IP Packet Loss and RTT Calculation Simulation Using Low-Cost Embedded Real-Time Systems

摘要

Research and development in the field of congestion control in computer networks is involving machine learning. However, the training process can be expensive and complicated because of the need for a pre-existing computer network with which to grade the success of an algorithm. Network simulators such as ns-2 and ns-3 also may not help because they are not real-time simulators. This paper proposes to capture the pattern of data that the neural network is identifying and the embedded system records. The interpolated data with similar patterns are generated and presented to the neural network to test if the trigger has been successfully identified. For the purposes of this paper, the data is a record of delays in multiple TCP packets.