On estimation in networked control systems with random delays and partial observation losses

The paper considers state estimation in networked control systems (NCS) where observations from multiple sensors are subject to random delays and packet losses. We derive the minimum error covariance estimator and the optimum estimator with constant gains as a low-complexity solution. Generalizations to account for the effects of measurements quantization and limited transmission bandwidth are investigated for a stable system. Assuming a simple scalar system, we show how the proposed framework can be exploited for the design of NCS. In particular we investigate i) cross-layer optimization of quantization processes and network resource allocation and ii) comparison between single-hop and multi-hop communication protocols. We show that simple BPSK and single-hop communication protocols provide close to optimum performance in applications dealing with state estimation of a stable system.