By Amanda Puna and Marine Dubois, pupils in the 1'st OC info. at the Blaise-Cendrars High School, June 2012
General Presentation
For this project, we wanted to use the Thymio robots. Our project was to take 2 Thymios, to connect them together by means of a Lego structure and to make them walk, like a quadruped.
Problems encountered
Assembly
Our first problem was to find out how to put the Thymios together. For that, we took Lego pieces which we could fix to the robots at places foreseen for this. We had to find a quite robust structure which would not come apart when the robots are moving or when there is a small shock. We also tried to match the pieces a little, so as not to have a quadruped of many colours, by using mainly yellows and reds.
The paws
After having found a satisfactory structure, we looked how we could make the Thymios walk. At first we wanted to give them real "paws", which would have been like those of some animals and which would be fixed to the wheels, so that they could move thanks to the motor. But we realised that in order for the robot to walk more or less correctly, it would need joints in these legs. We then tried to put a joint on each paw, similar to our ankles or knees. But that did not work, because the joints gave way under the weight of the robot so that the robot did not stay up. So we gave up the idea of having paws like animals have.
Following these infertile experiments, we decide to turn to legs resembling wheels a little more. We used two thin Lego plates which we assembled in a cross. At each extremity we added another piece, square if possible, which serves as the foot. Then we fixed this assembly to the wheel of Thymio with another square piece. One Thymio thus has 8 "feet", 4 per wheel.
But this way of doing it was not strong enough. The cross got loose from the robot's wheels too easily. To get over this problem, we changed the square pieces connecting the crosses to the wheels of Thymio. We replaced them with circular pieces, which hold things together better without taking up more space. This is an idea which we found on the Aseba web site, watching a funicular robot working.
After having understood how to fix this system, we tried it in practice, with only one robot. We noticed that it did not "walk" really straight. We then had to modify the feet of the robots and make them all the same size. Then the cross stayed properly attached to the wheels and Thymio moved straight forward.
Setting in motion
Next, we looked for a way to make the robots move forward. At first we thought to use the buttons present in each robot. But we realised that it would be necessary to press the buttons of the two Thymios at exactly the same moment, otherwise the structure was at risk of not holding together, the front robot pulling on the rear one or the rear robot holding back the front one. The robots have two different groups of buttons and they are not synchronised with one another. So we looked for another way of getting our quadruped going.
We turned to using sound to get things going. First we were hoping that the front robot would make sound as soon as it started moving and that the rear robot would react to this sound by "following" the front robot, that is by starting its wheels and feet moving. But we could not manage to make the front robot produce a continuous sound.
Then we wanted to try to do something with the front distance sensors of Thymio. For example, when the field is free, the robot moves forward and makes a sound and the rear robot follows it by reacting to the sound. That would also have allowed our quadruped to stop before walls or other obstacles in its way.
We tried to follow this track, but it was hardly conclusive. Already, there was a problem to make the 2 robots start at the same time. We wanted to use sound for this, but since we were using it already when the front robot was moving, we did not manage to arrange it so that the 2 "conditions" worked. Then there was also the problem of getting Thymio to produce a continuous sound. We succeeded notably by using the front distance sensors, but the sound produced was really disagreeable. It would have needed to be able to use an SD card on which a sound or some music was loaded, insert it in the robot and cause the contents of the card to be played.
We then concentrated on how to start the Thymios. We kept the idea of using sound. We programmed the two robots so that they start when the ambient sound level exceeded a certain fixed value. This latter roughly corresponded to a handclap. To determine the front and rear of our quadruped, we simply programmed one robot to "roll" forward and the other backward.
The file containing the code can be downloaded here.
Video Demonstration
Here is a video demonstrating our robot working:
Conclusion
There are still many possible improvements which we did not have the time to study and to implement. For example, to stop our quadruped, we pressed a button on the front robot and another button on the rear robot. We could have tried to find another way of doing this, so that the two robots could both stop at the same moment. Or, use the distance sensors to stop our system in front of walls, at the edge of a table, etc …