Predictive Compensation in Finite-Horizon LQ Games under Gauss-Markov Deviations

This paper presents a predictive compensation framework for finite-horizon discrete-time linear quadratic dynamic games in the presence of Gauss-Markov deviations from feedback Nash strategies. One player experiences correlated stochastic deviations, modeled via a first-order autoregressive process, while the other compensates using a predictive strategy that anticipates the effect of future correlation. Closed-form recursions for mean and covariance propagation are derived, and the resulting performance improvement is analyzed through the sensitivity of expected cost.