Sequential Evidence Accumulation VIA Integrating Dempster-Shafer Reasoning and Estimation Theory

Integrating evidence from multiple sources has been heavily researched with various approaches such as Bayes and Dempster-Shafer (D-S) being widely adopted. Integrating evidence from a single sensor over time is becoming more common due to the Internet of Things (IoT). Researchers have often adopted these same mechanisms used for multi-source integration for integrating evidence over time. There are many issues with this approach, however, including the facts that time series are order dependent and that changes in the environment, due to a dynamic environment or due to assignable errors in the sensor measurements may cause significant evidential conflict. While methods such as D-S theory are suitable for multi-source evidence accumulation, we propose an alternate approach for sequential evidence accumulation. Our approach integrates the set theoretic nature of Dempster-Shafer theory with an estimation structure based on Kalman filtering. This approach is motivated both from traditional signal processing as well as from research in human psychology where a very similar filtering structure has been proposed for modeling human evidence accumulation. The approach is demonstrated to be effective using a smart airbag deployment system application.