Cost-efficient collection and delivery of sensor data using mobile devices

Wireless sensor networks represent an important component of distributed pervasive computing infrastructure supporting a range of applications including health,military, environmental monitoring, civil structure monitoring, smart homes, etc. The primary factor driving such pervasive real-world applications is availability of data from sensors distributed in the environment. Traditional way of collecting data is to transmit the data from sensors to a collection point using wireless radio communications. However, the traditional approach is expensive and not always efficient. This thesis addresses a major challenge of cost-efficient collection of data from wireless sensor networks. Our data collection philosophy is to use mobile devices as sensor data collectors. The use of mobile devices as mobile data mules facilitates the formation of a mobile access network that can be used by sensors to connect to the external world. We propose, investigate, develop and validate a sensor data collection framework called sGaRuDa which enables interoperable capabilities and takes advantage of existing communication and hardware capabilities of the mobile data mule platforms enabling them to collect sensor data on-the-run. The sGaRuDa framework incorporates intelligent mobile data mule allowing them to dynamically make data collection and delivery decisions. The sGaRuDa framework and the corresponding data collection algorithms are targeted at sensor networks that use short range radio communication technologies like Bluetooth. We have also proposed, implemented and validated a novel three dimensional k-Nearest Neighbour query-based sensor data collection approach called 3D-KNN to address broadcast-based sensor network communication architectures. The 3D-KNN facilitates multi-hop data collection from infrastructure-less wireless sensor networks (e.g. Zigbee). We propose, develop, implement and validate a dynamic smart spaces modelling approach called Ranked-Context Spaces (R-CS). Our smart spaces modelling approach is driven by the notion of situation modelling and reasoning about context. Ranked-Context Spaces is capable of computing situation-based smart spaces model taking into account changing contextual information. R-CS is proposed as an extension to Context Spaces theory. The thesis presents implementation and evaluation details of the proposed sGaRuDa framework and the 3D-KNN algorithm. We have demonstrated the feasibility and cost-efficiency of the sGaRuDa system framework in real-world environments by implementing a proof-of-concept prototype on a range of mobile device platforms, namely, Personal Digital Assistants and mobile robot. Extensive evaluation and experimentation have been performed to prove the extent of energy conservation using the proposed data collection framework and the 3D-KNN algorithm. Finally, we have implemented the R-CS system to demonstrate its reasoning ability under uncertainty. Experiments based on synthetic sensor data streams have been performed to evaluate the proposed Context Spaces extensions incorporated into R-CS. During the course of the thesis work, 7 peer-refereed international conference papers, 1 peer-refereed workshop paper and 1 journal paper have been produced. One of the conference papers was awarded a BEST PAPER AWARD.