Permanent Magnet Synchronous Motors: Nonlinear Dynamic Modeling, Hardware Characterization, and High-Bandwidth Torque Control for Applications in Dynamic Robotics

Abstract

The high-level control algorithms that are responsible for achieving dynamic locomotion in legged robots depend on accurate torque production for matching real-life performance with simulated performance. To achieve accurate torque production, actuators must run high-bandwidth, low-level torque control. Developing high performance low-level controllers requires accurate actuator models. This thesis covers the physical model of a Permanent Magnet Synchronous Motors (PMSM), a very common type of actuator in dynamic robotics. This thesis details the derivation of the PMSM linear model, how to adapt the model dependent on the physical construction of a real motor, and the implementation of FieldOriented Control (FOC) to achieve torque control. This thesis also describes a novel design of a high-precision dynamometer, which allows a motor to be coupled with an impedance and a torque sensor in order to accurately characterize the torque production characteristics of the motor. Using this dynamometer and other experimental setups, this thesis validates the model and determines parameters for multiple different actuators. Finally, this thesis proposes an augmented PMSM model that considers the nonlinear saturation behavior of the motor, validating the principle with hardware experiments, and demonstrates a nonlinear torque model and gain-scheduled current controller that improve torque tracking performance.