Source-linked AI summary
Ultra-Reliable Low Latency Cellular Networks: Use Cases, Challenges and Approaches
He Chen, Rana Abbas, Peng Cheng, Mahyar Shirvanimoghaddam, Wibowo Hardjawana, Wei Bao, Yonghui Li, Branka Vucetic
TL;DR
Mission-critical URLLC applications require extremely low end-to-end latency and very high reliability, while current cellular networks contain substantial delay sources. The paper characterizes these requirements and delays, then proposes techniques spanning coding, signal processing, resource management, and broader network design.
Problem
Mission-critical URLLC applications require submillisecond to 10 ms end-to-end latency and extremely high reliability, while current networks incur delay from retransmissions, channel errors, congestion, and uplink transmission procedures.
Method
The paper analyzes LTE latency sources and investigates latency-reduction approaches in error-control coding, signal processing, radio resource management, and network design.
Results
The paper proposes and evaluates several techniques for reducing end-to-end latency, including joint coding and modulation and self-adaptive codes for shorter block lengths.
Takeaways & Limitations
URLLC design must address latency and reliability jointly across multiple protocol and network components rather than relying on a single technique.
Takeaways & Limitations
Short error-control codes remain constrained by decoding latency and capacity loss at short block lengths, and no known codes achieve the PPV limit.
Abstract
from arXiv · showhide
The fifth-generation cellular mobile networks are expected to support mission critical ultra-reliable low latency communication (URLLC) services in addition to the enhanced mobile broadband applications. This article first introduces three emerging mission critical applications of URLLC and identifies their requirements on end-to-end latency and reliability. We then investigate the various sources of end-to-end delay of current wireless networks by taking the 4G Long Term Evolution (LTE) as an example. Subsequently, we propose and evaluate several techniques to reduce the end-to-end latency from the perspectives of error control coding, signal processing, and radio resource management. We also briefly discuss other network design approaches with the potential for further latency reduction.
II. Emerging URLLC Applications
The paper presents tele-surgery, intelligent transportation, and industrial automation as mission-critical URLLC applications with stringent latency and reliability requirements. These requirements motivate wireless designs that can support timely, reliable communication while preserving deployment flexibility.
- Tele-surgery: Tele-surgery includes remote consultations and remote surgery, requiring reliable audio, video, and, for remote surgery, haptic feedback.Haptic feedback has the tightest delay requirement among the three traffic types.
- Tele-surgery: End-to-end round-trip times below 1ms are required for haptic feedback in remote surgery.
- Tele-surgery: Remote surgery requires BLER down to 10^-9 because noticeable errors can lead to catastrophic outcomes.
- Intelligent Transportation: Automated overtaking, cooperative collision avoidance, and high-density platooning require 5–10ms end-to-end latency and BLER down to 10^-5.
- Industrial Automation: Industrial automation requires end-to-end latency below 0.5ms and BLER of 10^-9, while existing wireless technologies cannot meet these requirements.
- Industrial Automation: Replacing wired links with radio links can reduce manufacturing, installation, and maintenance costs while improving flexibility and long-term reliability in motion applications.
III. Latency Sources in Cellular Networks
The paper decomposes LTE end-to-end latency into radio-access, core-network, processing, and propagation components. It identifies link establishment and retransmissions as critical radio-access delays, then focuses on reducing TTI, signal-processing, and radio-resource-management delays.
- Delay Sources: Link establishment and packet retransmissions caused by channel errors and congestion are the two most critical radio-access delay sources.
- Uplink Latency: An aligned LTE uplink transmission can incur up to 17ms of average radio-access delay even without retransmissions.The delay includes scheduling-request, grant-acquisition, decoding, and data-transmission steps.
- Downlink Latency: Downlink transmission sums to 7.5ms because it includes incoming-data processing, TTI alignment, transmission, and UE decoding without grant acquisition.
- End-to-End Latency: End-to-end latency also includes core-network, cloud, Internet-server, and propagation delays, increasing with transmitter–receiver distance and network load.At least 39ms is needed to contact the core-network gateway, while a Google-server response requires at least 44ms.
- Reduction Approaches: The paper groups delay sources into TTI, signal processing, radio resource management, retransmissions, and core-network categories.It mainly addresses the first three categories and briefly discusses the last two.
IV. Short Error Control Codes
Short block lengths are necessary for low delay but incur finite-blocklength capacity loss, motivating self-adaptive codes such as AFC for URLLC. AFC avoids transmitter-side CSI overhead and offers near-capacity performance with linear complexity, though feedback loss can add delay.
- Finite-blocklength limits: Short codes enable shorter transmission intervals, but finite-blocklength analysis shows severe capacity loss and no known codes achieve the PPV limit.The PPV analysis characterizes delay, throughput, and reliability tradeoffs for Gaussian channels and fixed-rate codes.
- Finite-blocklength limits: Adaptive wireless schemes increase latency through CSI estimation and feedback, code-rate and modulation selection, and encoding or decoding complexity.These overheads accompany selecting among fixed-rate codes for the estimated channel state.
- Self-adaptive codes: Self-adaptive, or rateless, codes send the exact amount of coded symbols needed for successful decoding and require no transmitter-side CSI.This removes channel-estimation overhead and delay at the transmitter.
- Self-adaptive codes: AFC jointly optimizes coding and modulation through specialized EXIT charts, providing seamless performance across SNRs with linear encoding and decoding complexity.AFC is represented by a single sparse non-binary generator matrix.
- AFC performance: Lost positive acknowledgements make AFC retransmit the same codeword, introducing extra delay and motivating more robust acknowledgment coding.The feedback is sent after successful decoding and may be carried in uplink or downlink slots.
V. Ultra-fast Signal Processing
LTE receiver latency is dominated by channel estimation, MIMO detection, and Turbo decoding, motivating ultra-fast signal-processing techniques. The section evaluates faster channel estimation and discusses parallel interference cancellation and parallel Turbo decoding as latency-reduction approaches.
- Latency bottlenecks: LTE receiver computational time is measured across major modules using an Intel Core i5 implementation and three typical bandwidths.The simulations use 4 transmit antennas, 2 receive antennas, 16-QAM, a 0.3691 code rate, and 10 dB SNR.
- Latency bottlenecks: MMSE channel estimation, MMSE-SIC MIMO detection, and Turbo decoding consume the most computational resources and dominate receiver computational time.These functions are identified as the primary processing bottlenecks in the LTE receiver.
- Channel estimation: The proposed channel-estimation method combines least-square estimation on reference symbols with low-complexity 2-D biharmonic interpolation across the resource block.The interpolation produces smoother curves than linear and nearest-neighbor methods.
- Channel estimation: Around 60% less computational time is achieved than with MMSE-based estimation at B = 5MHz while achieving almost the same system throughput.This result is reported for the proposed channel-estimation method.
- MIMO detection: PIC-DSC reduces detection latency relative to MMSE-SIC by using matched-filter-based parallel interference cancellation instead of time-consuming MMSE matrix inversion.The approach uses decoder and matched-filter iterations to approach MMSE-receiver performance.
- MIMO detection: 128 times lower computational complexity than MMSE and ZF is reported for PIC-DSC, with negligible performance degradation.Parallel Turbo decoding is another approach, offering 1.53 Gb/s average throughput and 50% hardware-resource reduction versus the original architecture.
VI. Radio Resource Management
The section examines NOMA and radio-resource-block slicing as latency-oriented resource-management approaches, with benefits that depend on traffic load and service demands.
- A. Non-orthogonal Multiple Access: Contention-based access suffers severe collisions and high latencies as the number of users increases.
- A. Non-orthogonal Multiple Access: Code-domain NOMA assigns each user a unique channel code, enabling grant-free transmission and potentially reducing medium-access latency while improving reliability.
- A. Non-orthogonal Multiple Access: When device numbers are small, OMA slightly outperforms NOMA in delay because collisions are unlikely and orthogonal transmission provides higher spectral efficiency.
- A. Non-orthogonal Multiple Access: When device numbers are large, NOMA outperforms OMA by exploiting interference, avoiding random access, and sharing radio resources.
- B. Resource Reservation via Resource Block Slicing: Resource-block slicing reserves resources for each service while allowing unused blocks to be shared, reducing cross-service overload effects and preserving spectral efficiency.
- B. Resource Reservation via Resource Block Slicing: Reducing median latency from 10ms to 5ms for ITS and 6ms for SG, resource-block slicing improved both services over legacy LTE in simulation.
- B. Resource Reservation via Resource Block Slicing: Allocating 80% of reserved resources to ITS improved ITS latency, while SG latency deteriorated with 20% allocation but remained better than legacy LTE.
VII. Other Potential Techniques
The section surveys additional latency-reduction approaches, including D2D communication, mobile edge computing, caching, lightweight security, and cross-layer error control, while identifying implementation constraints.
- Cross-layer Error Control: UDP is attractive for mission-critical wireless communication because it avoids TCP’s overhead, but its reliability must be significantly improved for URLLC.
- Cross-layer Error Control: Short AFC codes can improve UDP reliability through concatenated physical- and network-layer coding with soft-output and soft-input decoding.
- Device-to-Device Communication: Device-to-device communication can reduce latency through direct links, but shared-spectrum operation requires efficient interference management to preserve that gain.
- Mobile Edge Computing: Mobile edge computing reduces end-to-end latency by offloading computationally intensive jobs to modules located closer to sensing devices.
- Mobile Edge Computing: MEC deployment remains immature because computing services are incompatible with the existing LTE protocol stack and may require substantial reconstruction and reconfiguration.
- Mobile Caching for Content Delivery: Mobile caching reduces delay and improves backhaul efficiency by serving cached content directly at base stations when requests hit the cache.
- Mobile Caching for Content Delivery: Caching benefits are constrained by limited base-station cache size and unlimited possible contents, making popularity-based cache selection essential for maximizing hit rate.
- Lightweight security mechanisms: Cryptographic overhead can increase message size and transmission latency, motivating lightweight security mechanisms such as physical-layer approaches for URLLC.
VIII. Conclusions
The paper identifies URLLC applications and requirements, diagnoses LTE latency bottlenecks, and evaluates approaches spanning coding, signal processing, and radio resource management. Initial results indicate that substantial end-to-end latency reduction requires reducing each latency component, while additional cross-layer and edge-oriented approaches offer further potential.
- The paper covers URLLC use cases including tele-surgery, smart transportation, and industry automation, with corresponding latency and reliability requirements.
- The LTE delay breakdown identifies major latency bottlenecks and indicates that several orders of end-to-end latency reduction are required for mission-critical applications.Each latency component therefore needs significant reduction.
- Initial results support short analog fountain codes, ultra-fast signal processing, non-orthogonal multiple access, and resource reservation via resource block slicing as latency-reduction approaches.
- The paper emphasizes physical- and multiple-access-layer measures for latency reduction.
- Cross-layer error control, device-to-device communication, mobile edge computing, and mobile caching are discussed as potential additional latency-reduction measures.