Abstract:

As the field of sensor networks matures, research in this area become more and more focused on real-world applications and scenarios. Moreover, with the proliferation of mobile devices, applications are increasingly taking advantage of emerging mobile platforms. Sensor network applications share the common, but seemingly contradictory, requirements of increasing functionality and longevity. While most sensor network research has focused on dense networks with fixed nodes, these previous works are not always applicable to mobile networks. Even densely deployed mobile nodes can move to areas where there are few nodes, or to locations outside the intended deployment area, resulting in unpredictable energy profiles. In addition, in areas of low connectivity, many of the previously proposed techniques become highly inefficient or simply fail.

This thesis presents a vertical approach (hardware, middleware, service layers) to increasing the longevity of mobile sensor systems. On the hardware level, I present local techniques that increase energy efficiency by up to 96%. The middleware further improves operational efficiency by taking advantage of characteristics in idle and active states. It provides efficient energy control, error recovery, and hardware abstractions.

To improve efficiency in determining the physical location of sensor nodes, LOw-density Collaborative Ad hoc Location Estimation (LOCALE) is presented in detail. This delay-tolerant collaborative service reduces the reliance on per-node GPS by exchanging only 20 bytes of information when nodes meet, and tracking movements with hardware when nodes are not in range. With the support of movement tracking in the lower layers, this vertical approach reduces infrastructure needs by up to 64X, increases accuracy by more than 21X and drastically reduces energy consumption by more than 150X.

In addition, my dissertation also explores policies that improve system efficiency through adaption. These policies reduce energy variations experienced by mobile nodes, and further increase overall system efficiency. Through both simulations and test bed implementations, I show that the effective system lifetime is prolonged, while at the same time data packets received increase by more than 50% compared to non-collaborative methods. Allowing collaborative adaption methods achieves an additional 30% improvement in data reception.

The techniques presented in this thesis drastically improve overall mobile sensor system performance, especially in sparse conditions. These local, collaborative, and adaptive techniques not only increase system longevity , but also improve system functionality . As these techniques are applicable to a wide range of networks (from very dense to very sparse network scenarios), they can be employed in most mobile systems, and useful in the development of future mobile systems.