Control system design for robots used in simulating dynamic force and moment interaction in virtual reality applications

This dissertation presents an approach to simulating the dynamic force and moment interaction between a human and a virtual object using a robotic manipulator as the force transmitter. Accurate control of the linear and angular accelerations of the robot end effector is required in order for the correct forces and moments to be imparted on a human operating in a computer generated virtual environment. A control system has been designed which is robust in terms of stability and performance. This control system is derived from abbreviated linear and nonlinear models of the manipulator dynamics which are efficient enough for real-time implementation yet retain a sufficient level of complexity for accurate calculations. An efficient multiple-input multiple-output (MIMO) pole placement scheme has also been devised which locates the pre-specified system eigenvalues. The controller gains are given as explicit functions of a desired trajectory to be followed and, thus, are time varying such that the overall closed-loop system is rendered time-invariant. Key software elements were automatically derived and output in compiler-ready form demonstrating the feasibility of automatic, computer generated control laws for complex robotic systems. Test results are given for a PUMA 560 used to impart dynamic forces on a user operating in a virtual environment.