|M.Sc Student||Nusbaum Uriel|
|Subject||Path Design and High-Level Control of Redundant Manipulators|
|Department||Department of Autonomous Systems and Robotics||Supervisors||Professor Yoram Halevi|
|Ms. Miri Weiss Cohen|
Redundant machines are a common type of mechanisms in which the amount of available degrees of freedom (DOF) for the control system is larger than needed for task execution. This property poses difficulties since the problem of inverse kinematics have infinite number of solutions. However, redundancy also gives rise to the possibility of performance enhancement through utilizing the redundant DOF for extra tasks using various optimization methods. In this work redundancy will be utilized to minimize the energy consumption of the system while in some cases an additional task of tracking will also be required and achieved. Optimal control theory has been extensively used for optimization of dynamic systems, however complex tasks and redundancy makes these problems computationally expensive, numerically difficult to solve and in many cases ill-defined. Herein, this problem will be dealt by using evolutionary bi-level optimization which will perform decision vector optimization, optimal control problem solution derivation and computational load reduction. In addition, the proposed algorithm allows optimization of complex tasks usually unsolvable in practice using standard dynamic optimization tools. This is done by using a GA as the upper level optimization technique and optimal control algorithm as the low level one. The problem of global energy minimization will be presented and its natural relation to classic optimal control will be shown. Illustrative examples of a redundant system dynamic optimization will also be presented, solved and discussed.