DDPG-Enhanced Integration of the BLUE Algorithm for Adaptive Active Queue Management and Congestion Control
Öz
Active Queue Management (AQM) mechanisms are also used in TCP/IP networks to alleviate congestion, decrease queueing delay, and increase throughput. In related work, the BLUE Algorithm (Adaptive) is an AQM that controls congestion by adjusting the likelihood of dropping packets in response to network events. However, the performance of these solutions is heavily dependent on a set of fixed configuration parameters, resulting in limited adaptation to dynamic network conditions. To address this limitation, a Deep Reinforcement Learning-enhanced version of it, called BLUE (DRL-BLUE), is proposed in this paper, using the methodology of Deep Deterministic Policy Gradient (DDPG). Unlike traditional methods, which fix parameter values at the beginning of the process, this contribution proposes a flexible approach for queue estimation and congestion responsiveness, enabling dynamic adjustments to the parameter wq (queue weight). Instead, the DDPG agent continues to learn adaptive control by interactively exploring the network environment over time, enabling the BLUE algorithm to configure itself to different traffic scenarios without any explicit traffic models. Our proposed framework is tested under low- and high-density network scenarios and compared with the classical BLUE algorithm in terms of throughput, delay, and packet loss ratio (PLR). Experimental results show that the proposed DRL-BLUE approach improves network performance in most evaluated scenarios. Figure 10 shows the throughput, packet loss ratio, and delay in a high-density scenario with 10 active senders; high density is defined as having more than 5 active senders competing to receive packets. The overall results show that the adaptive tuning queue weight parameter with DDPG enhances congestion control performance through higher throughput, lower packet losses, even in obstructed transmission attempts, and keeps low queueing delays when faced with changing data from network usages.
Anahtar Kelimeler
Kaynakça
- [1] S. Floyd and V. Jacobson, “Random Early Detection Gateways for Congestion Avoidance,” IEEE/ACM Transactions on Networking, vol. 1, no. 4, pp. 397–413, Aug. 1993.
- [2] V. Jacobson, “Congestion Avoidance and Control,” in Proceedings of ACM SIGCOMM, Stanford, CA, USA, 1988, pp. 314–329.
- [3] W.-C. Feng, K. G. Shin, D. D. Kandlur, and D. Saha, “The BLUE Active Queue Management Algorithms,” IEEE/ACM Transactions on Networking, vol. 10, no. 4, pp. 513–528, Aug. 2002.
- [4] R. Pan, P. Natarajan, C. Piglione, M. S. Prabhu, V. Subramanian, F. Baker, and B. VerSteeg, “PIE: A Lightweight Control Scheme to Address the Bufferbloat Problem,” in Proceedings of IEEE HPSR, 2013, pp. 148–155.
- [5] K. Nichols and V. Jacobson, “Controlling Queue Delay,” Communications of the ACM, vol. 55, no. 7, pp. 42–50, Jul. 2012.
- [6] T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. Heess, T. Erez, Y. Tassa, D. Silver, and D. Wierstra, “Continuous Control with Deep Reinforcement Learning,” arXiv:1509.02971, 2015.
- [7] R. S. Sutton and A. G. Barto, Reinforcement Learning: An Introduction, 2nd ed. Cambridge, MA, USA: MIT Press, 2018.
- [8] V. Mnih et al., “Human-Level Control Through Deep Reinforcement Learning,” Nature, vol. 518, no. 7540, pp. 529–533, 2015.
Ayrıntılar
Birincil Dil
İngilizce
Konular
Yapay Zeka (Diğer)
Bölüm
Araştırma Makalesi
Yazarlar
Erken Görünüm Tarihi
14 Temmuz 2026
Yayımlanma Tarihi
-
Gönderilme Tarihi
27 Ekim 2025
Kabul Tarihi
29 Haziran 2026
Yayımlandığı Sayı
Yıl 2026 Sayı: Advanced Online Publication