Analysis and design of AQM based on state-space models for stabilizing TCP

In this paper, we formulate the AQM (active queue management) design problem for stabilizing a given TCP (Transmission Control Protocol) as state-space models. Thereby, we show that PD-type (proportional-derivative) AQM is a natural state-feedback control structure to stabilize the given TCP for the first time, to our knowledge, in the networking literature and by applying integral control action, a PID-control structure that is a unified framework for analysis and design of AQM. Next, we compensate for delays in congestion measure explicitly by using a memory control. Third, we obtain stabilizing optimal gains of the proposed feedback control structures for linearized systems of the given TCP. We interpret existing AQMs, including a simplified random early detection, random exponential marking, PI and a simplified adaptive virtual queue, as different approximations of the unified AQM structure. We discuss the impact of each structure on the performance from the results of the stabilizing optimal AQMs. Finally, we illustrate our results through simulations for TCP Reno.