Intrinsically Motivated Model Learning for a Developing Curious Agent (2012)
Reinforcement Learning (RL) agents could benefit society by learning tasks that require learning and adaptation. However, learning these tasks efficiently typically requires a well- engineered reward function. Intrinsic motivation can be used to drive an agent to learn useful models of domains with limited or no external reward function. The agent can later plan on its learned model to perform tasks in the domain if given a reward function. This paper presents the TEXPLORE with Variance-And-Novelty-Intrinsic-Rewards algorithm (TEXPLORE-VANIR), an intrinsically motivated model- based RL algorithm. The algorithm learns models of the transition dynamics of a domain using decision trees. It calculates two different intrinsic rewards from this model: one to explore where the model is uncertain, and one to acquire novel experiences that the model has not yet been trained on. This paper presents experiments demonstrating that the combination of these two intrinsic rewards enables the algorithm to learn an accurate model of a domain with no external rewards and that the learned model can be used afterward to perform tasks in the domain. While learning the model, the agent explores the domain in a developing and curious way, progressively learning more complex skills. In addition, the experiments show that combining the agent’s intrinsic rewards with external task rewards enables the agent to learn faster than using external rewards alone.
In Eleventh International Conference on Autonomous Agents and Multiagent Systems - Adaptive Learning Agents Workshop (AAMAS - ALA), June 2012.

Todd Hester todd [at] cs utexas edu
Peter Stone pstone [at] cs utexas edu