This paper studies two interconnected problems in mobile sensor network deployment, the optimal placement of
heterogeneous mobile sensor platforms for cost-efficient and reliable coverage purposes, and the self-organizable
deployment. We first develop an optimal placement algorithm based on a "mosaicked technology" such that
different types of mobile sensors form a mosaicked pattern uniquely determined by the popularity of different
types of sensor nodes. The initial state is assumed to be random. In order to converge to the optimal state, we
investigate the swarm intelligence (SI)-based sensor movement strategy, through which the randomly deployed
sensors can self-organize themselves to reach the optimal placement state. The proposed algorithm is compared
with the random movement and the centralized method using performance metrics such as network coverage,
convergence time, and energy consumption. Simulation results are presented to demonstrate the effectiveness of
the mosaic placement and the SI-based movement.
In this paper, we present a cognitive agent framework (CAF) based on swarm intelligence and self-organization principles, and demonstrate it through collaborative processing for target classification in sensor networks. The framework involves integrated designs to provide both cognitive behavior at the organization level to conquer complexity and reactive behavior at the individual agent level to retain simplicity. The design tackles various problems in the current information processing systems, including overly complex systems, maintenance difficulties, increasing vulnerability to attack, lack of capability to tolerate faults, and inability to identify and cope with low-frequency patterns. An important and distinguishing point of the presented work from classical AI research is that the acquired intelligence does not pertain to distinct individuals but to groups. It also deviates from multi-agent systems (MAS) due to sheer quantity of extremely simple agents we are able to accommodate, to the degree that some loss of coordination messages and behavior of faulty/compromised agents will not affect the collective decision made by the group.