Research On Collaborative Signal Detection In Energy Harvesting Wireless Sensor Networks

The wireless sensor network system usually deploys a large number of sensor nodes in the surveillance area. Distributed detection is one of its main applications. Each sensor node collects and sends the data to the fusion center. The fusion center makes the optimal decision fusion based on some rules. Due to the redundancy of the data from different sensor nodes, the multiple-sensor system can achieve better performance than the single-sensor system. Each sensor node is usually equipped with limited battery and the limited energy supply becomes the key factor which affects the overall performance of the sensor network system. With the development of new energy technologies and microelectronic technologies, energy harvesting sensor networks can be applied in practice. With the energy harvesting unit, each sensor node can harvest and store energy from the environment in order to maintain its operation. However, due to the randomness of the harvesting energy, how to schedule the harvesting energy and improve the overall performance becomes the main concern of energy harvesting systems.Considering the distributed detection problem in energy harvesting sensor networks, this paper analyzes the mathematical model of signal detection, the overall detection performance indexes and the mathematical model of harvesting energy. The analysis method of the energy scheduling optimization problem is also proposed. Both the finite horizon case and infinite horizon case are considered in this paper. Considering the finite horizon case, both the offline and the online energy scheduling strategies are proposed. For the offline energy scheduling strategy, the harvesting energy can be considered to be known, and the energy scheduling problem can be formed as an optimization problem and the strategy can be obtained through solving the optimization problem. For the online energy scheduling problem, the energy scheduling strategy can be solved by using dynamic programming. This paper also proposes a suboptimal energy scheduling strategy which could reduce computational complexity. Considering the infinite horizon case, the energy scheduling problem can be formed as a Markov decision process. Through the value iteration method, the optimal energy scheduling strategy can be obtained. Besides, a suboptimal energy scheduling strategy is proposed for the infinite horizon case.In addition to this, the decision fusion methods in the distributed detection problem need the parameters of each sensor node. However, the parameters of the low-precision sensor nodes are usually hard to be known in practice. This paper proposes an iterative estimation strategy of the sensor parameters. Without the knowledge of the sensor parameters, the parameters of the sensor nodes can be iterative estimated through using the fusion rules. Considering the heterogeneity of sensor networks, this paper also proposes an adaptive fusion strategy by using the high-precision guidance sensor. Through using the high-precision guidance sensor, the parameters of the low-precision sensors can be obtained through on-line estimation, and the fusion rules can be used to make the fusion decision.