Multi-Sensor Data Fusion: An Unscented Least Squares Approach.

摘要

This manuscript provides an approach to solving the nonlinear least squares problem that arises in decentralized fusion. In decentralized fusion, measurements are first processed at the sensor node before they are relayed to the central node. Even though almost all sensor noise can be modeled as additive noise, the additive nature of the measurement noise is lost when the signal is processed at the sensor node. The proposed unscented transformation- based approach helps to tackle the non-additive nature of the noise in the nonlinear least squares problem. Numerical simulations indicate that the proposed unscented transformation-based approach yields desired results.