Information fusion and cooperative control for target search and localization in multi-agent sensor networks.
Date of Issue2013
School of Electrical and Electronic Engineering
Multi-agent sensor networks (MSNs) are comprised of multiple static or mobile agents capable of collecting, processing, storing and transferring information from one agent to another. They can play a critical role in several application domains such as landmine detection and identification, monitoring of endangered species, monitoring of urban environments, manufacturing plants, and civil infrastructure, and intruder and target detection systems. These networks are expected to operate cooperatively and reliably in cluttered dynamic environments with little human intervention. However, coordinating such large heterogeneous sensor networks is challenging and requires the development of novel methods of communication, motion control and planning, computation, proactive estimation and sensing, and power management. This thesis summarizes research results on information fusion and cooperative control for target search and localization in MSNs. First, the distributed target estimation problem for linear time-varying systems in MSNs is addressed. A diffusion Kalman filtering algorithm based on the covariance intersection method is proposed, where local estimates are fused by incorporating the covariance information of local Kalman filters. Our algorithm leads to a stable estimate for each agent regardless of whether or not the system is uniformly observable locally with the measurements of itself and its neighbors as long as the system is uniformly observable with the measurements of all agents and the communication is sufficiently fast compared to the sampling. Simulation results validate the effectiveness of the proposed distributed Kalman filtering algorithm. Second, we study the cooperative control for target localization and pursuit in the ground MSNs. Conventional target tracking methods always require an explicit system observation model of the target positions, which, however, would fail if such model is not available. Thus, a distributed target localization and pursuit scheme is proposed based on discrete measurements of the energy intensity field produced by static or mobile targets. The accurate observation model of such field is not available except some critical bounds. By our control strategy, all agents are categorized into two groups: the leaders, responsible for the target pursuit, and the followers, responsible for the formation and connectivity maintenance. The influence of the system parameters on the convergence of leaders to the local maximum points is analyzed. Finally, the proposed scheme is demonstrated by simulation. Next, we study the cooperative search for multiple stationary ground targets by a group of unmanned aerial vehicles (UAVs) with limited sensing and communication capabilities, where targets and UAVs are moving in planes. The whole surveillance region is partitioned into cells where each cell is associated with a probability of target existence within the cell, which constitutes a probability map for the whole region. Each agent keeps an individual probability map and updates the map individually with measurements according to Bayesian rule. A nonlinear transformation of the probability map is introduced to simplify the computation by linearizing the Bayesian update. A consensus-like distributed fusion scheme is proposed for multi-agent map fusion. It is proven that all the individual probability maps converge to the same one that reflects the true existence or nonexistence of targets within each cell. Coverage and topology control algorithms are designed for the path planning of mobile agents. Moreover, the performance of the fusion scheme for asynchronous implementations of sampling and communication is analyzed. Finally, the effectiveness of the proposed algorithms is illustrated via simulations. Further, we consider the vision-based cooperative search for multiple mobile ground targets by a group of UAVs with limited sensing and communication capabilities and moving in a three dimensional space. The airborne camera on each UAV has a limited field of view and its target discriminability varies as a function of altitude. First, a general target detection probability model is built based on the physical imaging process of a camera. Based on the previous results, we propose a generalized distributed probability map updating model which includes the fusion of measurement information, information sharing among neighboring agents, information decaying and transmission due to environmental changes such as the target movement. Furthermore, we formulate the target search problem by multiple agents as a cooperative coverage control problem by optimizing the collective coverage area and the detection performance. The proposed map updating model and the cooperative control scheme are distributed, i.e., allowing that each agent only communicates with its neighbors within its communication range. Finally, the effectiveness of the proposed algorithms is illustrated by simulation. Finally, we investigate the adaptive sensing for three-dimensional target tracking in MSNs based on measurements from the time-difference-of-arrival (TDOA) sensors. An iterated filtering algorithm combined with the Gauss-Newton method is applied to estimate the target location. By minimizing the determinant of the estimation error covariance matrix, an optimal adaptive sensing strategy is developed. A gradient-based control law for each agent is proposed and a set of stationary points for local optimum geometric configurations of the agents is given. The optimal sensing strategy is further compared with other sensing strategies using different optimization criteria such as the Cramer-Rao lower bound. Possible modifications of the proposed optimal sensing strategy are also discussed. Finally, the proposed sensing strategy is demonstrated and compared with other sensing strategies by simulation, which shows that our method can provide good performance with even only one TDOA measurement at each time.
DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation::Control engineering