This thesis propose Ripple Flooding Scheme (RFS) which goal is to achieve extreme fast convergence speed to meet the stringent timing requirement of some sensor network applications such as time synchronization for distributed signal processing.
A typical distributed signal processing application as sniper localization system use Rapid Time Synchronization (RaTS) to synchronize all sensors nodes within 30 $\micros$ error require re-synchronization in every 30 s. However the convergence time of the synchronization process itself takes 4 s, which is 13% overhead. With the help of RFS, synchronization process in the same size network can be finished in 25 ms, which is only 0.08% overhead.
Instead of using CSMA MAC to avoid collision by transmitting rebroadcast packets at different time, RFS schedule nodes to forward flooding packet at the same to avoid collision. The feasibility of RFS is carefully studied from theory to experiment, and there is 15% packet lost in a single level forwarding due to delay uncertainty among the senders. RFS use three level rebroadcast and overhearing-retransmission scheme to recover the single transmission packet lost and achieve good overall reliability.