Risk-Sensitive Generative Adversarial Imitation Learning
This work addresses risk-sensitive decision-making in imitation learning for robotics and control applications, but it is incremental as it builds on existing GAIL methods.
The paper tackles risk-sensitive imitation learning by proposing RS-GAIL, which matches risk profiles between agent and expert using JS divergence and Wasserstein distance, showing performance improvements over GAIL and RAIL in MuJoCo and OpenAI control tasks.
We study risk-sensitive imitation learning where the agent's goal is to perform at least as well as the expert in terms of a risk profile. We first formulate our risk-sensitive imitation learning setting. We consider the generative adversarial approach to imitation learning (GAIL) and derive an optimization problem for our formulation, which we call it risk-sensitive GAIL (RS-GAIL). We then derive two different versions of our RS-GAIL optimization problem that aim at matching the risk profiles of the agent and the expert w.r.t. Jensen-Shannon (JS) divergence and Wasserstein distance, and develop risk-sensitive generative adversarial imitation learning algorithms based on these optimization problems. We evaluate the performance of our algorithms and compare them with GAIL and the risk-averse imitation learning (RAIL) algorithms in two MuJoCo and two OpenAI classical control tasks.