Merging of multistep predictors for decentralized adaptive control

Decentralized adaptive control is based on the use of many local controllers in parallel, each of them estimating its own local model and pursuing local aims. When each controller designs its strategy using only its model, the resulting control will be suboptimal since local models do not allow prediction of consequences of actions of the neighbors. We use probabilistic formulation of adaptive control to build predictive densities of future outputs. Mutual exchange of these densities on commonly observed variables is proposed to compensate for incompleteness of the local models. The task is to find a procedure how to use such information withing the control strategy design under the constraint that the resulting design procedure is of the same complexity as the one without the exchange. We present an approximate algorithm and illustrate its performance on a simple example.