On the Optimal Blacklisting Threshold for Link Selection in Wireless Sensor Networks

Empirical studies on link blacklisting show that the delivery rate is sensitive to the calibration of the blacklisting threshold. If the calibration is too restrictive (the threshold is too high), all neighbors get blacklisted. On the other hand, if the calibration is too loose (the threshold is too low), unreliable links get selected. This paper investigates blacklisting analytically. We derive a model that accounts for the joint effect of the wireless channel (signal strength variance and coherence time) and the network (node density). The model, validated empirically with mote-class hardware, shows that blacklisting does not help if the wireless channel is stable or if the network is relatively sparse. In fact, blacklisting is most beneficial when the network is relatively dense and the channel is unstable with long coherence times.