Quality of Information Aware Data Delivery in Wireless Sensor Networks

Efficient energy use is paramount if sensor networks are to function for long periods.Already, researchers have delivered many efficiency schemes with various levels ofsuccesses; however, there is still need for improvements. This thesis is an attempt tomake further improvements on energy efficiency within wireless sensor networks. Radio, sensing and light emitting diodes have dominated the power consumption oncurrent sensor network architectures. The radio consumes nearly as much energywhile it is listening as it does while transmitting. Naturally, if we reduce listening timesor if we reduce the number of messages transmitted then we would have reduced theenergy consumption. Some early successes in this area include power saving mediaaccess control protocols such as X-MAC that reduces the radio listening times bycycling through sleep and wake states and adaptive sampling techniques that havehelped to reduce the sensing and transmission numbers. This work is similar to theadaptive sampling techniques in that it focuses on saving energy by reducing thetransmission numbers. This thesis tackles the problem of energy efficiency from a Quality of Information(QoI) perspective. Data is delivered with the quality of information at the forefront ofdelivery decisions, thus, samples are taken at a rate so that important events are notmissed, then a decision is taken whether to transmit the packet or not. Further tothis, we implement a routing protocol that is information aware, allowing it to providea better quality of service to important packets. Our goal is to lower the number ofmessages without degrading the quality of information. This work is an improvement over static sampling methods because important eventscan be missed if the sampling rate is too low. We investigate our approach byimplementing a QoI aware routing algorithm based on the SPEED protocol and doevaluations using simulations and on a sensor network testbed with data from realdeployments. Our results show that it is possible to reduce the number of messages transmittedand to reconstruct the missing data at the Sink with high fidelity. We were able toachieve in some instances up to 75% message reduction in our temperaturemeasuring application with 94% of all errors falling below 0.5 degrees Celsius