Robotour 2007
Robotour 2007 is a second year of an outdoor competition of fully autonomous robots. "Fully autonomous" means that once the robot is started, it has to go by itself, with no remote control whatsover. The task is to travel around a prespecified route. This route is roughly one kilometer long.
Team members
- Pavel Jiroutek
- Jiøí I¹a
- Dan Polák
- Tomá¹ Stanìk
Hardware
Platform
The robot is based on the 1:8 RC Hummer toy car. In the beginning, this seems to be a good decision. Looking backward, it is not. Such a plastic toy is not designed for the weight and forces we need. Direction control was replaced, motors were rebuilt, ...
Sensors
- GPS: Although the competition spot is known for a very unreliable GPS signal from the previous year, our team decided to use GPS as a rough position indicator anyway.
- Odometry: Without encoders we would not even be able to control the speed of the robot.
- Camera: A very robust road tracking software was designed.
- Compass: An accurate sensor with no accumulative error.
Electronics
The software runs on a regular laptop. The low-level communication with hardware and its control is obtained through ATMega8-board. The motors use regular model regulators (40A).
Software
Idea
The idea is to combine a range of slightly inaccurate sensors into Monte-Carlo localization. The inaccuracies should cancel out one another. Due to the hardware problems we did not manage to finish the whole integration. Our rides were based on the pure vision-based road following, with encoder-based speed control.
Vision
I have developed an adaptive vision system (Visiir), which is used by the Short Circuits team.
 |
 |
Track following
Based on the current position, the robot also needs to decide, where to go next. I also wrote this piece of software. Including going reverse ... (which, together with low-level motor control, teared our motor axle apart just before the competition)
Track input
A route has to be given to the robot. I developed a tool for visual route input.
Position estimation
The critical part of the software is the fusion of information from sensors and position estimation. Pavel Jiroutek wrote an Monte-Carlo simulation based localization, which relies on odometry, GPS, camera and compass.
Video
Short Circuits 2007: Robot on the move