Affordance-Aware Planning

Planning algorithms for non-deterministic domains are often intractable in large state spaces due to the well-known “curse of dimensionality.” Existing approaches to address this problem fail to prevent the planner from considering many actions which would be obviously irrelevant to a human solving the same problem. We formalize the notion of affordances [7] as knowledge added to an MDP that prunes actions in a stateand rewardgeneral way. This pruning significantly reduces the number of state-action pairs the agent needs to evaluate in order to act optimally. We demonstrate our approach in the Minecraft domain, showing significant increase in speed and reduction in state-space exploration during planning. Further, we provide a learning framework that enables an agent to learn affordances through experience, removing the agent’s dependence on the expert. We provide preliminary results indicating that the learning process effectively produces affordances that help solve an MDP faster.