A robust PID and RLS controller for TCP/AQM system.

Abstract

Transmission Control Protocol (TCP) controlling congestion by peer-to-peer is challenging to handle communication with numerous concurrent TCP flows and heavy traffic loads. Therefore, TCP requires the active queue management (AQM) to assist in avoiding buffer bloat in intermediate devices. Thus, Some AQM controllers, such as Red, Codel, and Pie, have been proposed to control congestion better. However, these controllers are unable to overcome the drawbacks, including micro-burst streams, unfair competition of flows with different workloads, as well as sluggishly responding to congestion in heavy traffic, which degrade the performance of data transmission. To address these problems, in this paper, we proposed a queueing-delay-based AQM controller for TCP/AQM systems, namely PID-R. PID-R adopts two-level control components, i.e., the Proportional-Integral-Derivative (PID) algorithm and the recursive least squares (RLS) filter, and combines the advantages of the two parts. Meanwhile, we also constructed a new TCP/PID-R fluid model derived from TCP/AQM classical model. According to the model, PID-R parameters are determined to keep the TCP/PID-R system steady. The comprehensive simulation reveals that PID-R outperforms the state-of-the-art AQM controllers.