Coverage problems in mobile sensing

Abstract

Sensor-networks can today measure physical phenomena at spatial and temporal scales that were not achievable earlier, and have shown promise in monitoring the environment, structures, agricultural fields and so on. A key challenge in sensor-networks is the coordination of four actions across the network: measurement (sensing), communication, motion and computation. The term coverage is applied to the central question of how well a sensor-network senses some phenomenon to make inferences. More formally, a coverage problem involves finding an arrangement of sensors that optimizes a coverage metric. In this thesis we examine coverage in the context of three sensing modalities. The literature on the topic has thus far focused largely on coverage problems with the first modality: static event-detection sensors, which detect purely binary events in their immediate vicinity based on thresholds. However, coverage problems for sensors which measure physical quantities like temperature, pressure, chemical concentrations, light intensity and so on in a network configuration have received limited attention in the literature. We refer to this second modality of sensors as estimation sensors; local estimates from such sensors can be used to reconstruct a field. Third, there has been recent interest in deploying sensors on mobile platforms. Mobility has the effect of increasing the effectiveness of sensing actions. We further classify sensor mobility into incidental and intentional motion. Incidentally mobile sensors move passively under the influence of the environment, for instance, a floating sensor drifting in the sea. We define intentional mobility as the ability to control the location and trajectory of the sensor, for example by mounting it on a mobile robot. We build our analysis on a series of cases. We first analyze coverage and connectivity of a network of floating sensors in rivers using simulations and experimental data, and give guidelines for sensor-network design. Second, we examine intentional mobility and detection sensors.

(cont.) We examine the problem of covering indoor and outdoor pathways with reconfigurable camera sensor-networks. We propose and validate an empirical model for detection behavior of cameras. We propose a distributed algorithm for reconfiguring locations of cameras to maximize detection performance. Finally, we examine more general strategies for the placement of estimation sensors and ask when and where to take samples in order to estimate an unknown spatiotemporal field with tolerable estimation errors. We discuss various classes of error-tolerant sensor arrangements for trigonometric polynomial fields.