Bob has had a bit of minor surgery. I replaced the Darlington transistor motor driver with a L293D chip. Thing is, I forgot that the 293 has relatively large losses so Bob was much slower…
The idea was that the 293 would allow the motors to brake harder. With one side of the bridge permanently grounded, the motors would be actively driven or regeneratively braked. That worked fine but there simply was not enough voltage to get the motors up to any useful speed in the first place. After a poke about in the battery collection I found a LiPo 2-cell pack that would just fit in the battery compartment. I think the capacity is about 600mAh so it should be good for a couple of hours running.
The increase in speed was surprising and Bob nearly shot off the test track before I could catch it. However, when the speed is reduced to a level where it can reliably negotiate turns, I think I am no better off. The video shows that it still takes nearly 10 seconds to get around my little test track. I was hoping for more like 5 seconds but it just can’t turn hard enough when it encounters a bend to prevent the sensors losing the track altogether. I know the design is fundamentally flawed but I expected more. Next I shall try allowing Bob to control the motor direction as well as speed but I think I am wasting my time with it. There is supposed to be a nice PD controller in there but that too may be a waste of time.
I have used the trick of (very) briefly turning a motor into reverse for braking on our teaching wall-followers, to improve the turning response. These use a single chopper FET and relay direction change.
Braking will be more much more controllable with the L293 h-bridge that you have fitted – but wouldn’t a reasonable control loop do this for you anyway?
The bot is looking good though – particularly as a pattern for schools competitors. It is in the end very similar to the matrixmultimedia bot, even down to the sound it makes.
Maybe a F1 sound generator for F1 could mask it!