Utilizing model predictive controllers (MPC) as an active queue management scheme is investigated in this paper. Model based prediction of future output and determining optimized value of the control signal have made MPC as an advanced control strategy in various modern control systems. In this paper a new approach is proposed to alleviate the computational complexity of MPC in order to implement in fast dynamics systems like computer networks. Neural network approximation of MPC as an active queue management (AQM) method implemented here not only has less computational burden with respect to the common MPC approaches, but also results in better performance compare to the well-known AQM methods such as random early detection (RED) and proportional-integral (PI) control. The proposed AQM approach is implemented in a field-programmable gate array (FPGA) system and its feasibility is investigated by timing analysis.