Multi-agent case-based reasoning for cooperative reinforcement learners
ECCBR'06 Proceedings of the 8th European conference on Advances in Case-Based Reasoning
Similarity-Based Retrieval With Structure-Sensitive Sparse Binary Distributed Representations
Computational Intelligence
Hi-index | 0.00 |
Satellites represent scarce resources that must be carefully scheduled to maximize their value to service consumers. Near-optimal satellite task scheduling is so computationally difficult that it typically takes several hours to schedule one day's activities for a set of satellites and tasks. Thus, often a requestor will not know if a task will be scheduled until it is too late to accommodate scheduling failures. This paper presents our experiences creating a fast Analogical Reasoning (AR) system and an even faster Case-Based Reasoner (CBR) that can predict, in less than a millisecond, whether a hypothetical task will be scheduled successfully. Requestors can use the system to refine tasks for maximum schedulability. We report on three increasingly narrow approaches that use domain knowledge to constrain the problem space. We show results that indicate the method can achieve 80% accuracy on the given problem.