A. Nakajima’s research while affiliated with The University of Tokyo and other places

In this paper, we deal with congestion control of one bottleneck type for TCP/AQM network systems and investigate the H 2 control performance limitation. To this end, we first represent the dynamics of the congestion window size at source nodes and the queue length at the bottleneck link by a discrete time model. We then prove the stability of the plant and provide the Youla parametrization of all stabilizing controllers. With this parametrization, we consider an H 2 optimization problem of the queue length at the bottleneck link. We derive an explicit form of the best achievable H 2 norm and discuss the control performance limitation with respect to the ratio of the transmission time of the packets from source nodes to the link and the round trip time.

In this paper, we deal with congestion control of one bottleneck type for TCP/AQM network systems. We represent the dynamics of the congestion window size at source nodes and the queue length at the bottleneck link by discrete time systems in order to derive the parametrization of stabilizing controllers for the perturbation systems from equilibrium states. We prove the stability of the plant and give the Youla parametrization of the stabilizing controllers. With this result, a control performance problem of the queue length at the bottleneck link is considered and H2 optimization problem under a condition of the throughput at the bottleneck link is solved. We show that there exists a nonzero optimal queue length for this problem.

In this paper, we deal with congestion control of one bottleneck type for TCP/AQM network systems. We represent the dynamics of the congestion window size at source nodes and the queue length at the bottleneck link by discrete time systems in order to derive the parametrization of stabilizing controllers for the perturbation systems from equilibrium states. We prove the stability of the plant and give the Youla parametrization of the stabilizing controllers. With this result, a control performance problem of the queue length at the bottleneck link is considered and H/sub 2/ optimization problem under a condition of the throughput at the bottleneck link is solved. We show that there exists a nonzero optimal queue length for this problem.

In this paper, we propose two identification methods for a class of multi-input multi-output discrete-time linear parameter-varying systems. The methods are based on the subspace state space method. We present numerical simulations to demonstrate the efficacy of the methods.