Fusion of state estimates over long-haul sensor networks under random delay and loss

Long-haul sensor networks are deployed in a wide range of applications from national security to environmental monitoring. We consider target tracking over a long-haul sensor network, wherein state and covariance estimates are sent from sensors to a fusion center that generates a fused state. Fusion serves as a viable means to improve the estimation performance to meet the system requirement on accuracy and delay. Communications over the long-haul links, such as submarine fibers and satellite links, is subject to long latencies and high loss rates that lead to many lost or out-of-order messages and may significantly degrade the fusion performance. We propose an online selective fuser to combine the received state estimates based on estimated information contribution from the pending data. By concurrently using prediction and retrodiction, the fuser opportunistically makes timely decisions to achieve a balance between accuracy and timeliness of the fused estimate. Simulation results show that our method effectively maintains high levels of fusion performance under various communication delay and loss conditions.