Wireless sensor networks: a new computing platform for tomorrow's Internet

Summary form only given. Due to the rapid convergence of MEMS devices, ubiquitous connectivity, and low-power embedded processing, wireless sensor networks are emerging as an entirely new computing platform that promises to seamlessly couple the digital world with the physical world around us. The envisioned applications are societal-scale, ranging from transportation, manufacturing, healthcare, to environmental preservation. To cope with the enormous challenges of designing and maintaining such a massively distributed information fabric, we must address systemic issues such as networking, infrastructure establishment, collaborative signal processing, device tasking and control, data management, as well as programming tools. I described the recent progress in wireless sensor networks and applications, drawing examples from our own work as well as that of others. I started with the technical challenges posed by the severe resource constraints on power and bandwidth, the fragile and dynamic wireless connectivity, as well as the distributed and concurrent nature of the application programs, to illustrate the point that these systems must be designed with a holistic, cross-layer approach. As a concrete example, I described how distributed agents collaboratively seek, process, and aggregate information in a resource-constrained environment. These agents employ a decentralized decision-making, knows as the IDSQ protocol, to optimize for information utility while keeping the cost low. I then present recent work in sensor data management. Using in-network intelligence, I showed the dramatic improvements in the reduction of information needed to answer high-level queries. Next, I described significant progress the research community has made in developing programming abstractions and tools for the development of sensor network applications. In one approach, nodes are programmed as collectives, maintaining and migrating information states as required by the application. The goal is to significantly lessen the burden of managing lower-level network events on application developers. I concluded with an outlook for the future directions and problems that we must tackle in order to make societal-scale sensor network applications a reality.