In sensor networks comprised of energy- and complexity-limited inexpensive sensors, the underlying information field is distributed, endowed with certain dynamics in terms of physical movement (for example: seismic, acoustic waves traveling over a geographical region) and delay-sensitive that warrant real-time reconstruction. Further, in such multi-terminal settings, ``bit'' is not the universal currency of information, as the separation of multiuser compression and communication is sadly suboptimal. In this project these two issues are addressed by considering a new dynamic compression model and a new multi-partite graph-based architecture for distributed transmission of delay-sensitive information.
In this new compression model the information available at certain nodes is changing as a function of time. This model is referred to as source coding with feedforward. This induces a completely new dynamism in the information compression problem. Further, in the architecture for transmission of distributed information considered in this project, multi-partite graphs are used as a discrete interface between distributed compression and distributed communication. This leads to a modular design of multi-terminal sensor-based information processing systems. Using these mechanisms, the goal of this project is to realize a self-organizing sensor network enabled with a new architecture that is capable of performing more efficient representation, communication and real-time reconstruction of distributed and delay-sensitive complex information fields than possible by any system today. This research effort is complemented by an educational effort to train young engineers to become skilled in construction, management and development of such massively complex information systems. The specific issues considered in this project are development of project work in the courses, integrated undergraduate education and research effort, and curriculum development with a course on distributed signal processing and communication.