Robust locomotion control of legged robots via graph optimization-based disturbance estimation

Abstract

As the performance of legged robots has increased dramatically, various studies have been conducted for different purposes. These robots are not only used for exploring rough terrains but also for performing tasks using manipulators. In such operations, unexpected external forces can be encountered, potentially degrading the system's stability. To address these issues, this study proposes methods for estimating disturbances and an adaptive controller. The disturbance estimator does not relying on force sensors at the each foot, and a factor graph optimization formula is constructed using sensors such as cameras and IMUs to estimate the external force/torque. The obtained force and torque can also determine the point of force application. The adaptive controller, utilizing the estimated disturbances, employs Model Predictive Control to calculate the proper foot placement. It also determines the optimal gait types and the duration of contact with the ground, based on the direction and magnitude of these forces. Finally, the methodologies suggested in this study are expected to significantly enhance task performance and stability, even during interactions with external environments.