Cooperative localization and tracking of resource-constrained mobile nodes

Abstract

Tracking mobile nodes is a topic of significant scientific and commercial interest. However, due to the high energy consumption of GPS receivers, high-frequency, long-term tracking of resource-constrained mobile nodes remains challenging. Cooperative position tracking has been proposed to improve energy efficiency, however most existing schemes perform opportunistic cooperation and optimize for either energy or accuracy. In this thesis, a cooperative tracking algorithm is proposed using planned cooperation, where a cluster head centrally coordinates resource usage among cluster members. This technique assumes the existence of a reasonably stable cluster of mobile nodes carried by groups of people, animals, or vehicles. Several variants of the proposed cluster-based tracking algorithm are investigated, which differ based on how a node estimates its position after receiving a position update from the cluster head. These techniques include estimating a node's position to be the same as the position update, or using a Kalman filter to incorporate inertial measurement sensor data as well as the received position updates. A second group of localization techniques estimates position based on RSSI-based range estimates from a subset of nodes using GPS to estimate their position. Multilateration is used to estimate node positions based on these ranges. The theoretical bounds (Cramer-Rao Lower Bounds) for localization performance are derived in a practical scenario where both neighbouring node positions and RSSI range estimates have errors following normal (position) and log-normal (range) probability distributions. Targeting resource-constrained nodes, two different self-localization algorithms are proposed which use these different error distributions. The two algorithms vary in computational complexity and localization performance. In simulated experiments, the proposed self-localization algorithms reduce the root mean square error and bias-norm up to 50% while reducing computation time by 80% compared to previously published algorithms. Finally, a multi-mode tracking algorithm is proposed building on the proposed clusterbased tracking and self-localization algorithms. This algorithm uses either individual mea

Multilateration with Self-Calibration: Uncertainty Assessment, Experimental Measurements and Monte-Carlo Simulations

“…The log-likelihood function can then be developed as follows since the variables of the observation vector are statistically independent of each other [28]:…”