NEWS

CUBIC-LEARN: A REINFORCEMENT LEARNING APPROACH TO CUBIC CONGESTION CONTROL


(Received: 30-May.-2025, Revised: 3-Sep.-2025 , Accepted: 22-Sep.-2025)
Managing congestion effectively enables reliable and fast data transfer over networks. CUBIC delivers reliable results under normal circumstances, but cannot adapt effectively to changing network scenarios. We introduce CUBIC-Learn, an RL approach for improving congestion control in CUBIC. The central idea is to use a Q- learning algorithm to adjust congestion window thresholds based on current data on packet loss, throughput and latency. Simulations demonstrate more efficient and reliable congestion control when using CUBIC-Learn compared to standard CUBIC. CUBIC-Learn achieves a 47% reduction in packet loss, over a 59% increase in bandwidth utilization, approximately a 28% decrease in retransmissions and 47% lower latency. In addition, CUBIC-Learn shows significant improvements in congestion window (cwnd) growth behavior, fairness among competing flows and stability under heterogeneous traffic and network scenarios, including gigabit-scale bandwidth conditions. Statistical analysis further confirms the robustness of these gains, while the method introduces no additional computational overhead. Overall, CUBIC-Learn performs better than PCC, Reno, Tahoe, NewReno and BBRv3 in most metrics. These findings suggest that RL can markedly improve congestion control in high-speed networks.

[1] V. Jacobson, "Congestion Avoidance and Control," ACM SIGCOMM Computer Communication Review, vol. 25, no. 1, pp. 157–187, DOI: 10.1145/205447.205462, Jan. 1995.

[2] S. Ha et al., "CUBIC: A New TCP-friendly High-speed TCP Variant," SIGOPS Oper. Syst. Rev., vol. 42, no. 5, pp. 64–74, DOI:  10.1145/1400097.1400105, Jul. 2008.

[3] Z. D. Ghobadi et al., "An Overview of Reinforcement Learning and Deep Reinforcement Learning for Condition-based Maintenance," Int. J. of Reliability, Risk and Safety: Theory and Application, vol. 4, no. 2, pp. 81–89, DOI: 10.30699/IJRRS.4.2.9, Dec. 2021.

[4] R. Al-Saadi et al., "A Survey of Delay-based and Hybrid TCP Congestion Control Algorithms," IEEE Commun. Surveys Tuts., vol. 21, no. 4, pp. 3609–3638, DOI: 10.1109/COMST.2019.2904994, 2019.

[5] B. Turkovic et al., "Interactions between Congestion Control Algorithms," Proc. of the 2019 Network Traffic Measurem. and Analysis Conf. (TMA), pp. 161–168, DOI: 10.23919/TMA.2019.8784674, 2019.

[6] A. Kuzmanovic and E. W. Knightly, "TCP-LP: Low-priority Service via End-point Congestion Control," IEEE/ACM Trans. Netw., vol. 14, no. 4, pp. 739–752, Aug. 2006.

[7] R. Mittal et al., "TIMELY: RTT-based Congestion Control for the Datacenter," SIGCOMM Comput. Commun. Rev., vol. 45, no. 4, pp. 537–550, Aug.  2015.

[8] Z. Wang and J. Crowcroft, "Eliminating Periodic Packet Losses in the 4.3-Tahoe BSD TCP Congestion Control Algorithm," SIGCOMM Comput. Commun. Rev., vol. 22, no. 2, pp. 9–16, Apr. 1992.

[9] S. Shalunov et al., "Low Extra Delay Background Transport (LEDBAT)", RFC 6817, IETF, [Online], Available: https://datatracker.ietf.org/doc/rfc6817/, Dec. 2012.

[10] M. Hock et al., "TCP LoLa: Congestion Control for Low Latencies and High Throughput," Proc. of the IEEE 42nd Conf. Local Computer Networks (LCN), pp. 215–218, DOI: 10.1109/LCN.2017.42, 2017.

[11] L. S. Brakmo and L. L. Peterson, "TCP Vegas: End to End Congestion Avoidance on a Global Internet," IEEE Journal on Selected Areas in Communications, vol. 13, no. 8, pp. 1465–1480, Oct. 1995.

[12] K. N. Srijith et al., "TCP Vegas-A: Improving the Performance of TCP Vegas," Computer Communications, vol. 28, no. 4, pp. 429–446, Mar. 2005.

[13] D. X. Wei et al., "FAST TCP: Motivation, Architecture, Algorithms, Performance," IEEE/ACM Transactions on Networking, vol. 14, no. 6, pp. 1246–1259, Dec. 2006.

[14] S. Belhaj and M. Tagina, "VFAST TCP: An Improvement of FAST TCP," Proc. of the 10th IEEE Int. Conf. on Computer Modeling and Simul.  (Uksim’08), pp. 88–93, DOI: 10.1109/UKSIM.2008.50, 2008.

[15] A. Venkataramani et al., "TCP Nice: A Mechanism for Background Transfers," SIGOPS Oper. Syst. Rev., vol. 36, no. SI, pp. 329–343, DOI: 10.1145/844128.844159, Dec. 2003.

[16] S. Bhandarkar et al., "Emulating AQM from End Hosts," SIGCOMM Comput. Commun. Rev., vol. 37, no. 4, pp. 349–360, DOI: 10.1145/1282427.1282420, Aug. 2007.

[17] G. Marfia et al., "TCP Libra: Exploring RTT-Fairness for TCP," Proc. of the 6th Int. IFIP-TC6 Conf. on Ad Hoc and Sensor Networks, Wireless Networks, Next Generation Internet (NETWORKING’07), pp. 1005–1013, DOI: 10.1007/978-3-540-72606-7_86, 2007.

[18] D. A. Hayes and G. Armitage, "Revisiting TCP Congestion Control Using Delay Gradients," Proc. of the Int. Conf. on Research in Networking (NETWORKING 2011), pp. 328–341, DOI: 10.1007/978-3-642-20798-3_25, 2011.

[19] M. Dong et al., "PCC: Re-architecting Congestion Control for Consistent High Performance," arXiv: 1409.7092, DOI: 10.48550/arXiv.1409.7092, 11 Oct. 2014.

[20] N. Cardwell et al., "BBR: Congestion-based Congestion Control," Commun. ACM, vol. 60, no. 2, pp. 58–66, DOI: 10.1145/3009824, Jan. 2017.

[21] I. Petrov and T. Janevski, "Evolution of TCP in High Speed Networks," Int. Journal of Future Generation Communication and Networking, vol. 8, no. 2, pp. 137–186, Apr. 2015.

[22] R. King et al., "TCP-Africa: An Adaptive and Fair Rapid Increase Rule for Scalable TCP," Proc. of the IEEE 24th Annual Joint Conf. of the IEEE Computer and Communications Societies, vol. 3, pp. 1838–1848, DOI: 10.1109/INFCOM.2005.1498463, 2005.

[23] H. Shimonishi and T. Murase, "Improving Efficiency-friendliness Trade-offs of TCP Congestion Control Algorithm," Proc. of the IEEE Global Telecommunications Conf. (GLOBECOM ’05), vol. 1, p. 5, DOI: 10.1109/GLOCOM.2005.1577631, 2005.

[24] P. Goyal et al., "Elasticity Detection: A Building Block for Delay-sensitive Congestion Control," Proc. of the 2018 ACM Applied Network. Research Workshop, p. 75, DOI:10.1145/3232755.3232772, 2018.

[25] C. P. Fu and S. C. Liew, "TCP Veno: TCP Enhancement for Transmission over Wireless Access Networks," IEEE Journal on Selected Areas in Communications, vol. 21, no. 2, pp. 216–228, Feb. 2003.

[26] V. Arun and H. Balakrishnan, "Copa: Practical Delay-based Congestion Control for the Internet," Proc. of the 2018 ACM Applied Network. Research Workshop, p. 19, DOI: 10.1145/3232755.3232783, 2018.

[27] S. Liu, et al., "TCP-Illinois: A Loss- and Delay-based Congestion Control Algorithm for High-speed Networks," Performance Evaluation, vol. 65, no. 6, pp. 417–440, June 2008.

[28] S. Mascolo et al., "TCP Westwood: Bandwidth Estimation for Enhanced Transport over Wireless Links," Proc. of the 7th Annual Int. Conf. on Mobile Computing and Networking, pp. 287–297, DOI: 10.1145/381677.381704, 2001.

[29] L. A. Grieco and S. Mascolo, "TCP Westwood and Easy RED to Improve Fairness in High-speed Networks," Proc. of the Int. Workshop on Protocols for High Speed Networks, pp. 130–146, DOI: 10.1007/3-540-47828-0_9, 2002.

[30] L. Xu et al., "Binary Increase Congestion Control (BIC) for Fast Long-distance Networks," IEEE INFOCOM 2004, vol. 4, pp. 2514–2524, DOI: 10.1109/INFCOM.2004.1354672, 2004.

[31] C. Caini and R. Firrincieli, "TCP Hybla: A TCP Enhancement for Heterogeneous Networks," Int. Journal of Satellite Communications and Networking, vol. 22, no. 5, pp. 547-566, DOI: 10.1002/sat.799, 2004.

[32] K. Fall and S. Floyd, "Simulation-based Comparisons of Tahoe, Reno and SACK TCP," SIGCOMM Comput. Commun. Rev., vol. 26, no. 3, pp. 5-21, DOI: 10.1145/235160.235162, July 1996.

[33] A. Gurtov et al., "The NewReno Modification to TCP’s Fast Recovery Algorithm," RFC 6582, IETF, DOI: 10.17487/RFC3782, Apr. 2012.

[34] R. Wang et al., "TCP with Sender-Side Intelligence to Handle Dynamic, Large, Leaky Pipes," IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 235-248, DOI: 10.1109/JSAC.2004.839426, Feb. 2005.

[35] S. Floyd, "HighSpeed TCP for Large Congestion Windows," RFC 3649, IETF, DOI: 10.17487/RFC3649/, Dec. 2003.

[36] J. Gomez et al., "Evaluating TCP BBRv3 Performance in Wired Broadband Networks," Computer Communications, vol. 222, pp. 198–208, DOI: 10.1016/j.comcom.2024.04.037, Jun. 2024.

[37] D. Zeynali, et al., "Promises and Potential of BBRv3," Proc. of Passive and Active Measurement: 25th Int. Conf. (PAM 2024), vol. 14538, pp. 249–272, DOI: 10.1007/978-3-031-56252-5_12, 2024.

[38] J. Wang et al., "CUBIC-FIT: A High Performance and TCP CUBIC Friendly Congestion Control Algorithm," IEEE Commun. Lett., vol. 17, no. 8, pp. 1664-1667, Aug. 2013.

[39] S. Patel et al., "A Comparative Performance Analysis of TCP Congestion Control Algorithms: Newreno, Westwood, Veno, BIC and Cubic," Proc. of the 2020 6th Int. Conf. Signal Process. Commun. (ICSC), pp. 23-28, DOI: 10.1109/ICSC48311.2020.9182733, 2020.

[40] J. Y. Lee et al., "Coupled CUBIC Congestion Control for MPTCP in Broadband Networks," Computer Systems Science and Engineering, vol. 45, no. 1, pp. 99-115, 2022.

[41] C. McKenzie and M. D. McDonnell, "Modern Value Based Reinforcement Learning: A Chronological Review," IEEE Access, vol. 10, pp. 134704-134725, 2022.

[42] M. Sewak, "Policy-based Reinforcement Learning Approaches," Chapter in Book: Deep Reinforcement Learning: Frontiers of Artificial Intelligence, pp. 127-140, DOI: 10.1007/978-981-13-8285-7_10, 2019.

[43] B. Jang et al., "Q-Learning Algorithms: A Comprehensive Classification and Applications," IEEE Access, vol. 7, pp. 133653–133667, DOI: 10.1109/ACCESS.2019.2941229, 2019.

[44] M. Sewak, "Deep Q Network (DQN), Double DQN and Dueling DQN," Proc. of Deep Reinforcement Learning: Frontiers of Artificial Intelligence, pp. 95–108, DOI: 10.1007/978-981-13-8285-7_8, 2019.

[45] M. Lehmann, "The Definitive Guide to Policy Gradients in Deep Reinforcement Learning: Theory, Algorithms and Implementations," arXiv: 2401.13662, DOI: 10.48550/arXiv.2401.13662, Mar. 2024.

[46] I. Grondman et al., "A Survey of Actor-critic Reinforcement Learning: Standard and Natural Policy Gradients," IEEE Transactions on Systems, Man and Cybernetics, Part C (Applications and Reviews), vol. 42, no. 6, pp. 1291–1307, DOI: 10.1109/TSMCC.2012.2218595, Nov. 2012.

[47] J. Schulman et al., "Proximal Policy Optimization Algorithms," arXiv: 1707.06347, DOI: 10.48550/arXiv.1707.06347, Aug. 2017.

[48] E. H. Sumiea et al., "Deep Deterministic Policy Gradient Algorithm: A Systematic Review," Heliyon, vol. 10, no. 9, p. e30697, DOI: 10.1016/j.heliyon.2024.e30697, May 2024.

[49] H. Shen et al., "Towards Understanding Asynchronous Advantage Actor-critic: Convergence and Linear Speedup," IEEE Transactions on Signal Processing, vol. 71, pp. 2579–2594, 2023.

[50] Z. Xu et al., "Experience-driven Congestion Control: When Multi-path TCP Meets Deep Reinforcement Learning," IEEE Journal on Selected Areas in Communications, vol. 37, no. 6, pp. 1325–1336, June 2019.

[51] X. Nie et al., "Dynamic TCP Initial Windows and Congestion Control Schemes through Reinforcement Learning," IEEE Journal on Selected Areas in Communications, vol. 37, no. 6, pp. 1231–1247, June 2019.

[52] Y. Wang et al., "An Intelligent TCP Congestion Control Method Based on Deep Q Network," Future Internet, vol. 13, no. 10, p. 261, DOI: 10.3390/fi13100261, Oct. 2021.

[53] K. Xiao, et al., "TCP-Drinc: Smart Congestion Control Based on Deep Reinforcement Learning," IEEE Access, vol. 7, pp. 11892-118904, DOI: 10.1109/ACCESS.2019.2892046, 2019.

[54] S. Ketabi et al., "A Deep Reinforcement Learning Framework for Optimizing Congestion Control in Data Centers," Proc. of the 2023 IEEE/IFIP Network Operations and Management Symposium (NOMS 2023), pp. 1-7, DOI: 10.1109/NOMS56928.2023.10154411, 2023.

[55] L. Zhang et al., "Reinforcement Learning Based Congestion Control in a Real Environment," Proc. of the 2020 29th Int. Conf. on Computer Communications and Networks (ICCCN), pp. 1-9, DOI: 10.1109/ICCCN49398.2020.9209750, 2020.

[56] B. Fuhrer et al., "Implementing Reinforcement Learning Datacenter Congestion Control in NVIDIA NICs," Proc. of the 2023 IEEE/ACM 23rd Int. Symposium on Cluster, Cloud and Internet Computing (CCGrid), pp. 331-343, DOI: 10.1109/CCGrid57682.2023.00039, 2023.

[57] Z. Xia et al., "A Multi-objective Reinforcement Learning Perspective on Internet Congestion Control," Proc. of the 2021 IEEE/ACM 29th Int. Symposium on Quality of Service (IWQOS), pp. 1-10, DOI: 10.1109/IWQOS52092.2021.9521291, 2021. 

[58] M. Yamazaki and M. Yamamoto, "Fairness Improvement of Congestion Control with Reinforcement Learning," Journal of Information Processing, vol. 29, pp. 592-595, DOI: 10.2197/ipsjjip.29.592, 2021.

[59] K. Lei et al., "Congestion Control in SDN-based Networks via Multi-task Deep Reinforcement Learning," IEEE Network, vol. 34, no. 4, pp. 28-34, DOI: 10.1109/MNET.011.1900408, July 2020.

[60] W. Li et al., "SmartCC: A Reinforcement Learning Approach for Multipath TCP Congestion Control in Heterogeneous Networks," IEEE Journal on Selected Areas in Communications, vol. 37, no. 11, pp. 2621-2633, DOI: 10.1109/JSAC.2019.2933761, Nov. 2019.

[61] D. Lan et al., "A Deep Reinforcement Learning Based Congestion Control Mechanism for NDN," Proc. of the 2019 IEEE Int. Conf. on Communi. (ICC), pp. 1-7, DOI: 10.1109/ICC.2019.8761737, 2019.

[62] N. Jay et al., "A Deep Reinforcement Learning Perspective on Internet Congestion Control," Proc. of the 36th Int. Conf. on Machine Learning (PMLR), pp. 3050-3059, 2019.

[63] H. Shi and J. Wang, "Intelligent TCP Congestion Control Policy Optimization," Applied Sciences, vol. 13, no. 11, p. 6644, DOI: 10.3390/app13116644, Jan. 2023.

[64] O. Habachi et al., "Online Learning Based Congestion Control for Adaptive Multimedia Transmission," IEEE Transactions on Signal Processing, vol. 61, no. 6, pp. 1460–1469, Mar. 2013.

[65] W. Li et al., "QTCP: Adaptive Congestion Control with Reinforcement Learning," IEEE Transactions on Network Science and Engineering, vol. 6, no. 3, pp. 445–458, July 2019.

[66] M. Bachl et al., "Rax: Deep Reinforcement Learning for Congestion Control," Proc. of the IEEE Int. Conf. on Communications (ICC 2019), pp. 1–6, DOI: 10.1109/ICC.2019.8761187, 2019.

[67] J. Yang et al., "IEACC: An Intelligent Edge-aided Congestion Control Scheme for Named Data Networking with Deep Reinforcement Learning," IEEE Transactions on Network and Service Management, vol. 19, no. 4, pp. 4932–4947, Dec. 2022.

[68] R. Galliera, et al., "MARLIN: Soft Actor-Critic Based Reinforcement Learning for Congestion Control in Real Networks," Proc. of the IEEE/IFIP Network Operations and Management Symposium (NOMS 2023), pp. 1–10, DOI: 10.1109/NOMS56928.2023.10154210, 2023.

[69] A. Sacco et al., "Owl: Congestion Control with Partially Invisible Networks via Reinforcement Learning," Proc. of the IEEE Conf. on Computer Communications (IEEE INFOCOM 2021), pp. 1–10, DOI: 10.1109/INFOCOM42981.2021.9488851, 2021.

[70] R. Bhattacharyya et al., "QFlow: A Learning Approach to High QoE Video Streaming at the Wireless Edge," IEEE/ACM Transactions on Networking, vol. 30, no. 1, pp. 32–46, Feb. 2022.

[71] V. Sivakumar et al., "MVFST-RL: An Asynchronous RL Framework for Congestion Control with Delayed Actions," arXiv: 1910.04054, DOI: 10.48550/arXiv.1910.04054, May 2021.

[72] R. Galliera et al., "Learning to Sail Dynamic Networks: The MARLIN Reinforcement Learning Framework for Congestion Control in Tactical Environments," Proc. of the IEEE Military Communi. Conf. (MILCOM 2023), pp. 424–429, DOI: 10.1109/MILCOM58377.2023.10356270, 2023.

[73] A. Shahzad et al., "RS-RLNC: A Reinforcement Learning-based Selective Random Linear Network Coding Framework for Tactile Internet," IEEE Access, vol. 11, pp. 141277–141288, 2023.

[74] D. Markudova and M. Meo, "ReCoCo: Reinforcement Learning-based Congestion Control for Real-time Applications," Proc. of the 2023 IEEE 24th Int. Conf. on High Performance Switching and Routing (HPSR), pp. 68–74, DOI: 10.1109/HPSR57248.2023.10147986, 2023.

[75] A. Bentaleb et al., "BoB: Bandwidth Prediction for Real-time Communications Using Heuristic and Reinforcement Learning," IEEE Transactions on Multimedia, vol. 25, pp. 6930–6945, 2023.

[76] S. Emara et al., "Pareto: Fair Congestion Control with Online Reinforcement Learning," IEEE Transactions on Network Science and Engineering, vol. 9, no. 5, pp. 3731–3748, Sept. 2022.

[77] L. Jia et al., "ZiXia: A Reinforcement Learning Approach via Adjusted Ranking Reward for Internet Congestion Control," Proc. of the IEEE Int. Conf. on Communications (ICC 2022), pp. 365–370, DOI: 10.1109/ICC45855.2022.9838901, 2022.

[78] A. R. Andrade-Zambrano et al., "A Reinforcement Learning Congestion Control Algorithm for Smart Grid Networks," IEEE Access, vol. 12, pp. 75072–75092, DOI:10.1109/ACCESS.2024.3405334, 2024.

[79] X. Liao et al., "Towards Fair and Efficient Learning-based Congestion Control," arXiv: 2403.01798, DOI: 10.48550/arXiv.2403.01798, Mar. 2024.

[80] I. Rhee et al., "CUBIC for Fast Long-distance Networks," Request for Comments RFC 8312, Internet Engineering Task Force (IETF), DOI: 10.17487/RFC9438, Feb. 2018.