Center for Embedded Network Sensing

The information acquisition performance of a sensor network is critical to all applications based on it. This performance depends on factors which cannot be completely known at design or deployment time: sensing medium characteristics and the phenomenon distribution. Simplifying assumptions such as the homogeneous nature of sensing media do not hold in most practical scenarios due to the presence of sensing obstacles. Further, the medium and phenomena may change over time. We propose to use controlled mobility to enhance coverage at run time in an autonomous manner. However, extensive robotic capabilities and supporting services such as precise navigation may be infeasible in large scale sensor networks. We present feasible alternatives for physical reconfiguration using low complexity and low energy actuation. The key contribution of the paper is to show that even small degrees of actuation can lead to a significant coverage advantage. We also compare this approach to conventional means for achieving equivalent coverage by increasing node density without actuation. Further, we discuss the relevant trade-offs which affect the use of mobility in terms of the time required for actuation.