Congestion Control Policy Decision-making Based On Source And Achieve

This thesis discusses how to adapt video encoder's output rate to the network states offered by backward channel, under the IP network environment, which supplies no QoSs guarantee. The goal is to design a source-based real-time congestion control policy that not only utilize network resource reasonably, but also to arise no network congestion from transporting real-time video. There are three mechanisms for this congestion control: rate control, rate-adaptive video encoding, and rate shaping.Based on the characteristics of real-time video, there are two conditions for rate control: 1) inter-frame's delay constraints; 2) the constraints of the transport rate network. In the thesis, a new approach is proposed to convert the constraints of the inter frame delay to the constraints of the effective transport buffer capacity of the sender, by taking states of both transport buffer and decoder buffer into account. Model-based TCP friendly rate control method is used to calculate the available network bandwidth. The transmission rate control is based on the calculated bandwidth, and the throughput of video encoder is adapted to both the available network bandwidth and the effective transport buffer size.The adaptive strategy of used in this paper is Lagrangian's cost function based optimal forward adaptation, and encoding parameters is determined by the bandwidth and the buffer state.The paper also proposes and realizes:1. scheme of looking for the optimal Lagrangian multiplier by using the bisection method according to the rate constraint by the network bandwidth,2. and a fast searching algorithm for the optimal Lagrangian quantization vector for a given operator .Consider that the adaptive video encoding system is a long delay feedback control system, the I/O mismatch of the transport buffer is inevitable and a mechanism of rate shaping is required. To this end, an intelligent rate shape algorithmis suggested in the paper to match the transmitting rate and network bandwidth, which provides the primary strategy of selecting a quantization vector for the adaptive encoder and implements:1. an algorithm of controlling transmission rate to the network and of discarding certain frames intelligently, while a risk of overflow is forthcoming and the violation of delay constraints is encountered.2. an intelligent hopping-frame scheme in the encoder according to statistics of average bits of a GOP and the constraint condition.