Heuristic Machine-Learning Approach to the Control of an Anthropomimetic Robot Arm

This paper presents the novel solution to problems that emerge as a result of using the heuristic machine-learning approach that follows the cybernetic concept of black-box. The research is focused on experience-based learning as well as on point-to-point control problem of multi-jointed anthropomimetic robot arm. The model representation, hence kinematics, dynamics, and control parameters of the robot are considered unknown. However, the system is available for any input-to-output experiment. The biologically inspired control algorithm consists of two phases: feedforward and feedback control, both relying on the experience base. The acquiring and exploitation control stage are highlighted in this paper.