Quote:
Originally Posted by Thad House
The big issue with the DLinks was that the frequency backoff was hardcoded into the radio, and the value was the same for every radio produced. This meant that if they ever interfered with eachother, that would all start trying to separate from each other, but they would do so at the same rate, which just caused them to keep interfering. That was the root cause for the Christmas Trees. The new radios have that number legitimately randomize, which actually actively allow the radios to not interfere with each other. So even though they were FCC certified, for our uses they had issues, and our options for radios have to be specialized so those issues don't happen.
|
Why would that be an issue? You could christmas tree with a single field and all the radios sharing 1 or 2 channels. All 6 robots share those channels if this were the case as you propose a single collision would take out the whole field and clearly that did not happen as packet loss was constant. Where were they going to frequency hunt when they are channel locked? Now on the few occasions they had single channels with no channel bonding there were some christmas trees but they also had lots of traffic from video cameras all trying to flood into the field and competing with the TCP mechanism for re-transmit. QoS is supposed to alter this behavior to give FMS the advantage without making the cameras retry more and more.
Furthermore: even if they follow the same path to degrade they are all at various orientations, distances and signal strengths so the odds are very low they'd actually all block in that way.