Imitation and Reinforcement Learning in Agents with Heterogeneous Actions

Reinforcement learning techniques are increasingly being used to solve di cult problems in control and combinatorial optimization with promising results. Implicit imitation can accelerate reinforcement learning (RL) by augmenting the Bellman equations with information from the observation of expert agents (mentors). We propose two extensions that permit imitation of agents with heterogeneous actions: feasibility testing, which detects infeasible mentor actions, and k-step repair, which searches for plans that approximate infeasible actions. We demonstrate empirically that both of these extensions allow imitation agents to converge more quickly in the presence of heterogeneous actions.

• Type: Conference
• Year: 2001
• Where: AI
• Authors: Bob Price, Craig Boutilier
• Keywords: AI 2001, Artificial Intelligence, Heterogeneous Actions, Infeasible Mentor Actions, Reinforcement Learning