State estimation and fusion over long-haul links under linear constraints

We consider a number of sensors deployed over a large geographical area for tracking a target with linear constraints on its motion dynamics which are specified by Kalman filter conditions. The state estimates from the sensors are sent over long-haul networks to a remote fusion center, where they are fused to improve the tracking accuracy. The mismatches among the sensors in incorporating the target motion constraints into their state estimates, along with the information loss over the long-haul links, need to be accounted for by the state estimation and fusion algorithms. We propose using the null-space method to incorporate these constraints into three fusion algorithms based on information matrix, simple linear fuser and covariance intersection methods. Then using a tracking example, we study the impact of these factors and compare the accuracy of these fusion algorithms. Results show that incorporating knowledge of constraints directly or indirectly at the fusion center can effectively improve the overall tracking accuracy under various degrees of long-haul communication loss.