-component when compared with conventional state estimation methods. The proposed state estimator can potentially enhance the performance and autonomy of legged robots in various applications, including locomotion control, terrain mapping, and environment perception. Furthermore, its robustness and accuracy make it well-suited for scenarios where precise state estimation is crucial for safe and effective operation.