Throughput/energy aware opportunistic transmission control in broadcast networks

We address throughput/energy aware opportunistic transmission control in broadcast networks, under incomplete channel state information. The channels are modeled as i.i.d two-state Markov chains and the transmission controller makes transmit/idle decisions based on past 1-bit feedback from the broadcast users. With a reward structure that captures the trade-off between sum throughput gain (transmit) and energy savings (idle), we formulate the transmission control problem as an infinite horizon, discounted reward, partially observable Markov decision process. For special cases of the system parameters, we show that the optimal control policy is either greedy or partially greedy. For the general case, we follow an indirect approach towards the control problem. We first establish thresholdability properties of the optimal control policy in a two-user broadcast. We then extrapolate these properties to the general broadcast and propose a simple threshold control policy. Extensive numerical results suggest near-optimal performance of the proposed policy. In addition, the proposed threshold policy is easy to implement with complexity being polynomial in the number of broadcast users.