1 of 6 978-1-4244-2677-5/08/$25.00 ©2008 IEEE SENSING AND CONTROL IN BANDWIDTH-LIMITED SYSTEMS: A KALMAN FILTER APPROACH Gregory L. Barnette System Engineering Division Engineering Directorate Air Armament Center Eglin AFB, FL John M. Shea Wireless Information Networking Group Department of Electrical & Computer Engineering University of Florida Gainesville, FL Warren E. Dixon Nonlinear Controls and Robotics Mechanical and Aerospace Engineering Department University of Florida Gainesville, FL ABSTRACT Cooperative systems, like those found in networks of unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs), are typically distributed over large geographical areas. This geographic dispersion forces reliance on wireless communications to implement the control systems for these networks. As the number of nodes within the system grows, the bandwidth required to continuously update all nodes exceeds available bandwidth. This forces the system engineers to implement a prioritization scheme to determine which nodes will be updated during each frame. In this paper, we consider the design of protocols to prioritize the communications of sensing and control data. We consider a control system that uses a Kalman filter to supply optimal estimates of the positions for a group of maneuvering targets in the presence of measurement noise and lost data. The filter innovations provide estimates of the magnitude of statistical deviations from the state estimates, as well as deviations from planned profiles. Thus, we propose protocols that use the filter innovations to prioritize the sensing and control traffic. We compare the performance of this approach to a round-robin protocol and to the performance of a Kalman filter with no bandwidth constraints. We show that using the filter innovations offers a significant performance improvement over the round-robin protocol. This work was supported by the National Science Foundation under grant number CNS-0626863 and by the Air Force Office of Scientific Research under grant number FA9550-07-10456. BACKGROUND Cooperative systems, like those found in networks of unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs), are typically distributed over large geographical areas. This geographic dispersion forces reliance on wireless communications to implement the control systems for these networks. This type of system is commonly referred to as networked control system (NCS). In a recent survey of NCS research, Hespanha et al. [2] point out that NCS is a marriage of both control system and communications effects. Under normal circumstances, control theory would address the system’s stability and effective control with assumed perfect transmission of data between the controller and the system, while communications would address the effects of imperfect transmission of information over noisy media. The goal of NCS development is to optimize performance of the system under a combined set of constraints addressing both aspects of the system. As the number of nodes within the system grows, the bandwidth required to continuously update all nodes exceeds available bandwidth. Minimizing bit transmission and compression techniques may provide temporary fixes for marginal excesses [3,5,9]. However, prioritization techniques must be used to correct deficiencies that far exceed bandwidth capabilities. Numerous papers address the random dropping of packets within the system [2,4,8]. The analyses of random transmission dropouts provide upper bounds on techniques that employ an intelligent selection process. However, an intelligent selection process should provide