Ai 2 Robotics

Optimization of robotic cells operations using reinforcement learning

Keywords: Reinforcement Learning (RL); artificial intelligence (Ai); Industrial robotics, Pick&Place, Industry 4.0

Authors: Claudino Costa, António H. J. Moreira

Robots are increasingly used in manufacturing to perform repetitive and complex tasks. However in the future it is predicted that robots will need to be more autonomous, able to adapt to changing environments and capable of performing dynamic tasks. The use of a digital twin accomplishes these goals in a simulated environment reducing the cost of hardware and facilitating the implementation of artificial intelligence to manipulate the robot.
This work presents an early study of the applicability of reinforcement learning algorithms to control a manipulator in a ‘Pick and Place’ task, which are characterized for precise locations, using only distance to target as the reward function.