The thesis focuses on several key issues associated with modeling, analysis, and advanced control of a class of multi-module manipulator systems, which can vary their geometric configuration to quickly adapt to the environment and the requirements of a task. The studied manipulator system has a combination of revolute and prismatic joints, which lends itself to distinct advantages;​ for example, reduced dynamic coupling effects and singularities, simpler kinematics, and ease of obstacle avoidance , for a specific number of degrees of freedom.

In rigorous and hazardous environments, autonomous operation is preferred. A reliable and effective control approach is essential to properly carry out a task (e.g., trajectory tracking in parts joining, inspection and material application;​ satellite capturing for retrieval and repair) in an autonomous manner. In the present work, several controllers have been formulated, analyzed, designed, simulated, and tested. Specifically, linear quadratic regulator (LQR), feedback linearization technique (FLT), computed torque adaptive control, and neural network adaptive control, have been studied. Most of them are model-based controllers. In view of this and for use of computer simulations related to illustration and evaluation of the techniques developed in the thesis, complete dynamic formulations of ground-based and space-based models of the robotic system are presented in the thesis. The resulting coupled and nonlinear models may be incomplete and may have uncertainties in parameters. The neural-network adaptive controller developed here is able to accommodate these shortcomings. Details of this controller are presented and the performance is evaluated using several case studies.

The various control methods studied in the present research have distinct advantages and disadvantages, which make them perform satisfactorily under some conditions while deteriorating under some other conditions. As a result it is desirable to have a supervisory system to monitor the performance of the robotic system and then switch on an appropriate control scheme depending on the system behavior. When the system operates in different modes of operation or carries out several tasks, then, it would be possible to optimize its performance through controller switching. A new control scheme termed Supervisory Control Switching System (SCSS) has been developed and evaluated during the concluding stage of the present investigation. It is capable of selecting the most suitable controller for a particular task or a situation, from several separately designed controllers. The present study lays a sound foundation for further exploration of this class of novel and useful manipulators.