On Thu, Aug 13, 2009 at 1:15 PM, li peng<[email protected]> wrote: > Hi OM > > Based on your latest update of LinkEstimatorP.nc, after my test, I also > found the deficiency of the fix... > > As Mehmet mentioned, the link quality uncommonly becomes greater, because > the corresponding neighbor entry is always viewed as UN-MATURE, resulting > from the unfrequent execution of method updateNeighborTableEst... > > Actually, after a period time's mutual communication, the tree has been > formed. If one of the intermediate node is dead, then its children should > merge into other branches if these children > have the radio connection with other parents... But in my test, after a long > time running, these "isolated" nodes were unable to find new parent > throughly, this is still caused by same reason > shown above. I tested in the totally-no-loss and loss-existing environment, > for no-loss network, it is impossible to have Est updated since the > packetGap is always 1; even in the loss-existing > environment, it is really hard to get continuous BLQ_PKT_WINDOW's packet > loss, thus, the merge is hard to be shown. > > Your fix: > if (((NeighborTable[idx].rcvcnt + NeighborTable[idx].failcnt) >= > BLQ_PKT_WINDOW) > || (packetGap >= BLQ_PKT_WINDOW)) > > works much better than orginal fix in my test... And the tree's reforming > works as my wish flawlessly...
Thanks for testing this. I also did a quick test myself. I committed the fix. - om_p _______________________________________________ Tinyos-help mailing list [email protected] https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
