Modular Decentralized Control of Fruit Picking Redundant Manipulator

Industry requires automatic fruit picking due to increasing labour costs and human variability in picking skill. This requires motion of a manipulator in unmapped dynamic environments that has hyper-redundancy to find its way successfully through leaves and branches. This paper uses a modular arm technique to generate kinematic redundancy in constructing and controlling a fruit picking arm that can navigate in such a flexible manner. The need for a full environment reconstruction in one master controller is bypassed by having localized subgoals for each module. Decentralized control of the arm assists modular construction and selecting among the hyper-redundancy infinity of solutions. Singularities are avoided if the length of the deployed arm is greater than the distance to the object. If not, then the arm will point directly at the object and measure the shortfall in distance. The computational burden is kept down by restricting inverse kinematics to within each module. Results show stable controlled movement of the end effecter from multiple starting positions covering all possible approaches to the goal.