Collaborative signal and information processing (CSIP) in distributed sensor networks is an emerging research area, drawing upon traditionally disparate disciplines such as lower-power communication and computation, space-time signal processing, distributed and fault tolerant algorithms, adaptive systems, and sensor fusion and decision theory.
Recent advances in wireless networking, microfabrication (e.g. MEMS), and embedded processing have enabled a new generation of sensor networks for a wide range of tracking and identification problems in both civilian and military applications, including human-aware environments, intelligent transportation grids, factory condition-based monitoring and maintenance, and battlefield situational awareness. However, unlike centralized sensor platforms, distributed sensor nets are characterized by limited battery power, frequent node attrition, and variable data and communication quality. To scale up to more realistic tracking and classification applications involving tens of thousands of sensors, heterogeneous sensing modalities, multiple targets, and non-uniform spatio-temporal scales, these systems have to rely primarily on collaboration among distributed sensors to support energy-aware and low-latency computation and communication.
The PARC Collaborative Sensing Project has taken a systemic approach to address key CSIP issues of representing, processing, storing, and querying spatially distributed, multi-modal information from a sensor field. To extract reliable and timely information from a sensor field, CSIP must suport cross-node data aggregation, asynchronous execution, and progressive accuracy. We have developed information driven sensor querying (IDSQ) as a mechanism for mediating between sensor data and user queries in a network. IDSQ dynamically tasks sensor nodes to maximize information gain while minimizing energy and bandwidth consumption, thereby blurring the traditional separation between network layers and applications running on the network. In a recent DARPA field test in the Mojave desert involving 70+ wireless sensor nodes, IDSQ was successfully demonstrated to support a real-time, decentralized vehicle tracking application. In this talk, I will describe the IDSQ theory and algorithms, as well as experimental results from the field and lab testbeds.