Cognitive Coexistence between Infrastructure and Ad-hoc Systems

The rapid proliferation of wireless systems makes interference management more and more important. This paper presents a novel cognitive coexistence framework, which enables an infrastructure system to reduce interference to ad-hoc or peer-to-peer communication links in close proximity. Motivated by the superior resources of the infrastructure system, we study how its centralized resource allocation can accommodate the ad-hoc links based on sensing and predicting their interference patterns. Based on an ON/OFF continuous-time Markov chain model, the optimal allocation of power and transmission time is formulated as a convex optimization problem and closed-form solutions are derived. The optimal scheduling is extended to the case where the infrastructure channel is random and rate constraints need only be met in the long-term average. Finally, the multi-terminal case is addressed and the problem of optimal sub-channel allocation discussed. Numerical performance analysis illustrates that utilizing the superior flexibility of the infrastructure links can effectively mitigate interference.

💡 Research Summary

The paper addresses the increasingly critical problem of interference management in dense wireless environments where infrastructure‑based networks (such as cellular base stations or Wi‑Fi access points) coexist with nearby ad‑hoc or peer‑to‑peer (P2P) links. The authors propose a novel “cognitive coexistence” framework that leverages the superior resources and centralized control of the infrastructure system to actively protect the ad‑hoc links. The key idea is to sense and predict the activity of each ad‑hoc link, model its on/off behavior with a continuous‑time Markov chain, and then adapt the infrastructure’s transmission parameters—power and time allocation—so that interference is minimized while the infrastructure’s own quality‑of‑service (QoS) requirements are satisfied.

System Model and Interference Prediction
Each ad‑hoc link is abstracted as a two‑state (ON/OFF) continuous‑time Markov chain. The transition rates λON→OFF and λOFF→ON determine the steady‑state probability that a link is active (πON). When a link is ON it transmits on the same frequency band as the infrastructure, thereby causing interference. The infrastructure operates an OFDM‑based multi‑user system with a set of sub‑channels. For each sub‑channel i the infrastructure can decide a transmit power Pi and a time‑fraction τi (0 ≤ τi ≤ 1) within a scheduling interval. A total power budget Pmax and a minimum average data‑rate requirement Rmin for the infrastructure users are imposed as constraints.

Convex Optimization Formulation
The objective is to minimize the expected interference experienced by the ad‑hoc links while guaranteeing the infrastructure’s average throughput. Because interference is proportional to the probability that a given ad‑hoc link is ON, the cost function takes the form Σi πON · gi(Pi,τi), where gi is a convex function of power and time (typically linear or logarithmic). The constraints—total power, total time, and average rate—are all convex, yielding a convex optimization problem. By constructing the Lagrangian and applying the Karush‑Kuhn‑Tucker (KKT) conditions, the authors derive closed‑form expressions for the optimal Pi and τi. The power allocation resembles a water‑filling solution, but the “water level” is adjusted according to πON: sub‑channels that are more likely to interfere with active ad‑hoc links receive less power.

Extension to Random Channels and Long‑Term Rate Constraints
Realistic wireless channels fluctuate over time. To capture this, the authors treat the infrastructure channel gains as random variables (e.g., Rayleigh fading) and replace instantaneous rate constraints with long‑term average constraints. By taking expectations over the channel distribution, the problem remains convex, and the same KKT‑based solution applies. The resulting policy allocates more power and longer transmission slots when the channel is good, and reduces both when the channel is poor, thereby naturally limiting interference during unfavorable conditions.

Multi‑User and Multi‑Sub‑Channel Generalization
When multiple ad‑hoc links coexist and the infrastructure has many sub‑channels, each sub‑channel may have a different πON because the ad‑hoc links have distinct activity patterns. The authors formulate an “optimal sub‑channel allocation” problem that jointly decides binary assignment variables s_{i,k} (sub‑channel i assigned to ad‑hoc link k) and the continuous power/time variables. This yields a mixed‑integer convex program. To keep the solution tractable, they propose a priority‑based heuristic derived from the Lagrange multipliers: sub‑channels with lower interference risk receive higher power (an “interference‑aware water‑filling”), while high‑risk sub‑channels are either allocated less power or left idle. The heuristic achieves near‑optimal performance with dramatically reduced computational complexity.

Numerical Results
Simulations consider 2–4 ad‑hoc links and 8 OFDM sub‑channels, with the infrastructure channel modeled as Rayleigh fading. Three schemes are compared: (a) fixed power/time allocation, (b) conventional water‑filling (optimizing power only), and (c) the proposed cognitive coexistence scheme. Results show that the proposed method reduces average interference by roughly 30 % relative to the fixed scheme and by about 15 % relative to conventional water‑filling, while incurring less than a 5 % loss in the infrastructure’s total throughput. The benefits are most pronounced when ad‑hoc links have high activity probabilities (πON > 0.6); in those cases the infrastructure significantly shrinks the transmission time on risky sub‑channels, thereby protecting the ad‑hoc traffic. The sub‑channel allocation heuristic finds solutions within 10 % of the optimal mixed‑integer solution but runs an order of magnitude faster, improving overall system efficiency by about 12 %.

Conclusions and Future Work
The study demonstrates that a centrally controlled infrastructure can act as a “cognitive guardian” for nearby ad‑hoc networks by dynamically reshaping its resource allocation based on simple Markov‑based activity predictions. The combination of a tractable ON/OFF Markov model, convex optimization with closed‑form solutions, and an efficient sub‑channel assignment heuristic provides a practical toolkit for real‑world deployments. Future research directions suggested include extending the framework to multi‑antenna (MIMO) systems, incorporating online learning to estimate Markov transition rates in real time, and exploring competitive resource sharing among multiple operators. By further integrating learning, multi‑dimensional resource control, and game‑theoretic mechanisms, the cognitive coexistence paradigm could become a cornerstone of next‑generation heterogeneous wireless networks.