PaperlayerView exact revision on arXiv

Source-linked AI summary

Millimeter Wave V2V Communications: Distributed Association and Beam Alignment

Cristina Perfecto, Javier Del Ser, Mehdi Bennis

arXiv:1612.04217v2cs.NIcs.GTcs.IT

TL;DR

High-mobility mmWave V2V networks require resource allocation that meets stringent latency and reliability requirements while handling dynamic links, blockage, and beam misalignment. The paper combines distributed matching-based vehicle association with swarm-intelligence beamwidth optimization, reporting improved dropping-ratio resilience and delay performance in demanding ultra-dense scenarios.

  • Problem

    High-mobility V2V communications must provide millisecond-scale latency and packet error rates around 10^-5 to 10^-9 despite dynamic topology, blockage, and beam misalignment.

  • Method

    The framework uses matching theory to pair vehicles through utilities capturing relative dynamics, channel and queue information, and mmWave effects, then optimizes transmit and receive beamwidths.

  • Results

    The proposed schemes achieve the lowest dropping ratio and strongest delay resiliency under the most demanding ultra-dense setting, while PSO generally outperforms the baselines under stringent dropping constraints.

  • Takeaways & Limitations

    Beamwidth selection must adapt to traffic, mobility, and alignment conditions because narrower beams reduce delay while wider beams improve resilience to misalignment.

Abstract

from arXiv · show

Recently millimeter-wave bands have been postulated as a means to accommodate the foreseen extreme bandwidth demands in vehicular communications, which result from the dissemination of sensory data to nearby vehicles for enhanced environmental awareness and improved safety level. However, the literature is particularly scarce in regards to principled resource allocation schemes that deal with the challenging radio conditions posed by the high mobility of vehicular scenarios. In this work we propose a novel framework that blends together Matching Theory and Swarm Intelligence to dynamically and efficiently pair vehicles and optimize both transmission and reception beamwidths. This is done by jointly considering Channel State Information (CSI) and Queue State Information (QSI) when establishing vehicle-to-vehicle (V2V) links. To validate the proposed framework, simulation results are presented and discussed where the throughput performance as well as the latency/reliability trade-offs of the proposed approach are assessed and compared to several baseline approaches recently proposed in the literature. The results obtained in our study show performance gains in terms of reliability and delay up to 25% for ultra-dense vehicular scenarios and on average 50% more paired vehicles that some of the baselines. These results shed light on the operational limits and practical feasibility of mmWave bands, as a viable radio access solution for future high-rate V2V communications.

I. INTRODUCTION

The paper addresses high-rate, low-latency and reliable mmWave V2V communications despite mobility-driven channel, topology, blockage and alignment challenges. It proposes distributed CSI- and QSI-aware vehicle association with beamwidth optimization using matching theory and swarm intelligence.

  • Motivation: Vehicular sensing applications are expected to require transmission rates exceeding 100 Mbps for raw sensor data.The motivation includes sharing sensor information among nearby vehicles for enhanced environmental awareness and safety.
  • Challenges: Vehicle mobility creates changing topology, channel conditions, antenna misalignment and blockage in mmWave V2V links.These effects motivate time-slotted communications with separate scheduling and data-transmission timescales.
  • Proposed framework: The proposed RRM jointly uses CSI to indicate transmission opportunity and QSI to reflect traffic urgency when associating V2V links.The target scope is URLLC, with millisecond-scale latency bounds and packet error rates on the order of 10^-5 to 10^-9.
  • Proposed framework: Matching theory provides distributed transmitting- and receiving-vehicle association, while swarm intelligence optimizes transmission and reception beamwidths.The paper defines utility functions for the matching game and uses nature-inspired optimization to search complex beamwidth spaces.
  • System model: The system model considers same-direction traffic on a multilane highway, with one-to-one V2V pairing over mmWave TDD links.The scheduling model separates data-transmission slots from network-wide scheduling slots, and divides the first transmission slot into alignment and data phases.

B. Channel Modelling

The channel model represents 60 GHz mmWave V2V propagation, directional antenna gains, interference and alignment delay. Beamwidth selection must balance misalignment robustness, interference exposure and the time consumed by beam alignment.

  • Channel model: The 60 GHz channel uses a log-distance pathloss model whose parameters depend on distance, atmospheric attenuation and the number of blockers.Relative vehicle movement makes the channel gain time-varying, and aggregate CSI records link conditions across receivers.
  • Antenna pattern: The ideal sectored antenna model captures boresight direction, mainlobe and sidelobe gains, front-to-back ratio and half-power beamwidth.The model approximates directional vehicle antennas while retaining the features most relevant to mmWave link quality.
  • Antenna pattern: Misalignment depends on relative vehicle speed, mainlobe widths and scheduling-interval length, while beamwidths also determine sidelobe or mainlobe interference.The resulting SINR impact motivates dynamic beamwidth selection adapted to surrounding conditions.
  • Beam alignment: A two-stage beam-alignment process based on a simplified three-step beam-codebook approach selects refined steering at both ends of each V2V link.The approach is chosen for robustness and compliance with ongoing standards, with coarse alignment or neighbor locations assumed beforehand.
  • Transmission rate: The transmission rate depends on alignment-induced delay, measured SINR, interference from simultaneous transmitters, antenna gains and bandwidth.When no alignment is performed during a slot, the alignment penalty is zero and the rate expression uses the resulting SINR directly.

E. Queues and Delay Modeling

The model captures packet arrivals, queue evolution, delivery delays, and deadline-based dropping for delay-sensitive V2V traffic. It uses these dynamics to represent both transmission delay and reliability loss.

  • Each transmitter maintains a packet queue, with i.i.d. Poisson arrivals of mean λ observed after the scheduler acts.A maximum of ri,j(t)Tt/Ps packets can be successfully transmitted during a slot.
  • Queue dynamics track backlog and buffer limits through Qi(t), Qmax, and the aggregate queue-state vector QI(t).The global system state combines channel information HJ(t) and queue information QI(t).
  • Packets are either delivered or dropped within the maximum delay Dλmax after entering the queue.Good channel conditions permit transmission with delay dp_i,j, while insufficient conditions lead to deadline-based dropping.
  • Average packet delay is computed from the delays of successfully delivered packets and aggregated over each scheduling period.The per-slot delay averages over successfully delivered packets on the link.
  • Packets that cannot be delivered within Dλmax are dropped when new arrivals occur or a scheduling slot begins, incurring reliability loss.This hard requirement prioritizes newer traffic and enforces minimum-delay communications.

F. Elements of RRM and Problem Statement

The RRM policy jointly selects vehicle associations and transmit/receive beamwidths under queue, pairing, and beamwidth constraints. The resulting optimization is computationally hard, motivating a decomposition into matching and beamwidth subproblems.

  • The association variable Φ(ts) represents all feasible vTx/vRx mappings during scheduling slot ts, constrained by a coverage radius Rc.φi,j(ts) indicates whether transmitter i is associated with receiver j.
  • The effective link rate depends jointly on the matching policy, selected beamwidths, beam alignment, and interference among V2V links.The rate expressions differ between the alignment slot and subsequent transmission slots.
  • A delay-sensitive policy must account for both successfully transmitted packets and dropped packets produced by queue dynamics.The traffic influx rate is ρ = λPs, and the queueing policy prioritizes newer incoming traffic.
  • The optimization jointly designs Φ(ts), ϕtx(ts), and ϕrx(ts), enforcing finite queues, one-to-one pairing, binary associations, and beamwidth bounds.Constraints (17b)–(17h) encode queue, association, and beamwidth requirements.
  • Because the optimization is computationally hard in vehicular environments, it is decomposed into vehicle pairing and beamwidth optimization.Matching Theory handles Φ(ts), while Swarm Intelligence selects transmit and receive beamwidths.
  • The study numerically assesses reliability under different scheduling intervals, packet arrival rates, and packet sizes.Reliability is defined using packets successfully received within a maximum delay.

III. PROPOSED SCHEME

The proposed scheme is a decentralized, self-organizing mechanism in which vehicles use utilities to establish V2V links and then optimize beamwidths. Its workflow proceeds from learning and matching to beamwidth optimization and beam alignment.

  • The scheme lets transmitters and receivers interact through utilities to decide V2V associations in a decentralized manner.Matching Theory allocates mmWave V2V links at each scheduling slot.
  • The workflow performs learning, matching, beamwidth optimization, and beam alignment as successive stages.Figure 3 describes the interrelation among processes operating at different time scales.

A. V2V Link Selection as a Matching Game

V2V link selection is formulated as a matching game between transmitters and receivers with preference relations based on utilities. The proposed algorithm exchanges CSI/QSI information, constructs preferences, applies deferred acceptance, and seeks a stable matching despite preference externalities.

  • A matching game consists of transmitter and receiver sets with preference relations that rank players in the opposite set.The game output is a matching function subject to player quotas.
  • Preferences are complete, reflexive, and transitive relations over feasible opposing vehicles, with utilities defining the corresponding rankings.Utilities are assigned for receivers and transmitters based on the prospective partner.
  • The CSI/QSI-aware algorithm begins with unmatched vehicles, exchanges link-exploration information, and computes estimated queue-state information for preference construction.Vehicle receivers collect pilot-derived SINR entries, while estimated QSI is computed from packet size and estimated rate.
  • Deferred acceptance iteratively proposes, accepts, rejects, and rematches vehicles according to their preference lists until the matching phase converges.The procedure initializes eligible partners and unmatched transmitter and receiver sets before processing proposals.
  • A matching is pairwise stable when no blocking pair prefers each other over their current partners.A blocking pair would violate the stability condition by mutually preferring an alternative pairing.
  • Gale–Shapley deferred acceptance provides polynomial-time convergence for canonical one-to-one matchings without preference interdependencies.The algorithm is unsuitable when externalities dynamically change rankings as matching evolves.

B. Utility Formulation

The utility formulation uses weighted α-fair utilities to guide distributed vehicle matching toward lower system-wide delay. Utilities incorporate queue load, relative vehicle speed, and estimated link-rate conditions while accounting for matching externalities.

  • Utility Formulation: α-fair utilities guide transmitter and receiver preferences toward matching decisions that improve delay-related system welfare.The formulation uses α = 2 for weighted minimum proportional delay fairness, and maximizing welfare minimizes the system fitness.
  • Utility Formulation: The formulation links traffic influx ρ = λPs to the number of transmission slots required to serve queued bits at the estimated link rate.This connects utility values to the system-wide fitness used for resource allocation.
  • Utility Formulation: Transmitters favor receivers with similar speeds, while receivers favor transmitters with longer queues.Similar speeds reduce susceptibility to misalignment events, whereas longer queues are prioritized to help alleviate system load.
  • Utility Formulation: Utility weights incorporate relative speed and queue status to represent mobility-related alignment risk and traffic load.Relative speed is normalized by a maximum value, and queue status is defined for each transmitter-receiver pairing under the matching.
  • Utility Formulation: The link-rate utility depends on the complete matching because directional beams create interference externalities across paired links.Receivers would otherwise need instantaneous system-wide matching information to predict interference and link rates.

C. CSI/QSI Information Learning Procedure

The learning procedure estimates future V2V rates from locally collected CSI while combining that information with QSI to construct stable matching utilities. This removes matching externalities from the estimated rates and enables a one-to-one deferred-acceptance matching game.

  • CSI/QSI Information Learning Procedure: Vehicles dynamically enforce or release links, select beamwidths, and trigger beam steering using locally measured CSI and QSI.The procedure seeks a long-term distributed resource-allocation policy while reducing control-information exchange.
  • CSI/QSI Information Learning Procedure: Receivers probe random transmitter pairings and store timestamped transmitter identifiers with measured SINR values.Pilot transmissions allow each receiver to infer channel gains and SINR over the previous scheduling period.
  • CSI/QSI Information Learning Procedure: Expected pair rates are estimated from recent observations using weights that emphasize newer CSI samples.The weighting scheme assigns nonzero weight to available samples, increases weight with recency, and normalizes weights across candidate transmitters.
  • CSI/QSI Information Learning Procedure: Estimated rates are disseminated to neighboring transmitters so each can infer the average queue status associated with a receiver.The resulting queue estimate is based on packet size divided by the estimated rate for that pair.
  • CSI/QSI Information Learning Procedure: After learning, utilities remain fixed during game formation, allowing the V2V allocation problem to be solved as a one-to-one canonical matching game with deferred acceptance.This converts the learned CSI/QSI information into the matching procedure described by Algorithm 1.

D. Beamwidth Allocation using Swarm Intelligence

After vehicle pairing, Swarm Intelligence optimizes transmitting and receiving beamwidths by searching for high-rate configurations. The PSO procedure iteratively updates candidate beamwidth vectors using particle and swarm best solutions.

  • Beamwidth Allocation using Swarm Intelligence: Swarm Intelligence selects transmitting and receiving beamwidths after vehicles are paired.The optimization is intended for convex and non-convex beamwidth-allocation problems and accounts for alignment delay and throughput.
  • Beamwidth Allocation using Swarm Intelligence: PSO represents each candidate solution as a beamwidth vector spanning the effective mmWave links.The swarm size is K, and the vector dimension equals the number of links remaining after matching.
  • Beamwidth Allocation using Swarm Intelligence: Candidate quality is measured by the average data rate across active mmWave links.This rate objective is evaluated at the current scheduling slot.
  • Beamwidth Allocation using Swarm Intelligence: Each iteration updates velocities and beamwidth candidates using prior velocity, particle-best, and global-best solutions.The process repeats for a fixed number of iterations while inertia and learning parameters control the heuristic search.

IV. SIMULATION SETUP AND RESULTS

The evaluation uses a multi-lane highway simulation with varied vehicle densities, traffic configurations, packet sizes, matching baselines, and beamwidth strategies. Results examine rate, delay, transmission success, and pairing under these settings.

  • Simulation Setup: The simulated highway is 500 meters long with six 3-meter lanes and vehicle speeds ranging from 70 to 140 km/h.Vehicles comprise 80% cars and 20% trucks, with cars sampled from five models.
  • Simulation Setup: Experiments run for 30000 ms with 2 ms transmission intervals and scheduling intervals from 20 to 500 ms.Multiple packet arrival rates and packet sizes are used to assess queue dynamics.
  • Compared Schemes: WAF applies deferred-acceptance weighted α-fair matching with fixed equal beamwidths of 5°, 45°, or 360°.These configurations isolate matching performance under fixed beamwidth choices.
  • Compared Schemes: PSO extends weighted α-fair matching with beamwidth optimization using 30 particles and 50 iterations.Unlike WAF, PSO requires a central controller such as a roadside unit to coordinate beamwidth selection.
  • Results: Figure 4 compares rate and delay CDFs in the ultra-density scenario for short and long packets under different beamwidths.The long-packet case uses Ps = 2097144 bits with traffic arrival rate λ = 1/10 · Tt packets/s, while the short-packet case uses Ps = 3200 bits and λ = 1/Tt packets/s.
  • Compared Schemes: The compared methods include MIND minimum-distance pairing and ASYN asynchronous long-term pairing.MIND renews pairing every scheduling interval, whereas ASYN retains associations until a vehicle leaves the segment.
  • Results: Figure 5 compares delay and successful transmissions across vehicle densities and traffic-arrival configurations for short and long packets.The results are interpreted jointly because traffic pressure can reduce successful transmissions while changing average delay.

A. Discussion

The discussion evaluates delay, packet dropping, beamwidth choices, and vehicle pairing across traffic and packet-size settings. Results show that method rankings depend on reliability and delay thresholds, while PSO-optimized or narrow beams often improve performance.

  • Narrow and PSO-optimized beams produce better delay and rate CDF results than the other considered beamwidths.Longer packets increase utilization and interference, degrading SINR and link rate beyond the effect of longer service time alone.
  • The proposed schemes achieve the lowest dropping ratio and strongest delay resiliency in the most demanding ULTRA setting with 3200-bit packets.ASYN has the highest success ratio in LOW density, but pairs about 25% of vTxs versus about 60% for the other schemes and above 90% at higher density.
  • The evaluation jointly counts scheduling periods satisfying delay and packet-dropping upper bounds for representative ULTRA configurations.Table III covers 3200-bit packets with λ = 1/Tt; Table IV covers 2097144-bit packets with λ = 1/30·Tt.
  • 44.67% of ASYN scheduling periods in the ULTRA scenario meet average-delay and packet-dropping constraints below 0.1 ms and 1%.
  • WAF dominates under loose 10% packet-dropping constraints, whereas PSO outperforms as dropping restrictions become more stringent.PSO benefits from wider beams that improve resilience to vehicle misalignment, despite added delay from narrower-beam choices.
  • For large packets and low arrival rates, WAF dominates below 0.1 ms average-delay bounds, while PSO generally outperforms other baselines.WAF meets a 0.05 ms minimum average-delay bound with packet dropping below 0.1% in more than 49% of scheduling intervals; PSO meets it in none.

V. CONCLUSIONS AND FUTURE RESEARCH DIRECTIONS

The paper presents a distributed mmWave V2V framework combining matching theory for vehicle association with swarm intelligence for beamwidth optimization. Future work will test it in multi-vUE and nonlinear-road settings where vehicle misalignments are more likely.

  • The framework uses matching utilities capturing vehicle dynamics, channel and queue history, directionality, blockage, and alignment delay before optimizing link beamwidths.A deferred-acceptance algorithm performs distributed pairing, followed by transmit- and receive-beamwidth allocation.
  • Future research will assess the hybrid approach in multi-vUE configurations and nonlinear road networks with more likely vehicle misalignments.The authors also plan to study interactions among scheduling periods, packet arrivals, and vehicle density in real scenarios.
Loading 1612.04217v2…