Abstract

To study dynamic aspects of network protocols, like the transmission network protocol TCP, and networked control systems (NCS) with a UDP-like flow in a single framework, we employed UPPAAL, a tool that simulates discrete event systems as switched automata. Delays and packet losses can degrade the performance of a control loop, leading in the worst case to instability. In opposition to standard communication tasks, the data throughput is not so relevant for the control task — the quality of control (here measured by the ITAE criterion) is adopted in this paper. We focus on processes controlled over standard Internet. First, we present a simulation of TCP-Reno protocol modeled through formalities for discrete event and timed automata systems using UPPAAL. Results were compared with simulations in the NS-2 simulator found in literature. Over this TCP-Reno model we added a control system over the network, where we can observe and study how the communication network affect control performance behavior for different simulation scenarios and with different Active Queue Management (AQM) techniques. In addition, in order to improve the tradeoff between TCP-throughput and control system performance we present a new AQM technique called Explicit Non-Congestion Notification (ENCN) whose performance is compared with Random Early Detection (RED), Control Delay (CoDel) and Proportional Integral controller Enhanced (PIE) schemes. According to our study, CoDel queue policy gives us the best control performance but worsens the TCP throughput; however, as an expected compromise, the ENCN give us better tradeoff between control performance and TCP throughput.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call