Semi-flocking control for mobile sensor networks

Abstract

In this thesis, control algorithms are proposed for mobile agents to self-organize via local interactions. The mutual objective of the proposed algorithms is to monitor and track multiple targets within an area of interest (AoI). Based on mobility and sensing information collected via sensors and local communications among agents, mobile sensor networks (MSNs) with the proposed algorithms can efficiently search the AoI and provide the required coverage to each target. By incorporating path planning into the decision process, the proposed algorithms are capable of navigating mobile agents toward targets along time-efficient paths on terrains with non-uniform traverse costs. Furthermore, a distributed energy-efficient navigation approach is designed for controlling MSNs with the proposed algorithms on uneven terrains. First, a fully distributed semi-flocking algorithm is proposed for MSNs to search for emerging targets in the AoI and allocate mobile agents to track them. To facilitate distributed coordination among mobile agents, information maps and information lists are utilized to record area coverage and target information, respectively. Mobile agents can exchange information maps and information lists via local communications among nearby agents. A mode switching mechanism is designed for mobile agents to switch between searching and tracking modes. Based on information maps, mobile agents in searching mode are capable of visiting areas which have not been covered for the longest time to maximize their dynamic area coverage. Mobile agents constantly perform evaluations on different targets with the information lists. Each mobile agent uses its evaluation result to determine its next operating mode based on a state transition model. Semi-flocking-controlled MSNs can assign a number of mobile agents to track different targets and maintain small clusters around each target. Stability analysis on velocity matching and collision avoidance with single/multiple target(s) in multi-agent system is provided. Performances of the proposed semi-flocking algorithm, including area coverage, target tracking, and communication overheads are evaluated via extensive simulations. The uneven and rough landscapes in real-life applications have imposed extra challenges and raised demands for control algorithm design in MSNs. In order to perform area coverage and target tracking on terrains with non-uniform traverse costs, a path planning enabled semi-flocking algorithm is presented to tackle all the above challenges in such scenarios. We model an uneven terrain with irregular costs using a mobility map to represent the different maximum allowed speeds that a mobile agent can achieve when passing through patches in the terrain. A novel heuristic search algorithm is developed to perform path planning on the mobility map. Mobile agents can travel along time-efficient paths to reach a target and cooperate with other mobile agents that are tracking the same target. Results of extensive simulations show that semi-flocking-controlled mobile agents together with path planning can reach their targets faster with lower energy consumption compared to three existing flocking-based algorithms. Finally, to further reduce energy consumption resulting from the mobility of MSNs, an energy-efficient semi-flocking algorithm is proposed for controlling MSNs to perform area sensing and target tracking on rough terrains. A distributed navigation control, based on a terrain adaptation force and a navigation goal selection method, is introduced to guide mobile agents to move along terrain contours. MSNs with the proposed semi-flocking algorithm can achieve better area coverage and target tracking performance with lower energy expenditure when compared to the existing control protocols in the literature. Simulations are carried out to verify the effectiveness of the proposed algorithm.