Push Recovery Methods Based on Admittance Control Strategies for a NAO-H25 Humanoid

摘要

This paper discusses two main investigations regarding push recovery of a NAO humanoid robot. The first part indicates the simulation results of different strategies retrieved from human behaviors, namely: Ankle, Hip, and Hip-ankle strategies. These methods are separately assessed and their efficiencies are discussed. To the end of converting torque parameter to robot’s movement with a reasonable approximation in the aforementioned strategies’ obtained models, a new approach, the so-called Virtual leg model, is introduced. The second discussion assesses the admittance control strategy, which relies on disturbances’ energy absorption. Individual control algorithms are introduced for each disturbances’ categories based on NAO’s real conditions. This paper proposes a new strategy, which investigates the influence of admittance control coefficients against external disturbances, by considering a virtual interaction between robot’s feet and the ground, and introducing a Kalman-filter-based estimator. All proposed methods are simulated in NAO-H25’s simulation environment.