Abstract Congestion control is among the most fundamental problems in computer networking. With the networks becoming increasingly complex and saturated, efficient control of internet traffic congestion has become more important than ever. Transmission Control Protocol (TCP) is a set of procedures that facilitate reliable data transfer across a network. TCP is reliable in the sense that any segment of data that is lost or corrupted is retransmitted, and all packets of data arrive at the destination in the order that they are sent. Although TCP guarantees reliable data delivery, the physical and computational limitations of network hardware and software cause congestion resulting in wasted resources and transmission delays [1]. The development of a dynamic model of TCP has encouraged the application of control principles to Active Queue Management (AQM) schemes [3]. In particular, [2] demonstrated that the inherent presence of queuing delay in internet traffic cannot be ignored and thus approached it as an optimal control problem of a state-dependent delay system. In this project, we applied a novel Model-free control (MFC) [4] strategy to reduce congestion. While traditional approaches rely on the accuracy of the TCP behavior model, MFC is robust against modeling errors. Simulation results showed that our control algorithm can effectively reduce the queue length even in the presence of measurement noise. Mathematical analysis and numerical simulation results of our control algorithm will be presented. Moreover, the potential applicability, as well as the limitations of model-free control methods in systems with inherent state-dependent and input-action delay, will be discussed.