Robotic Arm Control and Task Training through Deep Reinforcement Learning
This work provides an incremental comparison of existing deep reinforcement learning methods for robotic manipulation tasks.
This paper compared Trust Region Policy Optimization and DeepQ-Network with Normalized Advantage Functions against Deep Deterministic Policy Gradient and Vanilla Policy Gradient for robotic arm control tasks like reaching random target poses and pick-and-place. The results showed the former methods performed better in both simulation and real-world experiments, with policies trained in simulation transferring to real environments with minimal changes.
This paper proposes a detailed and extensive comparison of the Trust Region Policy Optimization and DeepQ-Network with Normalized Advantage Functions with respect to other state of the art algorithms, namely Deep Deterministic Policy Gradient and Vanilla Policy Gradient. Comparisons demonstrate that the former have better performances then the latter when asking robotic arms to accomplish manipulation tasks such as reaching a random target pose and pick &placing an object. Both simulated and real-world experiments are provided. Simulation lets us show the procedures that we adopted to precisely estimate the algorithms hyper-parameters and to correctly design good policies. Real-world experiments let show that our polices, if correctly trained on simulation, can be transferred and executed in a real environment with almost no changes.