A Hybrid Approach for Reinforcement Learning Using Virtual Policy Gradient for Balancing an Inverted Pendulum
This work provides an incremental improvement in control robustness and training efficiency for balancing an inverted pendulum, primarily benefiting robotics and control systems researchers.
This paper tackles the problem of balancing an inverted pendulum using a hybrid reinforcement learning approach. They trained a single-hidden-layer neural network in a physically accurate simulation, achieving robust control that could be transferred to a real inverted pendulum. The method resulted in smoother control, faster learning, and better disturbance resistance compared to existing methods.
Using the policy gradient algorithm, we train a single-hidden-layer neural network to balance a physically accurate simulation of a single inverted pendulum. The trained weights and biases can then be transferred to a physical agent, where they are robust enough to to balance a real inverted pendulum. This hybrid approach of training a simulation allows thousands of trial runs to be completed orders of magnitude faster than would be possible in the real world, resulting in greatly reduced training time and more iterations, producing a more robust model. When compared with existing reinforcement learning methods, the resulting control is smoother, learned faster, and able to withstand forced disturbances.