Heterogeneous communication networks with their variety of application demands, uncertain time-varying traffic load, and mixture of wired and wireless links pose several challenging problem in modeling and control. In this paper we focus on the roundtrip time (RTT), which is a particularly important variable for efficient end-to-end congestion control. Based on a simple aggregated model of the network, an algorithm combining a Kalmanfilter and a change detection algorithm is proposed for RTT estimation. It is illustrated on real data that this algorithm provides estimates of significantly better accuracy as compared to the RTT estimator currently used in TCP, especially in scenarios where new cross-trafficflows cause a bottle-neck link to rapidly build up a queue, which in turn induces rapid changes of the RTT.