To adapt to the dynamic and variable network environment of the 5th Generation Mobile Communication System-Railway, and to address issues such as decreased throughput, increased transmission delay, and difficulty in ensuring the reliability of detection data transmission between train and ground, this paper analyzes the transmission control protocol (TCP) congestion control process and compares the existing problems with three basic congestion control algorithms: TCP-Reno, TCP-Vegas, and TCP-Westwood. Based on this analysis, a congestion control algorithm that combines transmission time interval (TTI) and round-trip time (RTT) called the TTI-RTT algorithm is proposed. This algorithm calculates the test statistic for pairwise comparison and the current network congestion factor based on the RTT measurement value, using both as a criterion for judging network congestion. It dynamically adjusts the additive and multiplicative factors based on the congestion factor reflecting the degree of network congestion. Additionally, the transmission time interval is flexibly adjusted during different conditions of congestion control. Using the NS-3 simulation software, a point-to-point communication network topology is constructed considering the error code impact of high-speed railway 5G-R wireless channel, the TTI-RTT congestion control algorithm is validated through simulation and compared with the three basic algorithms in terms of performance. The results show that under different packet loss rates, the average throughput of the proposed algorithm remains above 9 Mb · s-1, which is an increase of about 5.9% over the highest and most stable throughput of the basic TCP-Reno algorithm, with significant improvements in real-time throughput stability.
随着我国高速铁路迅速发展,行车安全问题至关重要,综合检测数据对指导铁路基础设施养护维修、保障铁路列车安全运行具有突出作用[1]。目前,综合检测数据的归集仍采用人工传输、应用系统管理等传统方式,效率低,时效差,安全性不足。实现车地检测数据的快速传输、安全汇聚及自动转储成为建设智慧铁路的迫切需要[2]。将5G通信技术应用在铁路上可以满足业务的高性能需求,而现有的网络架构和移动技术对网络的性能优化并不充分,当大量业务数据上传至核心网处理时,易导致网络拥塞严重、资源竞争激烈等问题[3]。多接入边缘计算(Multi-Access Edge Computing,MEC)作为满足5G网络关键性能指标的重要技术之一,通过在靠近终端侧对业务进行处理,可以有效缩短数据传输时延,满足业务可靠性要求,部署5G MEC网络架构成为推动铁路5G专用移动通信(The Fifth Generation Mobile Communication-Railway,5G-R)系统应用的关键。然而,5G-R传输层采用基础的传输控制协议(Transmission Control Protocol,TCP)进行数据传输,难以适应MEC动态易变的特性,且对于高速铁路无线网络来说,多普勒频移会导致下层无线链路误码率增加,频繁的频率切换会导致切换失败率增加、无线链路丢失和信号中断等问题[4]。因这些问题产生的随机丢包情况会被TCP误判为网络拥塞,从而减少吞吐量、增加传输时延,难以保证检测数据车地传输的可靠性。
在无线网络中,基础的TCP拥塞控制算法有TCP-Reno[5],TCP-Vegas[6]和TCP-Westwood[7],许多学者针对基础TCP算法的拥塞控制优化问题开展了研究。文献[8]提出了一种基于深度强化学习的TCP近端策略拥塞控制(TCP-Proximal Policy Congestion Control,TCP-PPCC)算法,并与TCP-NewReno算法[9]对比,验证了该算法可缩短数据包平均传输时延、增加网络吞吐量。文献[10]基于毫米波新无线技术提出了一种动态TCP(Dynamic-TCP,D-TCP)算法,并与TCP-Reno算法、TCP-Cubic算法[11]对比,验证了该算法可增加网络吞吐量和网络资源公平性。文献[12]基于网络流的瓶颈带宽和往返时延(Round-Trip Time,RTT)提出了一种测量瓶颈带宽和RTT(Measuring Bottleneck Bandwidth and RTT,BBR)的拥塞控制算法,其核心思想为在不引起网络拥塞的前提下尽量占用网络带宽,缩短数据包平均传输时延。文献[13]提出TCP增强型拥塞控制(TCP-Enhanced Congestion Control,TCP-E)算法,以适用于高速铁路无线网络环境,该算法在增加网络吞吐量、缩短数据包平均传输时延方面均有所改善。传输时间间隔(Transmission Time Interval,TTI)指数据从发送端发送开始,至数据在无线链路上传输完成经历的总时延。5G网络支持多种长度的TTI,且在TCP协议中使用较短的TTI,能够加速数据传输过程,提升系统整体性能[14]。文献[15]提出,可以通过灵活地改变TTI长度优化无线网络中的TCP传输性能。
MALiangmin, GUJianhua. Discussion on Overall Maintenance and Management Mode for Infrastructures as High Speed Railway in China [J]. Railway Standard Design, 2011, 55 (7): 18-21. in Chinese
TAOKai, DAIChunping, SHIJing, et al. Application of 5G Technology in Railway Infrastructure Inspection and Monitoring [J]. China Railway, 2022 (12): 58-62. in Chinese
YUYaojia. Research on Key Technologies of Train-Ground Communication for CTCS-3 Train Control Service Based on 5G-R [D]. Beijing: Beijing Jiaotong University, 2021. in Chinese
[9]
STEVENSW. TCP Slow Start, Congestion Avoidance, Fast Retransmit, and Fast Recovery Algorithms [R]. America: RFC Editor, 1997: 1-6.
[10]
BRAKMOL S, O'MALLEYS W, PETERSONL L. TCP Vegas: New Techniques for Congestion Detection and Avoidance [J]. ACM SIGCOMM Computer Communication Review, 1994, 24 (4): 24-35.
[11]
MASCOLOS, CASETTIC, GERLAM, et al. TCP Westwood: Bandwidth Estimation for Enhanced Transport over Wireless Links [C]// Proceedings of the 7th Annual International Conference on Mobile Computing and Networking. New York: ACM, 2001: 287-297.
[12]
LIJ, GUANY Y, DINGP P, et al. TCP-PPCC: Online-Learning Proximal Policy for Congestion Control [C]// 2020 21st Asia-Pacific Network Operations and Management Symposium (APNOMS). New York: IEEE Press, 2020: 243-246.
KANAGARATHINAMM R, SINGHS, SANDEEPI, et al. NexGen D-TCP: Next Generation Dynamic TCP Congestion Control Algorithm [J]. IEEE Access, 2020, 8: 164482-164496.
[15]
HA S, RHEEI, XUL S. Cubic: a New TCP-Friendly High-Speed TCP Variant [J]. ACM SIGOPS Operating Systems Review, 2008, 42 (5): 64-74.
[16]
CARDWELLN, CHENGY, GUNNC S, et al. BBR: Congestion-Based Congestion Control: Measuring Bottleneck Bandwidth and Round-Trip Propagation Time [J]. Queue, 2016, 14 (5): 20-53.
[17]
GUOY, YUY J, SUNB, et al. TCP-E: an Enhanced Congestion Control Algorithm for Reliable Communication in 5G-R [C]// Proceedings of the 5th International Conference on Electrical Engineering and Information Technologies for Rail Transportation (EITRT) 2021. Singapore: Springer Singapore, 2022: 99-109.
[18]
吴诗雨.基于可扩展传输时间间隔的混合业务性能优化研究[D].北京:北京交通大学,2019.
[19]
WUShiyu. Research on Performance Optimization of Hybrid Traffic Based on Scalable Transmission Time Interval [D]. Beijing: Beijing Jiaotong University, 2019. in Chinese
[20]
PEDERSENK I, BERARDINELLIG, FREDERIKSENF, et al. A Flexible 5G Frame Structure Design for Frequency-Division Duplex Cases [J]. IEEE Communications Magazine, 2016, 54 (3): 53-59.
[21]
YANGY R, LAMS S. General AIMD Congestion Control [C]// Proceedings 2000 International Conference on Network Protocols. Osaka: IEEE, 2000: 187-198.
WANGYongwei, SIBinbin, ZHAOLiang, et al. A Refined Congestion Control Scheme for Initial TCP Connection [J]. Computer Engineering and Applications, 2004, 40 (26): 152-154. in Chinese
NIULei, GUOBo, LIUDongdong. A Slow-Start Strategy Based on Round-Trip Time [J]. Journal of Fuyang Teachers College: Natural Science, 2013, 30 (1): 50-53. in Chinese
ZHANGNing, SONGXiaohong, HOUYaxin. An optimization Strategy of Network Bandwidth Utility in Transmitting Multi-Channel Surveillance Video Information [J]. China Railway Science, 2006, 27 (3): 84-88. in Chinese
[28]
JOSEPHA J M, DINIESHV C, SHIVARANJANIM. Throughput Stability and Flow Fairness Enhancement of TCP Traffic in Multi-Hop Wireless Networks [J]. Wireless Networks, 2020, 26 (6): 4689-4704.
LIZhichun. An Improved Network Available Bandwidth Measurement Method Based on Pathload [J]. Internet of Things Technology, 2015, 5 (5): 26-28, 31. in Chinese
[31]
MUS Q, WUS Y, LIUM, et al. Hybrid Transmission Time Intervals for TCP Slow Start in Mobile Edge Computing System [C]// Proceedings of the 2018 IEEE/CIC International Conference on Communications in China (ICCC). New York: IEEE Press, 2018: 808-813.
CHENQixiang, LIMaoqing, LINJunting. Channel Fading Properties Analysis of Train to Train Direct Communication Based on Ultrashort Wave [J]. Computer Engineering and Applications, 2015, 51 (8): 74-78. in Chinese
LINJiaru, WUWeiling. Analysis of Performance of LDPC on Rician-Fading Channels [J]. Journal of Beijing University of Posts and Telecommunications, 2004, 27 (2): 48-53. in Chinese