The use of direct end point sensing/control methods can make the current robotic systems more flexible. However, robust operation of such robotic systems requires intelligence integration into the end point control. The main objective of this thesis is to provide advanced intelligent supervisory aids to the direct end point control at the execution level. Therefore a global system architecture is designed and the architecture and functions of its Supervisory Level are given in details. The design and implementation of the key supervisory modules comprise the focus of the rest of this thesis. These modules include:​ motion trajectory planner;​ grasp planner;​ and automatic feature selector, which address the major supervisory issues.

The proposed motion trajectory planner has the capability of real-time trajectory generation and deals effectively with local minima and kinematic singularities. Within motion trajectory planning scheme, novel local and global path planning methods are proposed. The automatic grasp planner (AGP) is CAD-based and integrates sensory constraints into (motion) grasp planning. The grasp and motion trajectory planners are also capable of generating end-point trajectories relative to the task object. Automatic feature selection is an important original contribution of this thesis. Feature selection constraints, indices and the algorithms for the calculation of indices are presented. A feature selection strategy is proposed and the implementation of a CAD-based automatic feature and window selector (AFS) is achieved. Finally 3D motion, grasp, and feature selection simulators are designed and used for the simulation of the individual modules. The simulations demonstrate the usefulness and the effectiveness of the proposed supervisory control system, in particular for the realization of the tasks related to relative pose based visual servoing.