Learning Options from Demonstration using Skill Segmentation
This work addresses the challenge of enabling agents to autonomously discover reusable skills from demonstrations, which is incremental as it builds on existing segmentation and IRL techniques.
The paper tackles the problem of learning hierarchical skills (options) from segmented human demonstrations by using nonparametric Bayesian clustering for segmentation and inverse reinforcement learning for reward functions, then learning option initiation and termination conditions via one-class SVM. The method was demonstrated in the four rooms domain, where inferred options improved agent learning and planning.
We present a method for learning options from segmented demonstration trajectories. The trajectories are first segmented into skills using nonparametric Bayesian clustering and a reward function for each segment is then learned using inverse reinforcement learning. From this, a set of inferred trajectories for the demonstration are generated. Option initiation sets and termination conditions are learned from these trajectories using the one-class support vector machine clustering algorithm. We demonstrate our method in the four rooms domain, where an agent is able to autonomously discover usable options from human demonstration. Our results show that these inferred options can then be used to improve learning and planning.