Automated Control Method of Multiple Peg-in-Hole Assembly for Relay and Its Socket

In an automatic control system, multiple peg-in-hole (MPIH) assembly of a relay and its socket is often required. However, the existing hybrid position/force control (HPFC) methods cannot meet the requirement for MPIH assembly because of its great complexity. A reinforcement learning (RL)–based HPFC method is presented for the assembly of a relay and its socket, whose controller's parameters are adjusted to adapt to the complexity. First, the socket is modeled as a spring based on its internal structural characteristics. The force in assembly simulation is computed with this model by combining it with their relative positions. Second, the proposed RL-based HPFC method segregates the assembly motion control into a force control branch and a position control branch. In the force control branch, a proportional-derivative controller regulates radial motion to obtain compliance. In the position control branch, a proportional controller manages the axial motion, which is further constrained by the radial force to ensure safety. Finally, a parameter adjustment method for the controller based on RL is further developed. The sliding window state collection is used for training. A series of comparative experiments in both simulation and real assembly environments are conducted. The experimental results validate the effectiveness of the proposed method.