Data Fusion In Sensor Networks

The use of state estimates to obtain fused values by either a weighted matrix approach or other approaches bordering aggregation have been extensively studied. In almost all the data fusion approaches, the noise in the system is assumed to be known apriori with a zero mean and a known variance. It is also assumed that the process and measurement noise are uncorrelated and remain constant throughout. This is not the case in many practical sensor applications. An unknown noise model will severely cripple the way states are estimated thereby resulting in faulty fused estimates. As the Kalman filter is inherently a computationally expensive data fusion approach when used on sensor motes with power constraints, performing costly matrix operations to obtain faulty observation defeats the purpose of data fusion. Our approach proposes the estimation of the system noise and updates it dynamically in the system. We provide a method for estimating the unknown correlated noise parameters and a fusion algorithm which takes into account only the correlated measurements of sensors for data fusion is investigated.

Researcher

Brijesh Kashyap

Advisor

Dr. Sanjay Madria