[Gghc-discuss] topic of project

[Jonathan Manton]

In terms of the robot brains and wireless, I'd agree that Arduino and  
Xbee are the way to go.  Same with the base station - you avoid a lot  
of learning curve by using a laptop with Linux, rather than trying to  
deal with an embedded platform.

I would suggest having at least one sensor, either a touch sensor  
(just a stiff wire hooked up to a limit switch) or light/dark sensor.   
That allows you to do line-following or obstacle-avoidance, things  
that are pretty simple algorithmically and fun (my 11-year-old  
daughter has done this using Lego Mindstorms robots).

Some quick links as suggestions:
XBee shield at SparkFun:  http://www.sparkfun.com/products/9976
XBee pro module at SparkFun:  http://www.sparkfun.com/products/8742
XBee explorer USB at SparkFun:  http://www.sparkfun.com/products/8687
Motor Shield kit at Adafruit:  http://207.58.139.247/index.php?main_page=product_info&cPath=17_21&products_id=81
DFRobot 2WD Mobile Platform for Arduino at RobotShop (to speed initial  
development):  http://www.robotshop.com/dfrobot-2wd-mobile-platform-for-arduino-4.html

There are some other motor shields, plug-and-play sensors, and other  
stuff at RobotShop.  E.g., http://www.robotshop.com/dfrobot-en.html

Probably would also need battery power.  Easy way is to use  
rechargeable AA batteries.  Slicker way would be to use a LiPo battery  
pack on the robot.  E.g.,
Battery charger circuit (USB-powered):  http://www.sparkfun.com/products/10161
Battery pack (many, many options here):  http://www.sparkfun.com/products/339

Might need a step-up voltage converter to drive the motors - not sure,  
they might work on 3.7V.  I think everything else runs at 3.3V.  If  
not, you might need two LiPo packs and maybe two chargers.  Not sure.


I think the *dynamic* mesh network stuff is kind of ambitious, unless  
we can identify some canned software that is already out there and/or  
have some experts on the team that are conducting research in this  
area.  Some parts of that (e.g., network topology discovery and packet  
routing in a dynamic wireless network, motion planning based on  
wireless link continuity constraints) are PhD dissertation-level  
projects.

The good news is that if you remove the "dynamic" part from the  
wireless network, assume that all the nodes are connected to each  
other at all times, and that the bandwidth between two nodes is  
unaffected by what is going on with other nodes, the really hard parts  
pretty much go away.  Using ZigBee in the range of a gym or playground  
(< 100m) you can pretty much achieve those assumptions, unless you  
intentionally ratchet down the power or add interference.