On Using Stochastic Automata for Trajectory Planning of Robot Manipulators in Noisy Workspaces

We consider the problem of a robot manipulator operating in a noisy workspace. The robot is assigned the task of moving from Pi to Pf. Since Pi is its initial position, this position can be known fairly accurately. However, since Pf is usually obtained as a result of a sensing operation, possibly vision sensing, we assume that Pf i s noisy. We propose a solution to achieve the motion which involves a new learning automaton, called the Discretized Linear Reward-Penalty (DLR~) automaton. The strategy we propose does not involve the computation of any inverse kinematics. Altematively, an automaton is positioned at each joint of the robot, and by processing repeated noisy observations of Pf the automata operate in parallel to control the motion of the manipulator.The advantages and the possible disadvantages of the scheme are also discussed.