Multi-Agent Context Learning Strategy for Interference-Aware Beam Allocation in mmWave Vehicular Communications

TL;DR

This work tackles interference management in dense mmWave vehicular networks by formulating a Multi-Agent Context Learning (MACOL) framework built on contextual MAB. MACOL distributes learning across beam-level agents, uses masked contexts shared via a central node, and classifies observed contexts into interference-prone vs interference-free to guide transmissions and backoffs. An analytical beam-service model for highway scenarios pairs with MACOL to quantify vehicle sojourn, interference probability, and goodput-based rewards, achieving around 10% interference and near-perfect reliability with bandwidth efficiency (e.g., 150 MHz vs 250 MHz). The approach offers scalable, low-signaling coordination and rapid learning, enabling reliable, high-throughput V2X mmWave communication in dense deployments and guiding future extensions to more complex topologies and channel models.

Abstract

Millimeter wave (mmWave) has been recognized as one of key technologies for 5G and beyond networks due to its potential to enhance channel bandwidth and network capacity. The use of mmWave for various applications including vehicular communications has been extensively discussed. However, applying mmWave to vehicular communications faces challenges of high mobility nodes and narrow coverage along the mmWave beams. Due to high mobility in dense networks, overlapping beams can cause strong interference which leads to performance degradation. As a remedy, beam switching capability in mmWave can be utilized. Then, frequent beam switching and cell change become inevitable to manage interference, which increase computational and signalling complexity. In order to deal with the complexity in interference control, we develop a new strategy called Multi-Agent Context Learning (MACOL), which utilizes Contextual Bandit to manage interference while allocating mmWave beams to serve vehicles in the network. Our approach demonstrates that by leveraging knowledge of neighbouring beam status, the machine learning agent can identify and avoid potential interfering transmissions to other ongoing transmissions. Furthermore, we show that even under heavy traffic loads, our proposed MACOL strategy is able to maintain low interference levels at around 10%.

Figures (12)

• Figure 1: Illustration of a typical beam sector geometry layout with a BS at $P_1$ radiating a beam pointing at $\theta_k$, and a vehicle located at $P_2$ travelling in the direction $\psi_k$ within the beam. From the perspective of the beam, the vehicle is location at $r_k\angle{\phi_k}$ relative to the beam.
• Figure 2: Illustration of a highway layout with positions of BSs. The illustration shows the center BS situated at the south side of the highway radiating three north-pointing beams each with beamwidth of $60^{\circ}$
• Figure 3: Screen snapshot of Pymosim simulation running our proposed highway scenario based on geometric framework. Simulation code is given in https://github.com/cfoh/beam-analysis
• Figure 4: Illustration of numerical service distance CDF for various $p$ settings and simulated service distance CDF for MACOL and BestSNR techniques with traffic load of 30 vehicles, where $p$ is the probability that its neighbouring beam is active during a transmission service by a beam.
• Figure 5: Mean service distance for MACOL and Best SNR techniques with various traffic load conditions.

Theorems & Definitions (2)

• Definition 1: Beam Coverage for Geometric Framework
• Definition 2: Interference for Geometric Framework