After some experiments with Lego Mindstorms, we decided it was time to build a non-Lego robot...
Mechanics
The first major design decision was based on the following quote from “Back to the future”:
Marty
McFly:
Wait a minute, Doc. Ah... Are you telling me you built a time
machine... out of a DeLorean?
Dr.
Emmett Brown: The way I see it, if you're gonna
build a time machine into a car, why not do it with some style?
We decided to use a Tamiya TL-01 frame with a Carson Tuning Set. This turned out to be a stable, powerful and fast base. It is quite a chalange to control this power.
Electronics
The standard (mechanical) speed controller made it difficult to drive at low speeds. So it was replaced with a standard RC ESC (Electronic Speed Controller)., which gave much more controle over the motor power.
- An
odometer with a resolution of less then 0.5 cm.
- The
Devantech
Ultrasonic range finder wass tested, but got a lot of
reflections in the outdoor environment, which made it
useless. (2005 update: We know now this is caused by the high
sensitivity of this sensors. Reducing the sensitivity of this
type of sensor probably would have fixed the issue mentioned).
- The
Devantech
Digital Compass was added, which proves to be quite accurate,
even with the high-current motor in the vicinity. There is a some
deviation, because the sensor is not mounted fully horizontal.
- We
implemented a Fuzzy Logic controller to control the speed and steering
(which is quite a challenge with the speed-range this car has!). This
controller performs very well.
- Goniometric
routines to calculate routes, based on our current position and our
goal were added
The above enabled us to predefine a pattern in terms of speed, direction and distance and have the Roadbot smoothly execute it. The odometer and compass worked quite accurate. Initially, when driving a 20m square (80 meters length), Roadbot finished within 4 meters. The error was mainly the result of non-linearity of the compass and a basic correction made Roadbot finish within 2 meters. A more accurate correction and fusion of the compass-data with steering-data were never implemented.
We learned a lot of handling sensors and -
more important - sensor noise. The fuzzy logic controller performed
very well.
The application exhausted the microcontroller - both available PWM-outputs were in use and all of
the processing power was required for the fuzzy logic
controller (even when it was highly optimized), the 10khz interrupt
service routine for the odometer and and the i2c functions.
Links: My Robots