In this paper, we establish the connections of the fundamental limitations in feedback communication, estimation, and feedback control over Gaussian channels, from a unifying perspective for information, estimation, and control. The optimal feedback communication system over a Gaussian necessarily employs the Kalman filter (KF) algorithm, and hence can be transformed into an estimation system and a feedback control system over the same channel. This follows that the information rate of the communication system is alternatively given by the decay rate of the Cramer-Rao bound (CRB) of the estimation system and by the Bode integral (BI) of the control system. Furthermore, the optimal tradeoff between the channel input power and information rate in feedback communication is alternatively characterized by the optimal tradeoff between the (causal) one-step prediction mean-square error (MSE) and (anti-causal) smoothing MSE (of an appropriate form) in estimation, and by the optimal tradeoff between the regulated output variance with causal feedback and the disturbance rejection measure (BI or degree of anti-causality) in feedback control. All these optimal tradeoffs have an interpretation as the tradeoff between causality and anti-causality. Utilizing and motivated by these relations, we provide several new results regarding the feedback codes and information theoretic characterization of KF. Finally, the extension of the finite-horizon results to infinite horizon is briefly discussed under specific dimension assumptions (the asymptotic feedback capacity problem is left open in this paper).
In cite{butman1976} the linear coding scheme is applied, $X_t =g_tBig(Theta - {bf E}Big{ThetaBig|Y^{t-1}, V_0=v_0Big}Big)$, $t=2,ldots,n$, $X_1=g_1Theta$, with $Theta: Omega to {mathbb R}$, a Gaussian random variable, to derive a lower bound on the feedback rate, for additive Gaussian noise (AGN) channels, $Y_t=X_t+V_t, t=1, ldots, n$, where $V_t$ is a Gaussian autoregressive (AR) noise, and $kappa in [0,infty)$ is the total transmitter power. For the unit memory AR noise, with parameters $(c, K_W)$, where $cin [-1,1]$ is the pole and $K_W$ is the variance of the Gaussian noise, the lower bound is $C^{L,B} =frac{1}{2} log chi^2$, where $chi =lim_{nlongrightarrow infty} chi_n$ is the positive root of $chi^2=1+Big(1+ frac{|c|}{chi}Big)^2 frac{kappa}{K_W}$, and the sequence $chi_n triangleq Big|frac{g_n}{g_{n-1}}Big|, n=2, 3, ldots,$ satisfies a certain recursion, and conjectured that $C^{L,B}$ is the feedback capacity. In this correspondence, it is observed that the nontrivial lower bound $C^{L,B}=frac{1}{2} log chi^2$ such that $chi >1$, necessarily implies the scaling coefficients of the feedback code, $g_n$, $n=1,2, ldots$, grow unbounded, in the sense that, $lim_{nlongrightarrowinfty}|g_n| =+infty$. The unbounded behaviour of $g_n$ follows from the ratio limit theorem of a sequence of real numbers, and it is verified by simulations. It is then concluded that such linear codes are not practical, and fragile with respect to a mismatch between the statistics of the mathematical model of the channel and the real statistics of the channel. In particular, if the error is perturbed by $epsilon_n>0$ no matter how small, then $X_n =g_tBig(Theta - {bf E}Big{ThetaBig|Y^{t-1}, V_0=v_0Big}Big)+g_n epsilon_n$, and $|g_n|epsilon_n longrightarrow infty$, as $n longrightarrow infty$.
This paper considers the problem of secret communication over a multiple access channel with generalized feedback. Two trusted users send independent confidential messages to an intended receiver, in the presence of a passive eavesdropper. In this setting, an active cooperation between two trusted users is enabled through using channel feedback in order to improve the communication efficiency. Based on rate-splitting and decode-and-forward strategies, achievable secrecy rate regions are derived for both discrete memoryless and Gaussian channels. Results show that channel feedback improves the achievable secrecy rates.