NETWORK TOMOGRAPHY - ESTIMATING SOURCE-DESTINATION TRAFFIC INTENSITIES FROM LINK DATA

The problem of estimating the node-to-node traffic intensity from repe ated measurements of traffic on the links of a network is formulated a nd discussed under Poisson assumptions and two types of traffic-routin g regimens: deterministic (a fixed known path between each directed pa ir of nodes) and Markovian (a random path between each directed pair o f nodes, determined according to a known Markov chain fixed for that p air). Maximum likelihood estimation and related approximations are dis cussed, and computational difficulties are pointed out. A detailed met hodology is presented for estimates based on the method of moments. Th e estimates are derived algorithmically, taking advantage of the fact that the first and second moment equations give rise to a linear inver se problem with positivity restrictions that can be approached by an E M algorithm, resulting in a particularly simple solution to a hard pro blem. A small simulation study is carried out.