Evaluating Intelligence in Unmanned Ground Vehicle Teams

Evaluation of intelligence in Teams of Unmanned Ground Vehicles (UGVs) requires the development of consistent metrics and benchmarks. This is a complicated process as the implementation of the UGVs is problem and domain specific. Different performance requirements give rise to different set of metrics making the comparison of performance between two implementations difficult. In this paper, we focus on three aspects of intelligence, namely reconfiguration, adaptation and learning, and communications in UGV teams and investigate the development of metrics for measuring their performance. We also investigate the available benchmarks for intelligent systems and verify their suitability for measuring the performance of UGV teams. A hierarchical architecture called Adaptation and Learning at All levels (AL) for the UGV teams is presented. This architecture is designed to allow for a modular and hierarchical approach to implement deliberative and reactive behaviors in teams of autonomous vehicles. In this implementation, system intelligence is incorporated at all levels of the hierarchy. The performance of the proposed architecture is evaluated using the metrics identified.