Hidden Markov model approach to TCP link state tracking
Andrey V. Borisov, Gregory B. Miller · 2004 43rd IEEE Conference on Decision and Control (CDC) (IEEE Cat. No.04CH37601) · 2004
A new model of TCP link state evolution In the settings of hidden Markov models (HMM) is introduced. A class of special Markov processes, which can be considered as generalization of the classic Markov chains, is suggested. Solution of the optimal filtering problem given combined noised and unnoised indirect observations is also presented. The result is implemented to the tracking problem for the couple "TCP link condition - round trip time (RTT)" given both the noised RTT measurements and packet losses. The presented estimate is compared with the standard realized in TCP and the optimal filtering one given packet losses only.