Source-linked AI summary
A Survey on Resource Allocation in Vehicular Networks
Md. Noor-A-Rahim, Zilong Liu, Haeyoung Lee, G. G. Md. Nawaz Ali, Dirk Pesch, Pei Xiao
TL;DR
Vehicular networks must provide QoS despite wireless impairments, high mobility, diverse service requirements, congested spectrum, and heterogeneous devices. This paper surveys resource allocation schemes across DSRC, cellular, and heterogeneous vehicular networks, covering machine-learning applications and future research directions. It categorizes the literature by network type and summarizes advantages, disadvantages, and open challenges.
Problem
Resource allocation in vehicular networks is difficult because wireless conditions, mobility, QoS requirements, spectrum availability, and device capabilities vary substantially.
Method
The paper provides a comprehensive survey organized across DSRC, cellular, and heterogeneous vehicular networks, including machine-learning-based resource allocation.
Results
The survey reviews available resource allocation schemes, summarizes their advantages and disadvantages, and identifies challenging future research directions.
Takeaways & Limitations
The paper offers a quick, comprehensive understanding of the state of the art in vehicular-network radio resource allocation strategies.
Abstract
from arXiv · showhide
Vehicular networks, an enabling technology for Intelligent Transportation System (ITS), smart cities, and autonomous driving, can deliver numerous on-board data services, e.g., road-safety, easy navigation, traffic efficiency, comfort driving, infotainment, etc. Providing satisfactory Quality of Service (QoS) in vehicular networks, however, is a challenging task due to a number of limiting factors such as erroneous and congested wireless channels (due to high mobility or uncoordinated channel-access), increasingly fragmented and congested spectrum, hardware imperfections, and anticipated growth of vehicular communication devices. Therefore, it will be critical to allocate and utilize the available wireless network resources in an ultra-efficient manner. In this paper, we present a comprehensive survey on resource allocation schemes for the two dominant vehicular network technologies, e.g. Dedicated Short Range Communications (DSRC) and cellular based vehicular networks. We discuss the challenges and opportunities for resource allocations in modern vehicular networks and outline a number of promising future research directions.
I. INTRODUCTION
Vehicular networks support ITS, smart cities, and autonomous driving, but resource allocation is challenged by mobility, diverse QoS demands, spectrum congestion, and heterogeneous hardware. This survey reviews resource allocation across DSRC, cellular, and heterogeneous vehicular networks and identifies future research directions.
- Vehicular networks enable information exchange for road safety, situational awareness, travel comfort, traffic reduction, and lower infrastructure costs.
- V2X communications primarily use Dedicated Short Range Communications (DSRC) or cellular-based vehicular communication.DSRC includes IEEE 802.11p and IEEE 1609 standards, while C-V2X operates over LTE and 5G NR.
- Dynamic resource allocation is preferred because it exploits changing wireless-channel and network conditions more effectively than static allocation.
- Resource allocation must accommodate mobility from below 60 km/h to 500 km/h or higher, conflicting QoS requirements, congested spectrum, and heterogeneous device capabilities.High mobility can require additional time-frequency resources to address Doppler and multipath impairments.
- Earlier surveys focused on cellular vehicular networks while omitting DSRC and newer machine-learning-based resource allocation methods.This paper addresses that gap by surveying DSRC, cellular, and heterogeneous vehicular networks.
- The paper surveys DSRC and cellular network architectures, resource allocation literature, machine-learning applications, and future directions including network slicing and context awareness.
B. Cellular based Vehicular Network (C-V2X)
C-V2X offers broad coverage, high capacity, QoS, and multicast support, while resource allocation must balance vehicular services against other applications sharing infrastructure. The section discusses centralized and distributed allocation modes and the complementary role of heterogeneous networking.
- C-V2X attracts attention for its large coverage, high capacity, superior QoS, and multicast/broadcast support.
- LTE-V2V uses uplink resources with SC-FDMA, organizing transmissions into resource blocks formed from 12 subcarriers and one 0.5 ms time slot.
- Sidelink Mode 3 enables eNBs to dynamically assign or reserve resources for covered vehicles’ V2V transmissions.
- Sidelink Mode 4 uses distributed resource selection, including channel sensing, congestion measurements, random reservations, and Scheduling Assignments.
- DSRC and cellular networks have complementary limitations: DSRC has limited coverage and QoS guarantees, while cellular systems may overload in very dense traffic.
III. RESOURCE ALLOCATION IN DSRC NETWORKS
DSRC resource allocation studies mainly address MAC parameters, channel access, and transmission behavior under mobility and contention. Dynamic contention-window schemes adapt access to vehicle density, relative velocity, or mobility-related fairness and can improve throughput, delay, and packet reception.
- DSRC resource-allocation approaches are classified into MAC parameter allocation, channel allocation, and rate allocation.
- MAC Parameter Allocation: Identical default MAC parameters can disadvantage fast vehicles because slow vehicles remain longer within RSU coverage and obtain more communication opportunities.
- MAC Parameter Allocation: Harigovindan et al.'s contention-window scheme maintains a relatively flat data-transfer ratio as slow-vehicle velocity changes, supporting more equal RSU access for fast and slow vehicles.
- MAC Parameter Allocation: The Harigovindan et al. scheme can disadvantage slower vehicles when few fast vehicles coexist with many slow vehicles, increasing slower-vehicle losses.
- MAC Parameter Allocation: Rossi et al. optimized the maximum contention window from surrounding vehicle density, reducing average transmission delay and improving packet reception through fewer collisions.The evaluation used a 5-km, one-lane road, 100-m transmission range, path-loss exponent 4, and estimated neighboring-vehicle counts.
- MAC Parameter Allocation: Two dynamic contention-window schemes adapt to neighbor count or relative velocity and outperform default DSRC settings with CWmin = 3,7,15 in packet delivery ratio and throughput.The schemes were evaluated in an NS-2 three-lane highway simulation with vehicle velocities from 60 km/h to 120 km/h; each performed better in a specific scenario.
B. Channel Allocation for Emergency Messages
Emergency-message channel allocation in DSRC prioritizes safety traffic by assigning it channels with greater available bandwidth while preserving RSU–OBU QoS through periodic switching. DMAE consequently reports higher emergency packet delivery and lower delay than WAVE, including under heavy traffic.
- DMAE identifies available channel bandwidth and assigns the channel with the largest bandwidth to emergency messages while periodically switching channels to maintain RSU–OBU QoS.
- DMAE achieves higher emergency packet delivery ratio than WAVE by assigning emergency messages to the available service channel with maximum bandwidth.
- DMAE outperforms WAVE in delay under heavy traffic by assigning emergency messages to reserved channels.
C. Rate Allocation
DSRC rate-allocation work addresses the limitations of fixed modulation and coding by selecting transmission rates suited to traffic and roadway conditions. Its evaluation uses packet-delivery and throughput comparisons, while coded-packet transmission must satisfy the most urgent request's deadline.
- IEEE 802.11p supports multiple MCS with data rates from 3 Mbps to 27 Mbps, but constant-MCS assumptions may underperform across diverse traffic and roadway conditions.
- The reported rate-allocation evaluation examines packet delivery ratio and network throughput using simulation results for different CWmin sizes.
- Coded-packet transmission selects an MCS that offers the highest data rate while serving all included requests within the deadline of the most urgent request.Transmission time depends on the coded-packet size and selected MCS; encoded-packet size equals the largest data-item size among included items.
IV. RESOURCE ALLOCATION IN C-V2X
C-V2X resource allocation must support diverse services and QoS requirements while sharing infrastructure with other vertical applications. Surveyed approaches use optimization, graph-based assignment, and centralized coordination to improve throughput, spectrum efficiency, connectivity, or QoS.
- Motivation: C-V2X supports eMBB, mMTC, and URLLC services with distinct data-rate, connectivity, latency, and reliability requirements.eMBB targets high data rates, mMTC supports massive sensing connectivity, and URLLC targets 1 ms round-trip time with at least 99.999% reliability.
- Motivation: C-V2X resource allocation shares spectrum, bandwidth, storage, and computing across vertical applications while balancing QoS guarantees and competing data services.The shared physical infrastructure creates a trade-off among heterogeneous requirements.
- Traditional cellular systems: Graph-based interference-aware allocation models interference through edge weights and formulate resource assignment to maximize network sum rate with low computational complexity.Network sum rate is defined as the aggregate channel capacity of all V2I and V2V links.
- Traditional cellular systems: Zhang et al. use graph theory to obtain a suboptimal resource assignment under interference constraints in a simulated 20 m × 500 m roadway with 10 resource blocks.Vehicles were randomly placed and assigned speeds between 0 and 100 km/h; vehicle and base-station interference radii were 10 m and 100 m.
- C-V2X allocation approaches: Centralized resource reuse and connectivity-oriented schemes address dense or heterogeneous conditions through eNodeB coordination and minimum-spanning-tree-based allocation.The cited centralized scheme improves spectrum efficiency and maintains required QoS at high vehicle density, while the connectivity scheme seeks improved network connectivity.
B. RA for Vehicular Computing Systems
Vehicular computing resource allocation combines vehicle, RSU, cloud, and fog resources to provide real-time services. The surveyed fog approach reduces serving time through Lagrangian optimization followed by selection of a suboptimal solution.
- System models: Vehicular computing systems integrate vehicular and RSU computational resources to provide real-time services to onboard users.The survey distinguishes centralized vehicular cloud computing from decentralized vehicular fog computing.
- Allocation method: A vehicular fog allocation problem is formulated around serving requirements and solved in two stages: Lagrangian optimization finds suboptimal solutions, then a selection process chooses among them.The approach targets reduced serving time through distributed fog infrastructure.
C. RA for Secure Vehicular Networks
The cited secure-network work treats resource allocation as part of robust V2X message dissemination, while related vehicular allocation schemes jointly manage power, subcarriers, delay, and reallocation overhead.
- Secure vehicular networks: A joint channel and security-key assignment policy classifies V2X messages into four categories to support robust and secure message dissemination.The approach also uses V2X interfaces and dynamic or semi-persistent resource-allocation modes.
- Related allocation methods: One allocation method first assigns power with fixed subcarriers using bisection, then finds suboptimal subcarrier assignments with a greedy algorithm.This is a two-stage allocation procedure.
- Vehicle platooning: Vehicle-platooning schemes allocate intra- and inter-platoon resources to share control information efficiently and timely while guaranteeing minimum rates or delay requirements.Other work minimizes signaling-related process cost during dynamic reallocation and derives reallocation-rate and delay bounds using Lyapunov optimization.
E. RA for Out-of-Coverage Scenario
The surveyed heterogeneous and out-of-coverage approaches combine distributed scheduling, trajectory information, network slicing, cooperative relaying, and heterogeneous-link allocation. Their objectives include safety-message delivery, QoS differentiation, fairness, connectivity, and throughput.
- RA for Out-of-Coverage Scenario: Out-of-coverage LTE-V2V allocation assigns resource blocks by vehicle heading and then uses channel sensing to avoid conflicts.The scheme also uses past vehicle locations to predict future trajectories and dwelling time in the uncovered area.
- Network Slicing based RA: Network slicing creates multiple logical networks on shared infrastructure so different services can receive individualized design, deployment, customization, and optimization.C-V2X slices discussed in the survey include autonomous driving, tele-operated driving, infotainment, and remote diagnosis.
- Network Slicing based RA: Slice-specific numerologies adapt subcarrier spacing and symbol duration to service needs, with mMTC favoring massive connectivity and URLLC favoring low latency and stringent reliability.The survey notes that slicing can extend into the PHY, where differing numerologies create waveform and interference-design challenges.
- Heterogeneous vehicular networks: Cooperative relaying uses a weighted bipartite graph, Hungarian matching, and binary search to select transmission strategies, relays, and the number of relayed messages.The approach guarantees fairness and can improve data rates for vehicles far from the base station.
- Heterogeneous vehicular networks: A cascaded Hungarian channel-allocation algorithm addresses high mobility, heterogeneous QoS requirements, and imperfect channel-state information through chance-constrained throughput optimization.Its simulated allocation scheme shows enhanced performance because user scheduling fully utilizes transmit power, with greater benefits at higher transmit powers.
VI. MACHINE LEARNING BASED RA FOR VEHICULAR COMMUNICATIONS
Machine learning is surveyed as an emerging approach to vehicular resource allocation, with distributed and reinforcement-learning schemes addressing dynamic channels, mobility, traffic, and QoS constraints.
- ML can support data-driven decisions for vehicular networks that generate, process, and transmit massive amounts of sensor and service data.
- Conventional optimization may require repeated recomputation as channel quality and network topology change, creating substantial overhead.
- Deep reinforcement learning allocates V2V sub-bands and transmission power using channel, traffic, latency, interference, and neighboring-channel information.
- Fuzzy Q-learning can make real-time vertical handoff decisions using RSS, vehicle velocity, data type, and target-network traffic load.
- Distributed learning is used across V2V links, vehicles, base stations, and RSUs to limit communication overhead and computational complexity.
- Reinforcement learning avoids requiring pre-collected datasets and can converge through iterative feedback from dynamic vehicular environments.
VII. FUTURE RESEARCH DIRECTIONS
The paper identifies multiple attractive directions for future research in vehicular-network resource allocation.
- Future research directions are presented as important opportunities for improving resource allocation in vehicular networks.
A. RA for NR-V2X and IEEE 802.11bd
NR-V2X and IEEE 802.11bd introduce upgraded vehicular technologies whose resource allocation must handle both traditional and mm-Wave bands, distributed scheduling, and highly dynamic environments.
- A. RA for NR-V2X and IEEE 802.11bd: NR-V2X and IEEE 802.11bd are emerging upgrades intended to reduce the gap between cellular V2X and DSRC.
- A. RA for NR-V2X and IEEE 802.11bd: Resource allocation should assign mm-Wave bands to transmitters whose receivers are within short range.
- A. RA for NR-V2X and IEEE 802.11bd: Out-of-coverage NR-V2X requires investigation of what information vehicles should share for cooperative distributed scheduling without causing congestion.
- A. RA for NR-V2X and IEEE 802.11bd: Autonomous cluster-head selection remains open, including the information, environment adaptation, and connectivity needed for surrounding vehicles.
- C-V2X network slicing must address competition with vertical applications and ultra-fast resource optimization in highly dynamic environments.
C. Security Enhancement with Blockchain Technology
The survey highlights security, machine-learning adaptation, and context-aware transfer requirements as continuing resource-allocation challenges in vehicular networks.
- C. Security Enhancement with Blockchain Technology: Mission-critical messages require ultra-resilient security, whereas multimedia services prefer lightweight security because of their high data rates.
- Machine-learning mechanisms must adapt to vehicular dynamics, changing topologies, and traffic flows that differ from conventional learning scenarios.
- Existing resource allocation work mostly assigns frequency carriers or time slots rather than context-aware, on-demand data transfers.
- On-demand applications require resource allocation to consider requested-item deadlines and priorities for reliable service.
- The survey categorizes resource-allocation schemes into DSRC, cellular, and heterogeneous vehicular networks while reviewing their advantages, disadvantages, and open directions.