Foundations of Control and Estimation Over Lossy Networks
read more
Citations
Distributed Kalman filtering for sensor networks
Secure Estimation and Control for Cyber-Physical Systems Under Adversarial Attacks
For the Grid and Through the Grid: The Role of Power Line Communications in the Smart Grid
Control and Communication Challenges in Networked Real-Time Systems
A secure control framework for resource-limited adversaries
References
A survey on sensor networks
Neuro-Dynamic Programming.
Neuro-dynamic programming
Kalman filtering with intermittent observations
Control under communication constraints
Related Papers (5)
Frequently Asked Questions (15)
Q2. What are the future works mentioned in the paper "Foundations of control and estimation over lossy networks" ?
The authors are exploring this possibility. This suggests that controller design needs to be substantially reconsidered for such systems. This suggests that communication protocols targeted to networked control systems need to be developed.
Q3. What is the optimal control in the special case?
In the special case where the state is completed observed (C is invertible and there is no output noise i.e., R = 0), the optimal control is indeed linear.
Q4. What is the goal of this paper?
The goal of this paper is to design optimalLQG controllers and to estimate their closed-loop performance for both TCP-like and UDP-like protocols.
Q5. What is the expected value of the matrices Pk+1|k+1?
since the matrix Pk+1|k+1 is a nonlinear function of the previous time step matrix covariance Pk|k, as can be observed from Equations (11) and (15), the exact expected value of these matrices, Eγ [Pk|k], cannot be computed analytically, as shown in [33].
Q6. What are the two general protocols used in the Internet?
The well known Transmission Control (TCP) and User Datagram (UDP) protocols used in the Internet are specific examples of their more general notion of TCP-like and UDP-like communication protocol classes.
Q7. What is the compensation scheme for the loss packets between the controller and the actuator?
The stochastic variable νk models the loss packets between5 the controller and the actuator: if the packet is correctly delivered then uak = u c k, otherwise if it is lost then the actuator does nothing, i.e. uak = 0.
Q8. What is the critical probability for the TCP-like protocols?
For the same system the authors have pmin = pmax = 1 − 1/|A|2 = 0.173, therefore the critical probability for the TCP-like protocols is γc = νc = pmin as stated in Theorem 5.5.
Q9. what is the critical observation arrival probability?
Then there exists a critical observation arrival probability γc, such that the expectation of estimator error covariance is bounded if and only if the observation arrival probability is greater than the critical arrival probability, i.e.Eγ [Pk|k] ≤M ∀k iff γ̄ > γc. where M is a positive definite matrix possibly dependent on P0.
Q10. What is the controller for UDP-like protocols?
Although the true LQG optimal controller for UDP-like protocols is time-varying and hard to compute, the authors might choose to determine the optimal time-invariant LQG controller.
Q11. What is the way to show that the TCP-like case is a linear system?
The authors show that, for the TCP-like case, the classic separation principle holds, and consequently the controller and estimator can be designed independently.
Q12. What is the value function for the optimal LQG using ACK-based protocols?
Since J∗N (x̄0, P0) = V0(x0), from the lemma it follows that the cost function for the optimal LQG using ACK-based protocols is given by:J∗N = x̄ ′ 0S0x̄0 + trace(S0P0) + PN−1 k=0 trace(Sk+1Q)++ PN−1k=0 trace((A ′Sk+1A + Wk − Sk)Eγ [Pk|k])(27)where the authors used the fact E[x′0S0x0] = x̄ ′ 0S0x̄0 + trace(S0P0), and Eγ [·] explicitly indicates that the expectation is calculated with respect to the arrival sequence {γk}.
Q13. What is the way to estimate the arrival process?
In [9], Smith et al. considered a suboptimal but computationally efficient estimator that can be applied when the arrival process is modeled as a Markov chain, which is more general than a Bernoulli process.
Q14. What is the reason why the optimal controller ignores the observation yk?
This is because at times when packets are not delivered, the optimal estimator ignores the observation yk, therefore its value is irrelevant.
Q15. What is the approach to analyze MIMO systems?
This approach allows analysis of Multiple Input Multiple Output (MIMO) systems with many different controller and receiver compensation schemes [20], however, it does not include process and observation noise and the controller is restricted to be time-invariant, hence sub-optimal.