Torque feedback and saturation compensation for motion control of an indirect-drive unit using a harmonic drive

This dissertation is concerned with the motion control of Indirect-Drive (ID) robots with Harmonic Drives (HD). Modelling, analysis, simulations, and experiments are conducted for a test bed consisting of a single-joint ID unit with a HD. While ID units with HD possess advantages such as the ability to use small actuators to drive large payloads, they introduce several issues that must be properly addressed in the design of motion control systems. The issues include the nonlinear joint flexibility and nonlinear friction introduced by the HD mechanism, and the fact that the HD unit has a torque limit which must not be exceeded. In addition to these issues, actuator saturation is also considered. In this dissertation, two approaches are explored to address these issues.; The first is a practical approach where the issues are addressed almost independently of one another. For friction compensation, Coulomb friction is assumed to be the dominant HD nonlinear friction. Since compensation of Coulomb friction requires knowledge of motor velocity, a Lyapunov-based tuning guideline to estimate motor velocity is derived. Joint flexibility and HD torque limit issues are coupled. When joint vibration causes the HD torque limit to be exceeded, a virtual actuator saturation limit may need to be imposed. Vibration attenuation can make this virtual actuator saturation limit less restrictive. It is proposed to feed back the derivative of joint torque to attenuate joint vibration. This method is shown to be robust with respect to plant parameter uncertainties. The nonlinear joint flexibility can be characterized as a hardening spring. As a result, the pole and zero locations of the linearized model may vary under different operating conditions. This variation can be reduced when the derivative of joint torque is used for feedback. Having reduced some of the nonlinear effects, a tracking controller can be designed to achieve the tracking objective. To avoid a conservative tracking controller design with respect to the actual and virtual saturation limits, a compensation method called Linear Conditioning (LC) is used to address saturation.; The second approach is an a priori saturation compensation based on an extension of Artstein's theorem. There are two possible cases. The first is when the actuator cannot make the HD torque to exceed its limit. In this case, the extension of Artstein's theorem can be directly applied to the whole system. The second case, which is encountered in the experimental setup, is when the actuator can make the HD torque to exceed its limit. To deal with the second case, the extension of Artstein's theorem is applied to a part of the system. This determines the joint torque trajectory such that motion tracking is achieved while at the same time, the HD torque limit is not exceeded. Using backstepping, the required actuator torque is computed such that the joint torque follows this joint torque trajectory. To provide robustness with respect to plant parameter uncertainties, nonlinear damping is introduced.