Dynamic behaviors on the NAO robot with closed-loop whole body operational space control

摘要

Exploiting full-body dynamics in feedback control can enhance the balancing capability of a legged system using various techniques such as Whole-Body Control (WBC) or Centroidal Momentum control. However, motion control of the NAO robot based on full-body dynamics has not been extensively studied due to its limited computation power, limited sensors, and restricted access to its low-level controllers. Whole-Body Operational Space Control (WBOSC) is a promising WBC approach for NAO, since its closed form solution provides computational efficiency. But, users need to provide the velocity map (Jacobian) between operational space and configuration space to add the balancing control task. Thus, in this paper, we formulate the Jacobians incorporating the Capture Point (CP) technique [1] and the Centroidal Angular Momentum (CAM) [2], [3], and demonstrate the enhancement of balancing capability in a physics-based simulation. While WBOSC reduces the computational load, implementing WBC in the real system with limited sensing capability and built-in joint position control is challenging. We show that the combination of a virtual model as an interface to the real robot and an Extended Kalman-filter based orientation estimator results in a stable implementation of a closed-loop WBOSC. We demonstrate the validity of our approach by performing a dynamic kicking motion on the physical NAO robot. Overall, the contributions of this paper are: (1) to extend WBOSC by adding CAM and CP control tasks, and (2) to implement WBOSC in a restricted physical system by utilizing a virtual model and an orientation estimator.