Autonomous robot teaching using a smart robot in production line

In industrial robot applications, to adapt the environmental changes the robot teaching process has to be performed frequently by operators. This process increases the operational cost and reduces the manufacturing efficiency. This paper proposes to use a smart robot(named “adult” robot) with advanced sensing and decision-making capabilities to teach “child” robots in the production line. A Markov Decision Process(MDP) is formulated and solved using Q-Learning to correct the “child” robot´s tool position. The proposed algorithm was implemented to teach an industrial robot (child) to perform a high accuracy peg-in-hole assembly process. The corresponding “adult” robot is a mobile robot platform with an in-hand camera. The experimental results verify the effectiveness of the proposed method. Since there is no calibration between the vision system, the “adult” robot and the “child” robot, the flexibility of the proposed method is greatly increased. Hence it can be easily applied in industrial applications where robot teaching is needed.