Information Fusion, Power Optimization and Novel Modulation Schemes for Wireless Sensor Networks

This dissertation investigates information fusion and optimization of analog and digital wireless sensor networks. Also two modulation schemes, which can be used in wideband sensor networks, are proposed and studied.;Two novel modulation schemes for wideband sensor network applications are proposed and analyzed. DTM (delay time modulation) uses the delay time of a stochastic delay differential equation system (SDDE) to modulate input signals. Original signals can be easily recovered by measuring the variance of received baseband signals. It has the advantages of simple structure and low cost. DTM scheme is also used as an example to investigate wideband analog sensor networks. 10M (initial condition modulation) modulates information signals into the initial conditions of chaotic generators. Receivers can then trace back the initial conditions by using contractive properties of chaotic maps, and hence retrieve the transmitted data. The scheme can achieve very high data rate at the cost of hardware complexity. Both of the proposed schemes are analyzed in terms of noise performance, numerical experiments are carried out to verify the feasibility of the schemes.;Analog sensor networks are studied first. Fusion rules that aim to obtain better estimation of unknown signals from multiple physical sources are examined. Mean distortion level reflects average squared error between original signals and their estimates and is used as an index to measure fusion performance. Optimal transmission power allocation strategy is considered when sensor networks are under energy or distortion constraint. Information fusion and power optimization of digital sensor networks are studied next. The mean distortion levels of digital sensor networks are derived for different fusion methods. Data delay is included in the optimization process of digital sensor networks, since the sensed data are time-sensitive in many applications. Fusion performance and transmission power optimization of analog and digital sensor networks in narrowband and wideband applications are analyzed. Because analog sensors do not need analog to digital converters (ADCs), analog sensor networks may have the potential of low energy usage, especially when the sampling rates of ADCs are high. Energy consumptions of two types of sensor networks are then compared by numerical experiments. Also, with the same energy used for data transmission, the performances of two sensor networks are simulated and compared.