Evidentialist foundationalist argumentation for multi-agent sensor fusion

Abstract

In this paper, an approach to evidence-based argumentation called Evidentialist Foundationalist Argumentation (EFA) is formally defined in terms of the ASPIC framework. The EFA framework is then used as the basis for general argument patterns applied to the problem domain of Sensor Fusion. These general Sensor Fusion argument patterns serve as templates for concrete arguments constructed by agents in an in situ Sensor Web. These agents use EFA to solve specific instances of the Decentralized Sensor Fusion problem by strategically sharing evidence from their arguments using a Share on Disagreement protocol. Using real-world data, the performance of this multiagent system is compared to the performance of another multiagent system employing a Kalman Filtering approach. The results are statistically analyzed using omega-squared effect sizes produced by ANOVA with p values < 0.05. The EFA based system is found to outperform the Kalman Filtering system in terms of accuracy with mostly high and medium effect sizes. The Kalman Filtering system is found to outperform the EFA based system in terms of communication costs with mostly low effect sizes.