Modeling, Identification, and Control of Brushless Motors on Humanoid Robots.

A common approach for controlling humanoid robots is based on a hierarchical architecture composed of three different layers called: trajectory optimization, simplified model control, and whole-body quadratic programming (QP) control. When the whole-body QP control layer generates joint torques references, a reliable low-level joint torque control layer becomes pivotal. The humanoid robot iCub is actuated by permanent magnet synchronous motors (PMSM), thus the joint torque control inner layer should be composed of two nested control loops: a current and a torque control loop. This thesis aims to improve the overall torque control architecture of the humanoid robot iCub, by designing and implementing a low-level current vector controller as part of the fourth layer of the aforementioned architecture, that guarantees good current tracking performances. Firstly, an identification algorithm has been developed and tested for motor parameters estimation. Then, a current controller has been designed and implemented by exploiting the feedback linearization and the field-oriented control approaches. Experiments in the Simulink environment and on the knee motor group of the iCub humanoid robot validate the proposed controller.