A reinforcement learning approach to industrial robotic operation
In the current industrial market, all the companies strive to achieve shorter lead-times, reduced scrap and elimination of bottlenecks. To this regard, industries have recently employed human-robot teaming strategies in assembly lines to improve accuracy and to reduce human burden. But this system can fail to work efficiently as the pace of the worker in the human-robot team decides the pace of operation. Therefore, programming the robot to adapt to the pace of the worker is essential for effective human-robot team operations. This thesis uses a reinforcement learning approach to transform a conventional robotic system (robotic arm) trained using human demonstration into an artificially intelligent robotic arm using reinforcement learning. This work demonstrates the reinforcement learning approach to a human-robot team for a real industrial assembly process on a simulated platform in Python using the Pygame package.