Active Queue Management in Network Control Systems

Active queue management (AQM) represents a family of control strategies designed to regulate packet buffering in network routers and switches so as to forestall the onset of congestion. By monitoring queue length or delay, an AQM algorithm dynamically adjusts packet-dropping or marking probabilities to maintain buffer occupancy near a set point. This approach prevents long queue build-up and minimises latency, jitter and packet loss even under rapidly changing traffic conditions. Modern network control systems integrate AQM with feedback models of Transmission Control Protocol (TCP) to shape the end-to-end flow of data, leveraging control-theoretic tools such as proportional-integral-derivative (PID) controllers, disturbance observers, backstepping and sliding-mode observers. Advances in modelling have deepened understanding of time-delay effects intrinsic to transport networks, yielding robust and adaptive designs that ensure stability and performance in data centres, wireless networks and industrial communication. The convergence of AQM with machine learning and model-predictive control has further broadened its scope, enabling real-time anomaly detection and self-tuning behaviour in complex, heterogeneous environments.

Research from Nature Portfolio

Recent studies have extended adaptive control methodologies to nonlinear TCP/AQM systems with uncertain control gains. A novel controller was introduced by constructing an auxiliary system to handle unknown items alongside known virtual control coefficients. This design ensures practical boundedness of all signals in the closed-loop network and achieves precise queue-length tracking. Simulation results demonstrate the controller’s effectiveness and practicability under varying delay and traffic conditions, highlighting its potential to improve stability and throughput in real-world deployments.

Research from all publishers

Researchers have combined disturbance observers with Smith predictors to compensate simultaneously for modelling errors and large time delays in TCP/AQM networks. By incorporating a saturation function into the packet-drop probability, the integrated controller achieves higher throughput and goodput than classical PID and PD designs. Under typical delays of 100 ms and bottleneck capacities of 100 Mbps, the approach delivered near-optimal performance, reducing queue oscillations and stabilising data flows.

Another line of work utilises adaptive neural backstepping to develop finite-time congestion controllers for TCP/AQM networks with unknown hysteresis and external disturbances. The controller guarantees semiglobal practical finite-time stability, ensuring the queue length converges to its target within a finite time horizon. Simulation studies confirm significant reductions in settling time and robustness against traffic variations.

Fuzzy logic has also been applied to PID-like AQM controllers for wireless TCP networks, addressing the challenges of variable delay and user mobility. By tuning membership functions through optimisation algorithms, the fuzzy-PID controller achieves improved delay performance and reduced packet loss compared with standard PID schemes, demonstrating its suitability for next-generation wireless environments.

Active Queue Management in Network Control Systems publication trend

The graph below shows the total number of articles in active queue management in network control systems across all publications each year (not limited to Nature Index journals).

Technical terms

Active Queue Management (AQM): A family of router-based techniques that regulate packet queue lengths by dynamically adjusting drop or mark probabilities to prevent congestion.

Transmission Control Protocol (TCP): A core transport-layer protocol that adapts its sending rate based on network feedback to ensure reliable, ordered delivery of data.

Congestion Control: Strategies and algorithms to regulate traffic entering a network to maintain performance metrics such as throughput, latency and packet loss.

Backstepping: A recursive control design method that decomposes a nonlinear system into subsystems, allowing systematic synthesis of stabilising controllers.

Disturbance Observer (DOB): A model-based estimator used to reconstruct and compensate for uncertainties or external disturbances in control systems.

Smith Predictor (SP): A time-delay compensation technique that uses a model of the plant and delay to predict future output and adjust control actions.

Finite-Time Stability: A property of a control system whereby system states converge to an equilibrium within a finite time interval, rather than asymptotically.

References

1. Adaptive tracking control for nonlinear systems with uncertain control gains and its application to a TCP/AQM network. Scientific Reports (2023).
2. Active Queue Management Supporting TCP Flows Using Disturbance Observer and Smith Predictor. IEEE Access (2020).
3. Adaptive Finite‐Time Congestion Control for Uncertain TCP/AQM Network with Unknown Hysteresis. Complexity (2020).
4. Fuzzy type 1 PID controllers design for TCP/AQM wireless networks. Indonesian Journal of Electrical Engineering and Computer Science (2021).