Potential-Based Reward Shaping For Intrinsic Motivation
This addresses a critical issue for reinforcement learning practitioners using intrinsic motivation in complex environments, though it is an incremental extension of existing potential-based reward shaping methods.
The paper tackled the problem of intrinsic motivation reward-shaping methods inadvertently altering optimal policies in sparse-reward environments, and presented Potential-Based Intrinsic Motivation (PBIM) to convert such rewards into a form that preserves optimal policies, demonstrating in MiniGrid DoorKey and Cliff Walking environments that it prevents suboptimal convergence and speeds up training.
Recently there has been a proliferation of intrinsic motivation (IM) reward-shaping methods to learn in complex and sparse-reward environments. These methods can often inadvertently change the set of optimal policies in an environment, leading to suboptimal behavior. Previous work on mitigating the risks of reward shaping, particularly through potential-based reward shaping (PBRS), has not been applicable to many IM methods, as they are often complex, trainable functions themselves, and therefore dependent on a wider set of variables than the traditional reward functions that PBRS was developed for. We present an extension to PBRS that we prove preserves the set of optimal policies under a more general set of functions than has been previously proven. We also present {\em Potential-Based Intrinsic Motivation} (PBIM), a method for converting IM rewards into a potential-based form that is useable without altering the set of optimal policies. Testing in the MiniGrid DoorKey and Cliff Walking environments, we demonstrate that PBIM successfully prevents the agent from converging to a suboptimal policy and can speed up training.