Distributed Algorithms for Spectrum Allocation, Power Control, Routing, and Congestion Control in Wireless Networks

We develop distributed algorithms to allocate resources in multi-hop wireless networks with the aim of minimizing total cost. In order to observe the fundamental duplexing constraint that co-located transmitters and receivers cannot operate simultaneously on the same frequency band, we first devise a spectrum allocation scheme that divides the whole spectrum into multiple sub-bands and activates conflict-free links on each sub-band. We show that the minimum number of required sub-bands grows asymptotically at a logarithmic rate with the chromatic number of network connectivity graph. A simple distributed and asynchronous algorithm is developed to feasibly activate links on the available sub-bands. Given a feasible spectrum allocation, we then design node-based distributed algorithms for optimally controlling the transmission powers on active links for each sub-band, jointly with traffic routes and user input rates in response to channel states and traffic demands. We show that under specified conditions, the algorithms asymptotically converge to the optimal operating point.

💡 Research Summary

The paper tackles the joint resource‑allocation problem in multi‑hop wireless networks by proposing a fully distributed framework that simultaneously addresses spectrum allocation, power control, routing, and congestion control. The authors begin by formalizing the duplexing constraint—no node can transmit and receive on the same frequency band at the same time—and translate it into a conflict‑graph model. In this model, each link is a vertex and an edge connects two vertices if the corresponding links would interfere when using the same sub‑band. The chromatic number χ of the conflict graph therefore represents the minimum number of orthogonal sub‑bands required to avoid any conflict. By applying graph‑coloring theory, the authors prove that the number of sub‑bands K needed grows only logarithmically with χ, i.e., K = Θ(log χ). This result shows that even for dense networks the spectrum can be partitioned into a modest number of sub‑bands while guaranteeing conflict‑free operation.

To realize this theoretical bound in practice, the paper introduces an asynchronous, locally‑executed spectrum‑allocation algorithm. Each node periodically gathers the sub‑band usage of its incident links, detects any conflict, and selects an unused sub‑band from the locally available pool. If a conflict persists, the node initiates a lightweight “re‑coloring” step that preferentially chooses sub‑bands with the fewest neighboring users. The algorithm requires only neighbor‑to‑neighbor message exchanges and converges to a feasible allocation without any central coordinator, even under dynamic topology changes.

Once a feasible spectrum partition is established, the authors formulate a global cost minimization problem that jointly optimizes transmission powers p_{ij}^k, flow variables f_{ij}^k, and source rates r_i for each sub‑band k. The objective combines a convex power‑consumption term c_{ij}^k(p_{ij}^k) and a concave utility term g_i(r_i) (e.g., logarithmic utility), subject to per‑link capacity constraints derived from Shannon’s formula, per‑node power budgets, and flow‑conservation constraints. Because the problem is large‑scale and non‑separable, a centralized solution is impractical.

The authors therefore apply Lagrangian dual decomposition. Dual variables λ_{ij}^k (prices) are associated with the capacity constraints of each link‑sub‑band pair, while node‑level prices μ_i enforce flow balance. The resulting primal‑dual updates have a clear physical interpretation: each link adjusts its power according to a “price‑level” rule p_{ij}^k = (λ_{ij}^k/α)^{1/β}, each node routes traffic along the cheapest (price‑weighted) paths on each sub‑band, and each source adapts its input rate r_i by comparing marginal utility g_i′(r_i) with the aggregate price it experiences. All updates are performed locally; a node only needs the prices of its incident links and the rates advertised by its neighbors.

Convergence analysis assumes (i) strictly convex power cost functions, (ii) strictly concave, continuously differentiable utilities, and (iii) strong duality (which holds because the primal problem is convex and Slater’s condition is satisfied). Under diminishing step‑size schedules and bounded communication delays, the asynchronous primal‑dual iterations are shown to converge to the unique optimal point (p*, f*, r*). The paper also discusses robustness to packet loss and delayed price information, demonstrating that the algorithm’s stability is preserved as long as the delays are bounded.

Simulation experiments on random geometric graphs illustrate several key findings. First, the logarithmic scaling of K with χ is confirmed: even for networks with χ≈30, only K≈5–6 sub‑bands are needed. Second, the distributed power‑control component achieves near‑optimal energy consumption compared with a centralized benchmark, while respecting per‑node power caps. Third, the joint routing and congestion control yields total utility within 2–3 % of the global optimum, and the system quickly adapts to traffic spikes or link failures. Finally, the communication overhead is modest: each iteration requires only a few scalar price exchanges per neighbor, and the number of iterations to convergence grows sub‑linearly with network size.

In summary, the paper makes three major contributions: (1) a provably efficient spectrum‑allocation scheme that respects duplexing constraints with only logarithmically many sub‑bands; (2) a set of node‑centric, asynchronous algorithms that jointly optimize power, routing, and rate control in a fully distributed manner; and (3) rigorous convergence proofs and extensive simulations that validate both theoretical optimality and practical scalability. This work advances the state of the art in wireless network resource management by demonstrating that complex cross‑layer optimization can be achieved without centralized coordination, paving the way for scalable, self‑organizing ad‑hoc and sensor networks.