Episodic Curiosity through Reachability
This addresses the challenge of sparse rewards for reinforcement learning agents in complex environments, representing an incremental improvement over existing curiosity-based methods.
The paper tackles the problem of sparse rewards in reinforcement learning by introducing a curiosity method that uses episodic memory to create a novelty bonus based on reachability, which helps overcome 'couch-potato' issues in prior work. The method outperforms the state-of-the-art ICM in navigational tasks in ViZDoom and DMLab, and enables an ant to learn locomotion from first-person-view curiosity in MuJoCo.
Rewards are sparse in the real world and most of today's reinforcement learning algorithms struggle with such sparsity. One solution to this problem is to allow the agent to create rewards for itself - thus making rewards dense and more suitable for learning. In particular, inspired by curious behaviour in animals, observing something novel could be rewarded with a bonus. Such bonus is summed up with the real task reward - making it possible for RL algorithms to learn from the combined reward. We propose a new curiosity method which uses episodic memory to form the novelty bonus. To determine the bonus, the current observation is compared with the observations in memory. Crucially, the comparison is done based on how many environment steps it takes to reach the current observation from those in memory - which incorporates rich information about environment dynamics. This allows us to overcome the known "couch-potato" issues of prior work - when the agent finds a way to instantly gratify itself by exploiting actions which lead to hardly predictable consequences. We test our approach in visually rich 3D environments in ViZDoom, DMLab and MuJoCo. In navigational tasks from ViZDoom and DMLab, our agent outperforms the state-of-the-art curiosity method ICM. In MuJoCo, an ant equipped with our curiosity module learns locomotion out of the first-person-view curiosity only.