What is wrong with broadcast probing based ETX estimation for wireless links?

ETX (Expected Transmission Count) is shown to be the best link metric in terms of higher throughput for wireless multi-hop network with stationary-nodes having single radio. Some recent work, however, shows that ETX may exhibit high sensitivity to interference and a single TCP flow can increase the metric value by 10000% [1]. Such a change indicates that routing protocols, based on ETX, might be playing with random numbers without bringing any significant benefits of considering link quality. It is suggested that broadcast-based estimation of ETX could be the major cause of sensitivity as other metrics based on unicast probing method did not show similar performance degradation [1]. Several proposals came forward so far focusing on better estimation techniques of ETX metric mostly using unicast probing combined with cross-layer information; however, we have not seen any investigation study as to why does broadcast-based method result in inaccurate estimation of ETX. This paper presents preliminary results of an investigation study about the sensitivity of ETX estimation based on broadcast probing and compare it with the one based on unicast probing. It is work-in-progress.