Source-linked AI summary
5G NR-V2X: Towards Connected and Cooperative Autonomous Driving
Hamidreza Bagheri, Md Noor-A-Rahim, Zilong Liu, Haeyoung Lee, Dirk Pesch, Klaus Moessner, Pei Xiao
TL;DR
Autonomous-driving sensors are limited by line-of-sight, motivating complementary vehicle-to-everything communication. This paper surveys how 5G NR-V2X’s physical layer, Sidelink, architecture, security, positioning, and machine learning relate to connected and cooperative driving. It concludes that NR-V2X is designed to address stringent QoS requirements, and that machine learning can improve V2X communication performance.
Problem
Vehicle sensors and cameras are limited by line-of-sight, motivating C-V2X data exchange as a complement for autonomous-driving awareness.
Source: §I · p. 1Method
The paper surveys NR-V2X physical-layer and Sidelink features, resource allocation, architecture, security, positioning, machine learning, and advanced use cases.
Source: §I · p. 2Results
The paper concludes NR-V2X is designed to address autonomous-driving QoS requirements in throughput, latency, reliability, security, and positioning, and that ML can significantly improve V2X communication performance.
Source: §VI · p. 7Takeaways & Limitations
NR-V2X complements LTE-V2X’s basic safety services with advanced safety and cooperative connected-autonomous-driving use cases.
Source: §V · p. 6
Abstract
from arXiv · showhide
This paper is concerned with the key features and fundamental technology components for 5G New Radio (NR) for genuine realization of connected and cooperative autonomous driving. We discuss the major functionalities of physical layer, Sidelink features and its resource allocation, architecture flexibility, security and privacy mechanisms, and precise positioning techniques with an evolution path from existing cellular vehicle-to-everything (V2X) technology towards NR-V2X. Moreover, we envisage and highlight the potential of machine learning for further enhancement of various NR-V2X services. Lastly, we show how 5G NR can be configured to support advanced V2X use cases in autonomous driving.
I. INTRODUCTION
5G NR-V2X is presented as a way to complement vehicle sensors and support connected and autonomous driving. The paper surveys NR-V2X features, architecture, security, positioning, machine learning, and advanced use cases.
I. INTRODUCTION
- 5G NR can complement line-of-sight-limited vehicle sensors through C-V2X data exchange and support autonomous driving.The paper surveys NR-V2X design and use cases, including physical-layer functions, Sidelink, architecture, security, privacy, positioning, and machine learning.Source: §I · pp. 1–2
II. 5G NR: 3GPP ROADMAP
The paper traces NR-V2X from LTE-V2X toward 5G NR, whose later releases target more demanding autonomous-driving requirements. Release 16 adds enhanced URLLC, higher throughput, and unicast and multicast alongside broadcast.
II. 5G NR: 3GPP ROADMAP
- LTE-V2X, introduced in 3GPP Release 14, supports network-based communication over LTE-Uu and direct device communication over the PC5 Sidelink.The PC5 interface enables short-range direct communication such as V2V, V2P, and V2I without network infrastructure.Source: §II · p. 2
- Release 16 NR-V2X expands Phase 1 features, targeting enhanced URLLC and higher throughput while maintaining Release 15 backward compatibility, and supports unicast and multicast in addition to broadcast.Source: §II · p. 2
III. KEY FEATURES OF 5G NEW RADIO
The section frames NR features around the diverse and stringent requirements of autonomous driving from network, user, and application perspectives.
III. KEY FEATURES OF 5G NEW RADIO
- NR features are framed as addressing autonomous driving’s diverse and stringent requirements from network, user, and application perspectives.Source: §III · p. 2
A. The 5G NR Physical Layer (PHY) Design
NR-V2X must accommodate high vehicle mobility and conflicting service requirements while supporting flexible radio resources, direct communications, and connected network architectures. The paper outlines corresponding PHY, Sidelink, security, and positioning mechanisms.
A. The 5G NR Physical Layer (PHY) Design
- NR supports flexible frame configurations and subcarrier spacings to address mobility-related Doppler and ICI alongside diverse service requirements.NR supports up to 400 MHz per carrier; unlike LTE’s 15 kHz spacing, it offers 15–240 kHz spacings, with larger spacings of interest for high-mobility channels.Source: A. The 5G NR Physical Layer (PHY) Design · pp. 2–3
- NR adopts LDPC codes for user data and polar codes for control channels, with QC-LDPC structure allowing parallel hardware decoding.The coding design addresses throughput, packet length, decoding latency, mobility, rate compatibility, and HARQ requirements.Source: A. The 5G NR Physical Layer (PHY) Design · p. 3
- Network slicing can tailor logical networks and dynamically allocate time-frequency PHY resources to diverse services on shared infrastructure.The passage describes slicing as a way to balance V2X QoS against other vertical applications’ resource needs.Source: A. The 5G NR Physical Layer (PHY) Design · p. 3
- NR Sidelink adds feedback-based retransmission through PSFCH, along with carrier aggregation, higher-order modulation, and modified scheduling.PSFCH supports retransmission based on feedback and channel-state-information acquisition, avoiding LTE-V2X’s blind retransmission behavior that can waste resources.Source: A. The 5G NR Physical Layer (PHY) Design · p. 3
- Sidelink Mode 1 uses base-station scheduling under coverage, whereas Mode 2 allocates resources distributively without cellular coverage.Mode 2 includes autonomous sensing-based selection, cooperative distributed scheduling, preconfigured scheduling, and scheduling for neighboring vehicles.Source: A. The 5G NR Physical Layer (PHY) Design · pp. 3–4
- NSA interworks with 4G and 5G and leverages existing infrastructure, while SA uses a new NR end-to-end architecture and 5G core.NSA options 3, 4, and 7 commonly support dual connectivity and mobility robustness; SA is designed to enhance URLLC while fulfilling eMBB and mMTC requirements.Source: A. The 5G NR Physical Layer (PHY) Design · p. 4
- NR-V2X security spans users, vehicles, communication entities, functions, and interfaces, with SA able to support more security features than NSA.SEAF separates device access authentication from data-session setup and management; 5G also protects SUPI in ciphertext over the radio link, while PC5 sharing may expose private information.Source: A. The 5G NR Physical Layer (PHY) Design · pp. 4–5
- NR-V2X combines existing and new radio-based positioning methods with RTK, but no single approach is reliable enough across all environments.The paper identifies hybrid combinations of NR techniques and embedded vehicle sensors as the most promising approach for autonomous-driving positioning accuracy.Source: A. The 5G NR Physical Layer (PHY) Design · p. 5
IV. APPLICATION OF MACHINE LEARNING FOR NR-V2X
The paper surveys how machine learning could address NR-V2X challenges arising from high mobility, changing channels, and diverse service requirements. It discusses applications from PHY adaptation and resource management to trajectory prediction, while noting robustness and computing constraints.
Motivation
- Vehicle mobility, changing topology and channels, and URLLC requirements make end-to-end V2X network design complex, while ML can adapt decisions to environmental changes.The paper also notes that distributed ML can reduce complexity and signalling overhead compared with a centralized approach.Source: §IV · p. 5
PHY adaptation
- ML can learn and predict synchronization points, channel state information, and beamforming directions from historical and contextual information in volatile channels.This addresses synchronization and channel-estimation challenges from short channel coherence times, as well as fast beam tracking and switching at mmWave.Source: §IV · pp. 5–6
UE grouping
- ML may support switching between OMA and NOMA according to V2X use-case requirements.NOMA serves multiple users on the same time/frequency resources.Source: §IV · p. 6
Radio resource management
- Dynamic vehicular conditions can make optimization-based resource management costly, motivating ML for channel and power allocation, association, and handoff.For Sidelink, decentralized multi-agent learning could let vehicles learn and cooperate using local observations, avoiding the overhead of gathering global network information.Source: §IV · p. 6
Trajectory prediction
- Trajectory prediction from observed mobility traces and movement patterns can support collision avoidance, hazard warning, handoff, link scheduling, and routing.Predicted trajectories can inform relay selection and seamless handoff between V2V and V2I.Source: §IV · p. 6
Challenges
- Safety-sensitive ML errors may have major impacts, and limited onboard computing can rule out cloud-based resources under stringent end-to-end delay constraints.The paper points to robustness and security efforts and model reduction or compression as considerations for constrained vehicles.Source: §IV · p. 6
V. 5G NR-V2X USE CASES IN COOPERATIVE AND AUTONOMOUS DRIVING
NR-V2X targets advanced cooperative and autonomous driving services alongside LTE-V2X basic safety services. The paper describes use cases ranging from coordinated driving to remote driving and relates support to latency and reliability requirements.
Target use cases
- NR-V2X targets trajectory sharing and coordinated driving, platooning, extended sensor sharing, and remote driving.These use cases involve sharing vehicle intentions, periodic platoon data, sensor or video data, or remote control.Source: §V · pp. 6–7
Latency and reliability
- Services requiring less than 5 ms latency and above 99% reliability are supported only by NR-V2X, according to the paper's three-zone comparison.Fig. 4 presents end-to-end latency and reliability requirements for advanced NR-V2X use cases.Source: §V · p. 7, Figure 4 · p. 7
- LTE-V2X supports services requiring less than 90% reliability and 10–100 ms latency; services requiring 99% reliability and 5–10 ms latency may be supported by LTE-V2X but are surely supported by NR-V2X.Services requiring less than 90% reliability and 10–100 ms latency are described as supportable in the LTE-V2X zone.Source: §V · p. 7
VI. CONCLUSIONS
The conclusion presents NR-V2X as designed for connected and cooperative autonomous driving, addressing QoS requirements across several communication and positioning dimensions. It also identifies machine learning as a potential means to improve V2X communication performance.
VI. CONCLUSIONS
- The paper presents NR-V2X design and configuration for autonomous-driving QoS requirements in throughput, latency, reliability, security, and positioning.The conclusion frames these as stringent requirements associated with autonomous driving.Source: §VI · p. 7
- The authors state that machine learning can be exploited to significantly improve V2X communication performance.They conclude that 5G NR is believed to be transformative for a highly connected and cooperative vehicular world.Source: §VI · p. 7