A predictive sensor network using ant system

The need for a robust predictive sensor communication network inspired this research. There are many critical issues in a communication network with different data rate requirements, limited power and bandwidth. Energy con- sumption is one of the key issues in a sensor network as energy dissipation occurs during routing, communication and monitoring of the environment. This paper covers the routing of a sensor communication network by applying an evolu- tionary algorithm - the ant system. The issues considered include optimal energy, data fusion from different sensor types and predicting changes in environment with respect to time. A building consists of a sensor network with a multitude of wireless communication links and the sensors are inter- connected by means of RF communication links. The functionality of nodes in this application include sensing, collect- ing and distributing dynamic information within the network. Energy usage is a key issue as the sensors are typically tiny and wireless with limited memory and functionality given the fact that the batteries(y) have a limited power supply hence difficulties arise during computation(4). The communication links in such an unpredictable environment (node failures) are kept functional by applying a robust routing algorithm. The Ant system is a learning algorithm with charac- teristics such as robustness and versatility solves any NP hard communication problem. In this paper a novel approach is proposed. The agents use the sensor node's information to determine patterns to predict or anticipate changes in the environment with respect to time. The swarm agents are distributed along the network, where the agent communicates with its neighbors (agents). The current information on energy usage, prediction and decisions concerning environmental situation assessment are made possible by these swarm agents. Another optimization issue is the communication delay. During individual node failure, the swarm routing, unlike some other types of routing, automatically reroutes messages around node failures. The only data lost is data that was last prepared by the node or collected and processed by the failed node. The new route is deter- mined by applying the link status to the ant routing algorithm and other factors depending on the performance parame- ters including energy and distance. An evolutionary algorithm may not always provide the best global solution but does find the best local decision. Optimality(5), finding the solution that finds the best performance, and reachability(5), the global optimal is found instead of the local optimal, are the two important factors in choosing an appropriate algorithm. The focus of this paper, is to determine patterns using sensor node's information to predict or anticipate changes in the environment with respect to time. In the second section, the justification for using ant system and its impact on the sen- sor network is discussed. Section 3 discusses data mining and its importance in making predictable decisions to improve the performance of the network. Simulation results provide an insight of real-time situation assessment are provided in the fourth section. The paper concludes with the fifth section discussing the conclusion and future work.