Flocking And Its Application In Mobile Sensor Network

Like Genetic algorithm and Ant Colony algorithm, Flocking algorithm is inspired by natural flocking phenomenon and has its unique advantages in distributed control areas. For this reason, it draws attention of researchers from various fields. In recent years, the applications of flocking algorithm in robot networks, mobile sensor networks and Unmanned Arial vehicles (UAVs) push it to be a new hot-spot in the research of distributed control area.This paper mainly discusses the flocking algorithm and its applications in mobile sensor network. The contents can be divided into three parts:1. Considering the practical factors might be encountered in application of flocking algorithm, this paper extends the flocking algorithm by taking the time-delay of velocity into consideration, attaching the multi-hop network to the original network, designing the flocking algorithm with multiple leaders and considering the situation where only partial agents could receive leader’s information. The introduction of multi-hop network improves the connectivity of system and the converging speed of velocity consensus. Based on the traditional flocking algorithm with multiple leaders, this paper introduces a repulsive potential function to reduce the "embarrassing" relationship of adjacent agents, and the new algorithm could shorten the time for system to reach approximately stable. Flocking algorithm for the situation where only partial agents could receive leader’s information ensures all agents of system form a stable flock.2. Three consensus filters, namely Distributed Kalman Filter (DKF), Kalman Consensus Filter (KCF) and Sub-Optimal Discrete Kalman Consensus Filter (SOD-KCF), are qualitatively analyzed and compared. Meanwhile, the performance of three algorithms is analyzed under different connectivity of system and observing noise of nodes. From the aspect of average estimate error and estimate consistency factor, the results show that SOD-KCF could achieve better performance, which provides reference for the choosing of consensus filter part in coupled target tracking algorithm.3. The discrete coupled target tracking algorithm for mobile sensor network is designed based on the previous work. The stability of proposed algorithm is proven through theoretical deduction and numerical simulation. Furthermore, corresponding to the extension of flocking algorithm, this paper extends the discrete coupled target tracking algorithm by taking the time-delay of velocity observation into consideration, attaching the multi-hop network to original network, designing the algorithm with multiple targets and considering the situation where only partial sensor nodes could sense the status of targets. The stability of all extension algorithms are verified through theoretical deduction and numerical simulation. Lastly, the paper introduces the index including estimate error, estimate consistency factor and connectivity of network to evaluate the coupled target tracking algorithms.