Action Categorization for Computationally Improved Task Learning and Planning
This addresses computational efficiency for AI agents in complex environments, but it appears incremental as it builds on existing psychological concepts and applies them to known domains.
The paper tackles the problem of improving computational performance in task learning and planning by introducing the Action-Category Representation (ACR), which reduces the action space for agents, leading to demonstrated benefits in domains like StarCraft and Lightworld.
This paper explores the problem of task learning and planning, contributing the Action-Category Representation (ACR) to improve computational performance of both Planning and Reinforcement Learning (RL). ACR is an algorithm-agnostic, abstract data representation that maps objects to action categories (groups of actions), inspired by the psychological concept of action codes. We validate our approach in StarCraft and Lightworld domains; our results demonstrate several benefits of ACR relating to improved computational performance of planning and RL, by reducing the action space for the agent.