This robot was a plywood box about 2 feet square and a foot high with a pair of fairly high torque DC motors (servo motors from a tape drive, it turns out) that were directly coupled to a pair of 8-10" wheels, and some casters at the other end of the box. A car battery (originally) or some gel-cells (eventually) provided the traction power.
My friend, Doug Harber, and I spent a while figuring out how to connect the wheels to the shafts of the motors, before hitting on the interesting approach of screwing circular saw blades to the tape reel hubs, then gluing (using silicone) to the wheels. A terrible design for longevity, as it happens, because it puts big radial loads on the motor bearing.
Driving the motors at 12V made the thing scoot pretty quick (I think the nominal speed was probably something like 20 ft/second). An early attempt at a controller ended in disaster when it shorted ON, the box ran into the stucco wall (punching a hole), the car battery fell off the top, etc.
Later, I designed a transistor PWM controller using darlington power transistors (TIP120's?) driven by a wirewrapped Z80 and 8254 timer board. This controller literally exploded when I did a "plug reversal" (i.e. run full speed one way, then shift to full speed the other way) with the wheels off the ground. Since I was working on it in my office at work during lunch time, the loud bang (from the darlington transistor catastrophically failing) and (small) cloud of smoke did attract some attention. I then decided that a) one should have better matching of desired speed and motor characteristics and b) building motor controls is non-trivial.
Oddly, I have no idea where the motors from that robot went.
A big lesson learned is that you need to consider the gearing and speeds.