Aspect-oriented design of sensor networks

The rapid technology development in wireless communication and embedded micro-sensing devices has made the distributed sensor networks (DSN) an area of national importance. Wireless sensor networks are an important military technology with civil and scientific applications. More importantly, the design and analysis of sensor networks can be quite complicated, since each node must simultaneously interact with many other nodes to achieve multiple goals. In this paper, we show how this problem can be made tractable by designing separate protocols for each aspect of a node's behavior. We model this discrete event system by Petri Nets and then formulate three aspect hierarchies: sensing, communications, and command. Within each aspect hierarchy, a node is dynamically assigned roles. To combine the hierarchies, control specifications are derived that enforce consistency across the aspects. Controllers are created using three discrete event methodologies to show how computationally independent aspect-oriented designs can be integrated to form a unified distributed system. The controller methodologies used are: (i) Petri Nets, (ii) finite state automata (FSA) using the Ramadge and Wonham approach, and (iii) vector addition control using the Wonham and Li approach. Finally, we contrast the controller design methodologies by presenting the advantages and disadvantages for each method. In conclusion, for our Petri Nets modeled DSN system with n places and m transitions, constructing Petri Nets controller is computationally efficient but with controller execution time complexity of O(n × m²). On the other hand, FSA controller provides prompt response with time complexity of O(n × m) at the cost of manual offline state space search and encoding. Thus this method is only applicable to medium and small size system.