An Experimental Demonstration of Linear Quadratic Regulator Design Applied to Manipulator Path Tracking

This paper demonstrates the application of Linear Quadratic Regulator (LQR) design to path tracking control of a planar two link manipulator. It has been shown that established linear control design techniques can be effective even as the plant varies from the design model. In the path tracking task, the modeled nonlinear effects were included in a feedforward torque command that was computed to realize a specified nominal trajectory. LQR gains were generated from linear perturbational models at set points along this trajectory and then spline fitted to form a smoothly time-varying perturbation feedback control law. Both simulation and hardware results are presented which show good trajectory following performance, even in the presence of significant friction.