Wasserstein Distance Maximizing Intrinsic Control
arXiv:2110.15331
Abstract
This paper deals with the problem of learning a skill-conditioned policy that acts meaningfully in the absence of a reward signal. Mutual information based objectives have shown some success in learning skills that reach a diverse set of states in this setting. These objectives include a KL-divergence term, which is maximized by visiting distinct states even if those states are not far apart in the MDP. This paper presents an approach that rewards the agent for learning skills that maximize the Wasserstein distance of their state visitation from the start state of the skill. It shows that such an objective leads to a policy that covers more distance in the MDP than diversity based objectives, and validates the results on a variety of Atari environments.
References in corpus (9)
- Variational Intrinsic Control
- Intrinsically Motivated Goal Exploration Processes with Automatic Curriculum Learning
- Primal Wasserstein Imitation Learning
- Wasserstein Adversarial Imitation Learning
- On Learning Intrinsic Rewards for Policy Gradient Methods
- Wasserstein Distance guided Adversarial Imitation Learning with Reward Shape Exploration
- Relative Variational Intrinsic Control
- What Can Learned Intrinsic Rewards Capture?
- Adversarial Intrinsic Motivation for Reinforcement Learning