Spectrum Sharing between Wireless Networks

We consider the problem of two wireless networks operating on the same (presumably unlicensed) frequency band. Pairs within a given network cooperate to schedule transmissions, but between networks there is competition for spectrum. To make the problem tractable, we assume transmissions are scheduled according to a random access protocol where each network chooses an access probability for its users. A game between the two networks is defined. We characterize the Nash Equilibrium behavior of the system. Three regimes are identified; one in which both networks simultaneously schedule all transmissions; one in which the denser network schedules all transmissions and the sparser only schedules a fraction; and one in which both networks schedule only a fraction of their transmissions. The regime of operation depends on the pathloss exponent $\alpha$, the latter regime being desirable, but attainable only for $\alpha>4$. This suggests that in certain environments, rival wireless networks may end up naturally cooperating. To substantiate our analytical results, we simulate a system where networks iteratively optimize their access probabilities in a greedy manner. We also discuss a distributed scheduling protocol that employs carrier sensing, and demonstrate via simulations, that again a near cooperative equilibrium exists for sufficiently large $\alpha$.

💡 Research Summary

The paper tackles the classic problem of two independent wireless networks sharing the same unlicensed frequency band. Within each network, users cooperate to schedule transmissions, but between the networks there is pure competition. To keep the analysis tractable, the authors assume a slotted random‑access model: every user of network i attempts to transmit in a slot with probability pᵢ, which is the control variable each network can adjust. Interference is modeled using a standard distance‑based path‑loss law, received power ∝ d⁻ᵅ, where the exponent α captures the propagation environment (α≈2 for free‑space, larger values for indoor or obstructed settings).

The interaction is cast as a non‑cooperative two‑player game. Each network’s payoff is its average successful transmission rate, i.e., the probability that a scheduled packet meets a SINR threshold given the random access of both networks. By taking expectations over the spatial distribution of transmitters, the authors derive closed‑form expressions for the expected interference and thus for the payoff as a function of the two access probabilities (p₁, p₂). The Nash equilibrium (NE) is obtained by solving the best‑response equations ∂Payoffᵢ/∂pᵢ = 0 for i = 1,2.

Three distinct equilibrium regimes emerge, driven solely by the value of α:

Full‑Access Regime (small α). When the path‑loss exponent is low, interference decays slowly with distance, so the marginal cost of increasing pᵢ is modest. Both networks’ best responses are p₁ = p₂ = 1; every user transmits in every slot. This maximizes each network’s own attempt rate but leads to severe mutual interference and low overall spectral efficiency.
Asymmetric‑Access Regime (moderate α, density imbalance). If one network is denser (more users per unit area) than the other, the denser network’s best response remains p = 1, while the sparser network reduces its access probability to a value p < 1 that balances the extra interference it suffers. The equilibrium is thus asymmetric: the “dominant” network monopolizes the spectrum, the weaker one backs off.
Partial‑Access Regime (α > 4). For sufficiently large path‑loss exponents, interference drops sharply with distance. In this case the best‑response functions intersect at interior points 0 < p₁, p₂ < 1. Both networks voluntarily limit their transmission attempts, achieving a cooperative‑like outcome without any explicit coordination. The total throughput is maximized, and the equilibrium is Pareto‑superior to the previous two regimes.

The authors validate the analytical predictions with an iterative greedy simulation. Starting from arbitrary (p₁, p₂), each network in turn optimizes its own p while holding the opponent’s value fixed, mimicking best‑response dynamics. The process converges to the analytically derived NE in all three regimes, confirming the robustness of the equilibrium characterization.

Beyond the theoretical model, the paper proposes a practical distributed scheduling protocol based on carrier sensing. Each node measures the instantaneous channel occupancy (energy detection) and adapts its transmission probability according to a simple rule that approximates the best‑response function derived analytically. Simulations of this protocol show that, for α > 4, the network self‑organizes into the partial‑access equilibrium, demonstrating that the cooperative outcome can be achieved with only local sensing and no centralized controller.

Key insights and contributions are:

A clean game‑theoretic formulation of inter‑network spectrum sharing under random access, with explicit dependence on the physical propagation parameter α.
Identification of a critical threshold (α ≈ 4) beyond which selfish behavior naturally leads to a socially optimal, partially coordinated equilibrium.
Demonstration that simple carrier‑sensing based adaptation can realize the equilibrium in realistic, decentralized settings.
Evidence that in high‑loss environments (e.g., indoor millimeter‑wave deployments) rival operators may “cooperate” implicitly, reducing the need for regulatory spectrum‑partitioning or complex coordination mechanisms.

From a system‑design perspective, the work suggests that measuring or estimating the effective path‑loss exponent of the deployment environment should be a prerequisite for configuring access‑probability control algorithms. In scenarios where α is low, external coordination (e.g., spectrum etiquette, time‑division agreements) may be required to avoid the inefficient full‑access equilibrium. Conversely, in high‑α environments, operators can rely on autonomous, locally adaptive protocols to achieve near‑optimal spectrum utilization without sacrificing fairness.