Computational intelligence for a wire-in-hole task in a micromanipulation system

The paper researches the typical assembly operation of flexible objects, the \´wire-in-hole\´ operation, with a magnetically actuated micromanipulator. In order to realize the efficient wire-in-hole operation, a learning approach is proposed. In the proposed approach, a fuzzy controller is used to control the end-effector\´s motion. Each fuzzy rule has a set of possible strategies. The selection probability of strategies is updated by the Q-model learning automaton, and the output parameters of strategies are updated by the steady-state genetic algorithm approach. A 2D wire-in-hole operation is simulated in the paper. After learning, the wire can be inserted into the hole with a much smoother path than "try-adjustment" approaches does. Simulation results prove the feasibility of the proposed approach.