Performance and optimization of dynamic-TDD intercell coordination with elastic traffic

We study the intercell coordination problem between two interfering cells combined with dynamic TDD serving elastic data traffic. We model the system at the so-called flow level, where the number of flows varies dynamically and each flow has a random service requirement. Due to the interference between the stations the system is modeled as a set of four interacting processor sharing queues. Our objective is to consider, by using several approaches, the gains from dynamic policies that utilize instantaneous state information to minimize the total mean number of flows. Assuming that the capacity is shared dynamically according to the notion of balanced fairness yields a general policy that is also Pareto optimal, and we are able to explicitly analyze its performance. For exponential service times, we also apply the theory of Markov decision processes and the policy iteration algorithm to minimize the number of flows in the system. Finally, we define priority policies for certain special cases and show that they are even stochastically optimal. Our numerical results show that the gains from the dynamic policies compared with a statically optimized policy can be significant. Surprisingly, the gains can be achieved largely by our novel robust heuristic policy that only relies on instantaneous information about the number of flows.