Upravljanje antropomorfnom robotskom rukom koristeći genetičke algoritme

Abstract

In this paper it is treated control of an anthropomorphic robot arm (ARA) with three of freedom. Analysis of multijoint coordination in humans is an important source of information for synthesis of dynamic patterns in machines. Here, it is stated and solved the Bernstein problem which is related to kinematic redundancy of ARA. One of approaches for solving previous problem is based on introducing a „cost function“ that is a presumably optimized during movements where the problem of coordinating multiple systems can be state as an optimization problem which is most likely to biological „principle of minimum interaction“. The solution is generated in a sequence of steps involving a heuristic techniques of genetic algorithm that provides reliable initial guesses. To solve local problems, the minimum principle is used where the multi-level univariate hierarchical strategies is proposed. The problem is divided into two-level optimization problem which is solved iteratively until the desired performance is achieved.