“The Value of Information in Military Networks”
Dr. J. Willard Curtis, U.S. Air Force Research
Traditional measures of information do not depend upon the context or relevance of data to the control system’s purpose. Measuring the value of information, from a control system perspective, would allow better bandwidth and storage utilization in networks of autonomous systems.
The general problem is NP-hard, so we seek current problems where informational value can be computed to enhance closed-loop performance beyond the state-of-the-art. Specifically, we consider an encoder-decoder pair separated by a communication channel where the number of uses of the channel is limited. The optimal decoder transmission policy, for the scalar case, was recently discovered. However this policy is optimal only with respect to mean-squared estimation error. We seek to find the optimal encoder policy given knowledge of the control system which is driven by the decoder’s estimate. Essentially this requires discerning the value of the sensor information from the controller’s perspective.
J. Willard Curtis received his bachelor’s degree in applied and engineering physics from Cornell University. He went on to study electrical and computer engineering at Brigham Young University, receiving a PhD in 2002. Since then he has been with the Munitions Directorate of the Air Force Research Laboratory at Eglin Air Force Base, where he is the technical adviser for the Guidance and Control Branch. His research interests are in multi-agent and cooperative control, decentralized Bayesian estimation and networked control systems.