Abstract
Generally in a large and complex system each controller is unable to get perfect information on the system state, and he has different information from the other controllers. By transmitting his information to the others and communicating it with each other, better performance can be achieved.
Here a linear-quadratic-Gaussian-team problem is considered. The nonclassical information pattern for the team problem may be converted to the classical or to the different types of nonclassical information pattern by transmitting or communicating the information. The optimal linear control gains and the optimal linear estimator gains for each information pattern are derived. In the team problem the optimal linear control laws do not satisfy the separation theorem. A relative value of information for two different information patterns is defined. Compared with the information cost, the relative value of information is utilized to design the information pattern by transmitting or communicating the information. The interplay between the state and the control are discussed for the case where the transmission or communication are corrupted by noise.
