Autonomous Market-Based Approach for Resource Allocation in A Cluster-Based Sensor Network

We consider the resource allocation problem in a structured sensor network. While new technologies are making sensors smarter, smaller, and cheaper, an emerging problem is how to allocate limited energy, radio bandwidth, and other resources to achieve efficient global behavior for high performance, QoS, and long network lifetime. Conventionally, resource allocation is treated as an optimization problem. The solution is calculated at each round of scheduling according to the status of all resources and given tasks in a centralized manner, which is very computation and communication intensive and not suitable for multi-hop sensor networks. Recently, some distributed approaches with less computation and communication complexity have been reported. Most of these approaches are completely decentralized without using the advantage of underlying network structures. A large scale sensor networks is usually built with a hierarchical and reconfigurable structure that introduces efficient sensing, computing and networking. In this paper, we propose a hierarchical framework for the resource allocation in a cluster-based sensor network. The framework combines decentralized control scheme with local centralized control scheme. In each cluster, there is a centralized agent that can optimally allocate the resources in the cluster, while in each node there is a decentralized agent that manages the resources at the node. Instead of low-level sensor programming, such as manually tuning sensor and other resource usage, we explore market approach for dynamic allocation of system resources. Network customers can use the price of resources to loosely control the global behavior of the sensor network. All radio transmissions are supported by the routing protocol and reconfiguration function of the underlying cluster-based sensor network. We implement our approach to the task of mobile target tracking. Experiment results show that our approach promises a faster and more accurate tracking. Furthermore, it can significantly extend the network lifetime.