Case-Based Parameter Selection for Plans: Coordinating Autonomous Vehicle Teams

摘要

Executing complex plans for coordinating the behaviors of multiple heterogeneous agents often requires setting several parameters. For example, we are developing a decision aid for deploying a set of autonomous vehicles to perform situation assessment in a disaster relief operation. Our system, the Situated Decision Process (SDP), uses parameterized plans to coordinate these vehicles. However, no model exists for setting the values of these parameters. We describe a case-based reasoning solution for this problem and report on its utility in simulated scenarios, given a case library that represents only a small percentage of the problem space. We found that our agents, when executing plans generated using our case-based algorithm on problems with high uncertainty, performed significantly better than when executing plans using baseline approaches.