Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Mechanical Engineering


Dr. Hoda A. ElMaraghy


Excitation of lightly damped drive system resonances severely limits the performance of most current robotic manipulators. Also, many industrial applications require the control of the contact forces in order to execute contact tasks properly. This dissertation discusses the development and experimental verification of advanced high performance control algorithms in order to control the position and contact force of nonlinear manipulators with flexibilities which reside in the drive systems. An experimental two-link planar manipulator has been designed and constructed to test the control algorithms. The developed manipulator exhibits damped drive system resonances and highly nonlinear and coupled dynamics.

Several techniques for controlling the position and force of flexible joint robot manipulators are developed and evaluated. A collocated Proportional-Derivative controller is implemented to demonstrate the limitation of such a controller which is commonly used for industrial robots. Then, a fourth order feedback linearizable model is constructed and analyzed. A feedback linearization controller is designed, simulated and implemented in the joint space as a preliminary investigation of the feedback linearization performance. Results proved a tracking performance superior to that of the PD controller and support applying feedback techniques for position and force control.

Two model-based position and force control approaches based on feedback linearization concept are developed, simulated and implemented. The dynamic hybrid controller deals with the contact with rigid environments while the impedance control deals with performing free and/or contact with compliant environment motions. The results indicate excellent performance in tracking the position and regulating the contact force. These two approaches serve as basic structure for the design of robust and adaptive controllers.

A robust sliding mode controller is designed, simulated and implemented to maintain the model-based tracking performance in the presence of payload variations and other parametric uncertainty. The problem is formulated by using the feedback linearizable model and the constraint imposed by the contact environment, to characterize the disturbance caused by parametric uncertainty in the system. In addition, the effect of parametric uncertainty on computing the unmeasured state elements is included in the analysis and the design of the robust controller. Results indicate that the controller obtains an excellent tracking performance in the presence of parametric uncertainty. This controller solves the problem of the necessity of the exact state of the feedback linearizable system for feedback. In addition, a specific relation between the uncertainty bounds and the achievable accuracy is derived. Another alternative to deal with parametric uncertainty is an adaptive cascade controller which is developed, simulated and implemented. It consists of a direct adaptive controller for the rigid dynamics and a Model Reference Adaptive Controller (MRAC) for the flexible dynamics. Results indicate good force tracking with slow position response. The cascade approach requires extremely fast actuator to achieve the feedback linearization results. A robust sliding mode observer is developed in order to estimate the unmeasured state elements of the feedback linearizable system in case of parametric uncertainty. This observer is useful for the successful application of adaptive feedback linearization algorithms. An adaptive feedback linearization control algorithm combined with the sliding mode observer is developed and simulated. The formulation enables designing adaptive laws using the original system state and the angular accelerations instead of the state of the feedback linearizable system and characterizing the stability of the overall system. The simulation results show the ability of the control algorithm to achieve position and force tracking by robust estimation of the state vector and updating of the robot parameters.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."