Delay-based congestion avoidance for video communication with fuzzy logic control

Early detection of network congestion is important in streaming video, as packet loss has a recurring impact on video quality. Packet delay itself, rather than packet loss, can give early notice of congestion provided it can accurately reflect the congestion level at the network path's tight link. In this paper, the one-way delay of video packets serves as an incipient network congestion indicator, which acts as one input to fuzzy logic control of congestion avoidance to help optimize the response to network congestion. The fuzzy logic models are shown to be robust under: changes in the complexity and motion content of the video stream under control; a wide range one way end-to-end link delays up to 120 ms; and variations in available bandwidth. Tests reported that the fuzzy logic approach compares favorably to standard TFRC congestion control. ©2007 IEEE