Marginal Utility for Planning in Continuous or Large Discrete Action Spaces
This addresses the challenge of action generation for planning algorithms in complex environments, offering a generalizable method that improves performance over domain-specific or existing learned approaches, though it is incremental in advancing planning techniques.
The paper tackles the problem of generating effective candidate actions for sample-based planning in continuous or large discrete action spaces by introducing a novel objective called marginal utility, which measures the increase in value of an action over previously generated ones, and demonstrates that this approach outperforms hand-coded schemes, trained stochastic policies, and other objectives in domains like curling and a location game.
Sample-based planning is a powerful family of algorithms for generating intelligent behavior from a model of the environment. Generating good candidate actions is critical to the success of sample-based planners, particularly in continuous or large action spaces. Typically, candidate action generation exhausts the action space, uses domain knowledge, or more recently, involves learning a stochastic policy to provide such search guidance. In this paper we explore explicitly learning a candidate action generator by optimizing a novel objective, marginal utility. The marginal utility of an action generator measures the increase in value of an action over previously generated actions. We validate our approach in both curling, a challenging stochastic domain with continuous state and action spaces, and a location game with a discrete but large action space. We show that a generator trained with the marginal utility objective outperforms hand-coded schemes built on substantial domain knowledge, trained stochastic policies, and other natural objectives for generating actions for sampled-based planners.