Nonlinear position/force control of robot manipulators

This dissertation focuses on the development of nonlinear controllers for robot manipulators under constrained motion, i.e., robotic systems that require the simultaneous control of the robot's end-effector position and interaction force with the environmental constraint. The primary objective was to extend previous nonlinear control design and analysis tools developed for unconstrained robots to constrained robot systems. This extension is not obvious due to the structure of the nonlinear differential-algebraic equations that describe the dynamics of constrained robot systems. Specifically, we develop new nonlinear position/force control algorithms that compensate for parametric uncertainty and/or require minimal number of sensors for implementation. While the compensation of unknown parameters is accomplished by utilizing a Desired Compensation Adaptation Law (DCAL) like control structure, the reduction in the sensor count is achieved by (i) designing a model-based observer or a high-pass filter to estimate the robot's link velocity signals and (ii) using an open-loop force controller to eliminate the need to measure the robot's end-effector force.;We first introduce the DCAL controller by presenting a reexamination of the original DCAL design/analysis scheme proposed by Sadegh and Horowitz for the robot position control problem. By exploiting the nonlinear damping argument and using a Lyapunov-like stability analysis, we simplify the stability analysis of a DCAL-like controller with nonlinear feedback and illustrate how the controller with linear feedback can achieve a semi-global stability result. Furthermore, we show that proposed control structure can be easily utilized to examine some non-adaptive control extensions.;A DCAL-like control structure coupled with a filter is then used to design an adaptive position/force tracking controller for constrained robots without link velocity measurements. The controller, which is implemented measuring link position and end-effector contact force, is shown to drive both the position and force tracking errors to zero asymptotically fast.;Finally, by assuming exact knowledge of the system parameters, we show how a position/force tracking controller can be developed based solely on link position measurements. To eliminate the need for velocity measurements, two different approaches are used: a model-based observer and a filter. The estimated velocity signals are then used in the design of two different position/open-loop force controllers which drive the position and force tracking errors to zero exponentially fast.