Fuzzy-logic-based reinforcement learning of admittance control for automated robotic manufacturing

An approach to admittance control using fuzzy-logic-based reinforcement learning is proposed for the robotic automation of typical manufacturing operations. The proposed approach provides the necessary nonlinear control actions required in a typical automated robotic manufacturing task. Simultaneously, it reduces the controller development time due to the incorporation of pre-existing process knowledge in a neural-network form. The pre-existing knowledge is further refined using reinforcement learning via a CMAC (Cerebellar Model Articulation Controller) based critic network. Automated robotic deburring offers an attractive alternative to manual deburring in terms of reduced costs and improved quality of the finished parts. Hence. robotic deburring is used as an example of a typical manufacturing task to verify the performance of the proposed approach. However, the approach is general enough to be easily extended to similar manufacturing tasks. Simulation results are presented, which demonstrate the effectiveness of the proposed strategy in controlling the automated robotic deburring task. (C) 1998 Elsevier Science Ltd. All rights reserved.