This is a basic research program in detection and estimation for multiple sensors with communication constraints. Communication constraints arise naturally whenever sensors are located at physically separated sites, or when data must be stored before final processing. The research focuses on optimal distributed detection when statistics are uncertain and possibly time-varying, and on optimal quantizer design for parameter estimation. The uncertain statistics will be modeled parametrically, either in the received signal, the noise, or in unknown channel transition probabilities. The optimal detection system will be adaptive, with adaptation occurring locally as well as globally. The project will address the ability of the system to adapt to unknown parameters and to converge asymptotically to the optimal design. Optimal quantizer design for nonlinear parameter estimation will be examined. Quantizer design in the context of distributed detection and estimation is fundamental when communication links have limited capacity. Optimal quantizer design ensures that only the most essential information for parameter estimation is shared. The resulting optimal quantizer will differ from the Lloyd-Max quantizer because the goal here is parameter estimation. The project will address parameter estimation with communication constraints particularly for the multiterminal situation where no one sensor can make a useful individual estimate of the unknown parameter.
