Curiosity Based Reinforcement Learning on Robot Manufacturing Cell

  • 2020-11-17 16:19:47
  • Mohammed Sharafath Abdul Hameed, Md Muzahid Khan, Andreas Schwung
  • 0


This paper introduces a novel combination of scheduling control on a flexiblerobot manufacturing cell with curiosity based reinforcement learning.Reinforcement learning has proved to be highly successful in solving tasks likerobotics and scheduling. But this requires hand tuning of rewards in problemdomains like robotics and scheduling even where the solution is not obvious. Tothis end, we apply a curiosity based reinforcement learning, using intrinsicmotivation as a form of reward, on a flexible robot manufacturing cell toalleviate this problem. Further, the learning agents are embedded into thetransportation robots to enable a generalized learning solution that can beapplied to a variety of environments. In the first approach, the curiositybased reinforcement learning is applied to a simple structured robotmanufacturing cell. And in the second approach, the same algorithm is appliedto a graph structured robot manufacturing cell. Results from the experimentsshow that the agents are able to solve both the environments with the abilityto transfer the curiosity module directly from one environment to another. Weconclude that curiosity based learning on scheduling tasks provide a viablealternative to the reward shaped reinforcement learning traditionally used.


Quick Read (beta)

loading the full paper ...