On output selection for a large scale networked system with constraints

Based on properties of the Kalman filter, a distributed sensor placement method is proposed for output selection in a large scale networked system with sensor location constraints and/or performance constraints. This method only utilizes parameter matrices of the subsystem and its connection matrices with other subsystems to evaluate all the possible sensor locations of the large scale networked system. An attractive property of this method is that its computational efficiency is much higher than the greedy algorithm's. In addition, a lower and upper bound for the trace of the state estimation error covariance of the Kalman filter is determined to reveal the fundamental design limitations and structural design limitations. The results of numerical simulations confirm the effectiveness and the high computational efficiency of the method.