Hi all,
 
I'm developing a swarm robotics application that will require a group of SRV-1's to operate continuously for a long period.  I was very surprised and disappointed to find that the behavior of a fully charged SRV-1 running around on a table top changes dramatically after about 2 minutes.  The program (written in picoC) processes the camera image and will decide to move straight, make a random turn, or back up.  The actual motor command used is motors(50, 50) where the sign of the 50 will depend on the robot's current state.  The motor command is followed by:
 
    delay(50);
    motors(0, 0);
    delay(50);
 
In the following video you can see that the behavior changes abruptly after about 2 minutes (this roughly coincides with when the robot gets hung up on a block, but I am sure this is unrelated).
 
    http://www.cs.mun.ca/av/pub/SRV1First.mpeg
 
I also did a test with the robot sitting on a block with its treads in the air and executing motors(50,50) followed by the delay, motors, delay sequence shown above.  After about 18 minutes of continuous operation there is an abrupt transition and the motors slow to a crawl.  The "battery()" function returns a value of 1 throughout this test.  Again the robot was fully charged before the test began.
 
I know that one potential solution may be the addition of some form of feedback (from wheel encoders or the NAV board).  However, I would first like to understand what is going on here.  If anyone has any tips or suggestions, they would be most appreciated.
 
Best,
Andrew
 

Thanks for the suggestions, T. Jump.  However, I am fairly sure it is not related to my code.  I have just conducted the test you suggested.  I uploaded and executed the following picoC program:
 
    motors(50,50);
 
I put a fully-charged robot on a block so that the tracks were free to spin.  It ran happily for 7 minutes, maintaining a roughly constant speed.  After 7 minutes some subtle changes in the oscillation seemed to arise.  After 9 minutes the motors stopped quite abruptly and the only activity was an electronic whining sound.  The aluminum chassis appeared to be at room temperature (or possibly slightly above, but not at all hot).
 
This is a serious problem for my application because it seems to limit running time on the ground to 2-3 minutes.  Perhaps this can be extended by changing the frequency of activation of the motors, but I'm sure there is a better solution.
 
Andrew

I think it is the H-bridge overheating!  This IC is not directly accessible but by unscrewing the radio/motor card from the base of the SRV-1 I was able to insert a digital thermometer directly underneath it.  I ran the same test as before (executing motors(50,50)).  The thermometer I used was designed for body temperature.  Its operating range is 32 - 43 degrees C.  Initially I couldn't get a reading because the temperature was below 32.  However, after 1 minute I was seeing a temperature of 32.  From the 1 minute mark to 5 minutes the temperature rose steadily to 43.9.  After this point the thermometer would not display a value.
 
I expected to see the motors stall at about the 9 minute mark as before.  However, they kept going and going!  I ended the test after 30 minutes.  My assumption is that by lifting the radio/motor PCB off the base I created an air gap that was able to cool things down.  The datasheet for the FAN8200 motor driver indicates that the normal operating range is -20 to +75 C.  It also mentions a built-in thermal shutdown function.  I strongly suspect that this was the source of my problem (well, actually overheating was the real culprit).  There is actually a small plastic cavity that the radio/motor PCB sits on which completely isolates the area underneath the FAN8200 from the ambient air.  You could probably cook a tiny cake in there!
 
My current plan is to insert some spacers to raise the radio motor board off its base and hopefully allow the h-bridge to cool passively.  I will let you know how this goes.
 
Thanks again T. Jump!
 
Andrew

Hi again,
 
I did another test on the build up of heat around the FAN8200 h-bridge IC.  In this case I attached a linear active thermistor (MCP9701) to the PCB directly below the FAN8200.  After executing motors(50, 50) I tracked the temperature every 30 seconds.  The temperature rose from 26 degrees (a few seconds after startup) to a peak of 64.6 C after 11.5 minutes.  I terminated the test at this point and the temperature slowly dropped back to room temperature.
 
Note that this test was applied with 0.5 inch standoffs separating the radio/motor board from the robot's plastic base.  However, opening this small air gap is not enough.  The FAN8200 is completely boxed in from above by the Lantronix card.  When the robot is running about on a tabletop it still shows an abrupt reduction in motor performance after 3-4 minutes.  I am investigating adding a heat sink to the FAN8200 or replacing it with another motor driver that has a built-in heat sink.  However, I haven't been able to find any suitable replacements yet.  I would be grateful if people could post any ideas for replacing the FAN8200 or finding some other means to keep it cool.
 
Cheers,
Andrew