SUMMARYAn increasing number of different applications face the challenge of providing end-to-end quality of service (QoS) support such as bandwidth, delay, jitter, and packet loss. In this paper, we have focused on DiffServ architecture to improve its accuracy. We proposed a new algorithm, called Adaptive Model Predictive Control Scheduler (AMPCS), to schedule differentiated buffers in routers, using Adaptive Model Predictive Control as the controller. AMPCS regulates the service rates of aggregated traffic classes dynamically in a way that some constraints on proportional delay or absolute delay can be guaranteed. Our contribution is to apply a model predictive controller to the scheduling problem to control the QoS parameters accurately. Simulation results show that the AMPCS performs acceptable QoS differentiations at core routers while it maintains end-to-end delay constraints.