[mjcoss]

As I've said in other threads. To me there a several issues all intertwined, and it would be nice if we could decouple them.

First, all the robot are sharing the same wide channel @ 5Ghz. So they will compete for whatever bandwidth is available. How they behave will be dependent on the configuration of the various robots on the field. At best, you're looking at 150 Mbits for 6 robots, and even that is probably a stretch. Stream cameras, smart dashboards, custom dashboard, out of band messaging, and driver station traffic, and for those of us using 2CAN CAN bus traffic within the robot. It's a mess.

Second, the DLINKs seem to have a pretty high failure rate. Some have been known to work only when mounted vertically. Others intermittently fail. Is it power to the unit? Is it the barrel connector? Is it the PCB inside the DLINk? Is it firmware issues on the DLINK itself? What about the voltage regulator?

I set up a hardware capture during the MAR championship, and saw a number of teams stuck in a request to connect loop with their DLINKs and the field. Why some teams had this issue and others did not is unclear. The DLINKs are clearly not designed to be used on robots, powered as they are, and manhandled as only a FRC competition can. Also, I saw a lot of corrupted packets.

I have heard but cant confirm that there is an issue with the DLINKs and the number of SSIDs that they observe. If the number exceeds some threshold, they fail. It's certainly possible that this could be an issue.

I'd like to setup the network with 1 channel per robot. It would mean that they'd need 6 access points for the field, but it would increase the available bandwidth for all robots on the field, and each would be isolated from the other robots. There is enough spectrum that you can run 6 independent 5Ghz wide channels on the field. It would eliminate one variable from the equation.

Replacing the DLINK with a rugged AP would be a good thing as well.

And finally real time monitoring of the power to the DLINK or what ever AP is used would give a picture of what is happening as robots connect, and run during a match.

I work for BL research, and I offered to MAR to have some of our network researchers take a look at this problem, as I believe it is a real problem with the field, and not *just* a robot issue. The response was that they are looking into the problems. At the end of the day, someone will say that the problem is "fixed", and we will revisit the issues next year with a new set of parts and robots and we will never know the root causes.

Just my thoughts.