Dynamic Packet Scheduler Optimization in Wireless Relay Networks

In this work, we investigate the optimal dynamic packet scheduling policy in a wireless relay network (WRN). We model this network by two sets of parallel queues, that represent the subscriber stations (SS) and the relay stations (RS), with random link connectivity. An optimal policy minimizes, in stochastic ordering sense, the process of cost function of the SS and RS queue sizes. We prove that, in a system with symmetrical connectivity and arrival distributions, a policy that tries to balance the lengths of all the system queues, at every time slot, is optimal. We use stochastic dominance and coupling arguments in our proof. We also provide a low-overhead algorithm for optimal policy implementation.

💡 Research Summary

The paper addresses the dynamic packet‑scheduling problem in a wireless relay network (WRN) composed of two layers of parallel queues: subscriber stations (SS) and relay stations (RS). Each queue receives an independent stochastic arrival stream, while the wireless links between SS and RS are randomly on or off in every time slot. The authors model the system state by a vector of all queue lengths and define a cost function C(Q)=∑ f(Qi), where f is any non‑decreasing function (e.g., linear or quadratic). The objective is to find a scheduling policy that minimizes the stochastic process {C(Q(t))} in the sense of stochastic ordering, i.e., the cost under the optimal policy should be dominated by the cost under any other admissible policy at every time instant.

A central contribution is the proof that, under a symmetry assumption—identical numbers of SS and RS queues, identical arrival distributions, and identical link‑connectivity probabilities—a “balance policy” (BP) is optimal. BP operates by, at each slot, selecting among the currently connected SS‑RS pairs the SS queue with the largest backlog; if several queues share the maximum length, BP breaks ties in a round‑robin fashion. The authors employ stochastic dominance and coupling arguments: they construct a joint probability space where the BP and any other policy experience the same arrivals and link states, then show that the BP’s queue‑length vector majorizes that of the competitor. Because f is non‑decreasing, majorization implies stochastic dominance of the cost processes, establishing BP’s optimality.

Beyond the theoretical proof, the paper proposes a low‑overhead implementation of BP. In each time slot the algorithm (1) identifies all active SS‑RS links, (2) compares the backlogs of the associated SS queues, (3) picks the maximal‑backlog queue, and (4) serves the corresponding packet. This can be realized with a priority queue or a heap, yielding O(N log N) computational complexity where N is the number of queues per layer. No predictive modeling of channel states is required; the policy uses only instantaneous connectivity information, which is readily available from physical‑layer acknowledgments or channel‑state reports.

The authors discuss the practical relevance of their result. Although exact symmetry is rarely met in real deployments, many networks exhibit approximate symmetry (e.g., homogeneous cells or similar traffic patterns), suggesting that BP will still perform near‑optimally. The cost function’s flexibility allows the framework to capture diverse quality‑of‑service metrics such as average delay, overflow probability, or energy consumption, provided they are monotone in queue length. Limitations include the reliance on symmetric statistics and the assumption of a single‑hop relay; extensions to asymmetric topologies, multi‑relay or multi‑hop scenarios, and energy‑aware objectives are identified as promising future work.

In summary, the paper delivers a rigorous stochastic‑ordering analysis of WRN scheduling, proves that a simple backlog‑balancing rule is globally optimal under symmetry, and supplies a practical algorithm with modest computational burden. This bridges a gap between abstract optimal control theory and implementable scheduling mechanisms for modern wireless relay systems.