Resource management for wireless networks of bearings-only sensors

• Main Author: Le, Qiang
• Format: Others
• Language: en_US
• Published: Georgia Institute of Technology 2006
• Subjects: Resource management; Kalman filters; Wireless networks
• Online Access: http://hdl.handle.net/1853/10548

The thesis focuses on resource management or sensor allocation when we use bearings-only measurements to track targets in an unattended ground sensor (UGS) network. Intelligent resource management is necessary because each UGS sensor node has limited power and it is desirable that estimation performance not degrade very much when only a few nodes are active to maximize the effective tracking lifetime. For scheduling to prolong the tracking lifetime, a new energy-based (EB) metric is proposed to model the number of snapshots remaining for a hypothesized node set, i.e., the remaining battery energy divided by the energy to sense and share information amongst the node set. Unlike other methods that use the total energy consumed for the given snapshot as the energy-based metric, the new EB metric can achieve load balancing of the nodes without resorting to computationally demanding non-myopic optimization. The metrics to choose nodes at a given snapshot could be geometry-based (GB) to minimize the estimation error, EB, or multiobjective. In determining the active set, each node only knows the existence of itself, the active set of nodes from the previous snapshot and the node's neighbors, i.e., the set of nodes within a distance of r_nei. When measuring the tracking lifetime of the system, we propose an adaptive transmission range control, known as the knowledge pool (KP) where the transmission range is determined by the knowledge of the network and the currently remaining battery level. The KP saves more energy usage than another adaptive transmission range control bounded with the GB metric when the global location information is available. We also provide practical search algorithms to optimize a constraint metric (multiobjective function) using one metric as the optimization metric under the constraint of the other. We also demonstrate the resource management schemes for multitarget tracking with the field data.