This proposal presents a career development plan for integrating research and education. The research component of this program focuses on the study of collaborative signal and information processing (CSIP) algorithms as well as the supporting computing models and protocols in sensor networks. The educational component is aimed at enhancing the newly created computer engineering program of the department through innovative curriculum development, active student mentoring, and by incorporating research findings into classroom teaching. The project consolidates research and educational efforts, responds to the unique challenges presented by sensor networks, and contributes significantly to the PI's career development. A sensor network forms a loosely-coupled distributed environment where collaborative processing among multiple sensor nodes is essential in order to compensate for each other's limited capability in sensing, processing, power supply, and to tolerate faults. The extremely constraint resources of sensor networks have presented unique challenges to CSIP, the biggest of which is the contradictory requirements between energy efficiency and fault tolerance. While energy-efficient approaches try to limit the redundancy such that minimum amount of energy is required for fulfilling a certain task, redundancy is needed for providing fault tolerance since sensors might be faulty, malfunctioning, or even malicious. A balance has to be struck between these two objectives. Intellectual Merit: An integrated research plan is proposed that tackles the unique challenges presented by sensor networks. This plan concerns not only the development of effective processing algorithms, it also studies the design of supporting computing paradigms and protocols which play an important role in facilitating the collaborative processing. In particular, the research plan focusses on three themes: First, the design, evaluation and implementation of a new paradigm, the mobile-agent-based paradigm, for realizing collaborative processing in sensor networks. In this model, instead of each sensor node sending local information to a processing center for integration, as is typical in a client/server-based computing, the integration code is moved to the sensor nodes through mobile agents. We discuss the potential of mobileagent- based collaborative processing in providing progressive accuracy while maintaining certain degree of fault tolerance. Second, the development and evaluation of a decentralized reactive clustering protocol to help adapt to the changing environment. Due to the sheer amount of nodes deployed, collaboration is usually carried out among nodes within the same cluster. We propose a decentralized reactive clustering (DRC) protocol in which the clustering procedure is initiated only when events are detected. Performance gain in terms of energy consumption and network lifetime is analyzed. Third, the study of distributed optimization for in-network data processing. We tackle the challenging CSIP problems including multiple target/event detection and unknown target/event identification and develop decentralized algorithms that achieve both energy-efficiency and bandwidth-efficiency advantages, as well as detection performance gain compared to its centralized counterpart. Broader Impact: The PI is dedicated to excellence in teaching and enthusiastic student mentoring. CSIP in sensor networks, as one of the areas that integrate multidisciplinary characteristics, will be used as an experimental concentration area to initiate innovative curriculum design, which helps bring in unique features to increase the quality and visibility of the newly established computer engineering program of the department. Although sensor networks have revealed great potential in CSIP, the extremely constraint resources have largely limited their practical deployment. The performance evaluation of computing models, protocols, and processing algorithms for CSIP would help understand the capability of sensor networks as well as provide theoretical guide in the design of CSIP algorithms. Both the research and educational results will be made available through the Twiki collaboration platform on the Internet. 1
