On Apr 15, 2008, at 1:48 AM, Michiel Konstapel wrote:
> We have found that CTP is sensitive to short bursts of packet loss
> on otherwise good radio links (ETX 1.0 to 1.2) and found something
> that may be of interest. LinkEstimatorP uses an ALPHA parameter of 2
> (actually 0.2) for decaying its exponential average of eetx:
>
> void updateEETX(neighbor_table_entry_t *ne, uint16_t newEst) {
> ne->eetx = (ALPHA * ne->eetx + (10 - ALPHA) * newEst)/10;
> }
>
> Written as it is, this means it takes 80% of the new estimate and
> 20% of the old – is it possible that this was supposed to be the
> other way around? We find that a short burst of lost packets will
> quickly increase ETX to levels where CTP will select a new parent,
> which causes excessive beaconing as CTP resets its beacon interval
> and other nodes start switching parents as well.
>
What TinyOS version are you using? CVS uses an alpha parameter of 9.
The alpha of 2 was from before we tested CTP in networks with bursts
of losses.
Phil
_______________________________________________
Tinyos-help mailing list
[email protected]
https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
