Robust, Evidence-Based Data Fusion

Abstract This paper presents a strategy for determining system state estimates from unreliable and noisy system inputs. The concepts of sensor error orthogonality, meta-sensors, output equivalence, and adaptive state estimation are discussed. Although the proposed data fusion strategy is applicable to a large range of both 2-D and 3-D applications that consume unreliable inputs to support real-time control, an autonomous ground vehicle is used as a motivating example.