Delayed feedback capacity of finite-state machine channels: upper bounds on the feedforward capacity

Abslracl The delayed feedback capacity of a finitestate machine channel is achieved by a Markov source whose memory length is equal to the sum of the channel memory length and the feedback delay. Further, the optimal feedback is achieved by computing the causal posterior distribution of the Markov source state. We formulate the optimization of the Markov source as a stochastic optimal control problem, and solve it by dynamic programming. The delayed feedback capacity is a tight upper bound on the feedforward capacity.