[Tinyos-help] More reliable radio messaging, anyone?

Michael Schippling schip at santafe.edu
Sun Sep 11 14:27:20 PDT 2005


I'm curious if there is any interest in making a TOS radio
messaging system that is more reliable, or at least easier
to recover. From reading papers on mica and mica2 reliability,
it seems that 90-95% of close range message reception is the
best that can be expected from the current code. I think
that there is little interest in improving this because:
1) there are no equivalent studies of micaz messaging, and
2) the xbow people seem rather blase about the whole thing.

Just a couple top'o'th'head suggestions to start off:

The radio message structure should contain the source ID,
a sequence number, and maybe some extra bits to toggle. This
would make it easier to push retry protocols down into the stack.

And some attention should be paid to the receive-ACK mechanism,
which currently seems to depend on there being dead air after
eash message -- because the ACK does not contain any source or
message ID.

So, are there enough people out there who would like to do
something about it? I'm working on a self-funded project that
can squeek by with the current stuff, but I won't be able to
scale it up to get external funds without some changes...

I'm taking my little mica-bot guys to NYC for a conference
this week, so I may not be available with my usual so-helpful
advice for a while. Don't give up on me if I don't reply right
away.

MS

ps...I was thinking about the 1/50000 CRC false positive thing.
At first blush the CRC is just a comparison of two 16 bit numbers
which has a 1/64000 probability of being true, so maybe 1/50K is
about the best we can expect?


More information about the Tinyos-help mailing list