Date of Award

Spring 2008

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Micro and Nanoscale Systems

First Advisor

Rastko R. Selmic

Abstract

Wireless Sensor Networks (WSNs) consist of a large number of sensors, which in turn have their own dynamics. They interact with each other and the base station, which controls the network. In multi-hop wireless sensor networks, information hops from one node to another and finally to the network gateway or base station. Dynamic Recurrent Neural Networks (RNNs) consist of a set of dynamic nodes that provide internal feedback to their own inputs. They can be used to simulate and model dynamic systems such as a network of sensors.

In this dissertation, a dynamic model of wireless sensor networks and its application to sensor node fault detection are presented. RNNs are used to model a sensor node, the node's dynamics, and the interconnections with other sensor network nodes. A neural network modeling approach is used for sensor node identification and fault detection in WSNs. The input to the neural network is chosen to include previous output samples of the modeling sensor node and the current and previous output samples of neighboring sensors. The model is based on a new structure of a backpropagation-type neural network. The input to the neural network (NN) and the topology of the network are based on a general nonlinear sensor model. A simulation example, including a comparison to the Kalman filter method, has demonstrated the effectiveness of the proposed scheme. The simulation with comparison to the Kalman filtering technique was carried out on a network with 15 sensor nodes. A fault such as drift was introduced and successfully detected with the modified recurrent neural net model with no early false alarm that could have resulted when using the Kalman filtering approach.

In this dissertation, we also present the real-time implementation of a neural network-based fault detection for WSNs. The method is implemented on a TinyOS operating system. A collection tree network is formed, and multi-hoping data is sent to the base station root. Nodes take environmental measurements every N seconds while neighboring nodes overhear the measurement as it is being forwarded to the base station for recording it. After nodes complete M and receive/store M measurements from each neighboring node, recurrent neural networks are used to model the sensor node, the node's dynamics, and the interconnections with neighboring nodes. The physical measurement is compared to the predicted value and to a given threshold of error to determine a sensor fault. The process of neural network training can be repeated indefinitely to maintain self-aware network fault detection. By simply overhearing network traffic, this implementation uses no extra bandwidth or radio broadcast power. The only costs of the approach are the battery power required to power the receiver for overhearing packets and the processor time to train the RNN.

Share

COinS