Learning insertion task of a flexible beam by virtual agents

Proposes a method to learn the typical assembly operation of inserting a flexible wire into hole. The input space is divided into different contexts. Several virtual agents based on a learning automaton are constructed in the output space. Through learning, the agents can learn optimal actions according to different contexts, and achieve the insertion task together. The paper also proposes a computation approach based on a multi-body model to simulate the insertion process. The simulation results of a 2D insertion operation prove the feasibility of the proposed methods.