12/04/08
|
A major refurbishing of this home page is now under way due to Christine's new appearance. We apologize for the lack of info during the last few months, but we wanted the new robot to be a surprise. We will be sorting through the footage tomorrow and start updating the tech-page.
|
15/04/07
|
The gallery section has been updated with photos from this years event and with some of our training runs of the track.
Our preparation mainly consisted of placing the usual order at Conrad and Holte hobby plus spending a good part of the Easter holidays writing software for the track. A sturdier rear wheel steering was also added, as one of the welds broke last year during training.
We were probably the team using the most time on completing the track. A full run took about 7 minutes.. but well.. we opted for accuracy and knew that we would not be able to be the fastest anyway. We were kind of puzzled about the performance on the race track where the controller seemed to act against instinct in the curve. The explanation turned out to be two errors in the line check function - something that only manifested itself, when we started pushing the limits of the equipment.
We have put a plot of our transient recorder log in the gallery section. It shows what happens when Christine is going down the stairs (a track part we loathe to test). The recorder logs in a ring buffer that contains the last 30 seconds of the run.
There were definitely more people coming to watch the event this year and the level of the robots was better. The visit from the astronauts, the minister of science, and the rector also left a nice impression. As Jonas wrote on his homepage a good part of the crowd unfortunately got bored and left the scene before the second run. Much work was put into programming of the navigation, which is not that exciting to do or watch.
Some things that could be challenging to do would be to remove the line on the seesaw or to put up a bridge that the robot has to activate in order to access a plateau with a gate. A series of buttons that light up in the order in which you must push them may also be interesting. Torsten likes the idea of a death drome where you have to build up some speed to access a gate mounted with an angle. Some years ago someone also suggested putting a gate on turn table. This remains one of my personal favourites.
It was great seeing and chatting with everybody again and finding out what people had come up with. A huge thank you goes out to the organizers, who put a lot of work into setting up the event.
|
12/04/07
|
RoboCup 2007 came to an end a few hours ago. This year Christine got 2nd place!
We will sort through the huge amount of footage we have got and update the gallery section this weekend. First though.. some sleep.
|
14/05/06
|
RoboCup 2006 was held this week and we got 3rd place!
It did not start well though. Sunday evening when we were testing the software, we discovered an unpleasant problem when the transient recorder wrote to certain addresses in the ARM7 memory. It turned out not to be due to bad clusters as we initially thought, but that the compiler caused the trouble. We installed a new one Monday - the day before the competition started..
Tuesday morning the track was set up and the testing could commence. With the near elimination in the qualification last year in clear memory, we went for a robust and well tested run this time and also managed to get a couple of hours sleep. It worked rather well. We came in second behind Team Rose with 10 points in the first run and 12 in the second run.
We then went on testing the tunnel, which we had a great deal of respect for after it caused all the trouble last year. Then - when testing the stairs - disaster struck. While driving down the stairs the car suddenly went dead, the golf arm came crashing down and smoke accompanied with an unpleasant burned smell rose from the rear of the car. Needless to say, we thought that we were finished. Everything was taken apart and examined, and 'fortunately' it turned out that the battery casing had short circuited when (we believe) the springs holding the batteries in place came in contact with a screw on the chassis. The springs had gotten a blue colour and were melted into the plastic. The solution was a lot of insulation (gaffa tape) and a new casing. We would like to thank Bertil from Automation DTU for the spare part.
This cost us 4 hours and probably the minus gate. It was a shame, because the light sensor did not give a single false detection during the runs. Finally we gave up and decided to drop it and make the rest of the run as stable as possible.
In the final we were not very lucky. In the first run the car lost the line and drove into the wall, and in the second run it missed the golf ball and changed the section too late (bad programming - we did not test it). We only got 9 points and were forced into a tie break. We then went with our stable qualification run here, and got the third place.
The pictures and videos we shot during the 3 days have been put in the gallery section.
|
13/04/06
|
Since the last entry, a distance sensor has been added to the front of the vehicle, and a new board with an Atmel Mega 8 (The Caddie) is being prepared for control of our golf club. Furthermore, the software has been revised and the new track parts are under way.
|
12/02/06
|
The code and equipment have been dusted off, and Christine has got new steering on the rear wheels. Our main problem last year was the large turning circle, but it already seems to be functioning much better. You can see the improvement in the video of the racetrack in the gallery section.
|
26/04/05
|
The videos from RoboCup 2005 have been uploaded today. Last week Ingeniøren brought an article about the competition, and Christine was mentioned as "the car built of card board boxes from Aldi".. We don't know how they got that idea - it must be our test track, they are referring to..
The RoboCup work will now be paused for a few months while we concentrate on our exams, but we will return with some major reconstruction of Christine's rear, when time permits.
|
18/04/05
|
We came in 9th in Thursday's final, where we got 9 points. The first strategy was to play it safe and take the points we knew would work (almost) every time. This was the complete opposite to the qualification, where we were nearly eliminated, as we set the car to take 14 gates instead of going for something robust.
We have discovered a lot of things that should be revised for next year, but the main issue must be the lousy turning radius, which prevented us from taking any of the race track points this year. The line sensor should also be altered, as we need more intensity for line detection on the quarry tiles.
We will be uploading pictures from the competition within the next couple of days.
|
11/04/05
|
We have finished the body work, and you can view the end result in the gallery section.
Tomorrow the track is open for testing.
|
10/04/05
|
The body work is almost complete. It was assembled yesterday from Evergreen styrene sheets and painted metallic blue today. The degreasing was carried out with cheap vodka as we didn't have any denaturated spirits. Now we only need to buy the adhesive foil for the last details.
The last programming involves the red and yellow gates, the unmarked gate, and a few adjustments to the speed controller.
|
04/04/05
|
We spent a good deal of this weekend at the entrance of building 101A at DTU placing card board and adhesive tape on the ramp and stairs - much to the surprise of people passing by. Eventually - as you can see on this picture - we ended up in the middle of a car commercial shooting for Suzuki.
Our usb-to-serial converter suddenly wouldn't load the ARM software with a baud rate above 2800, so we chose to invest in a new one today. It works perfectly at 115kbps.
|
28/03/05
|
Movies of the wall and ball game sections have been uploaded. The tunnel is also ready now. Again - the programming continues..
|
21/03/05
|
The tennis ball arm, lcd display, and the classical spring+magnet construction for the tunnel obstacle are now all mounted. We have added a short movie of the vehicle making a 3-point turn to the gallery section. The programming of the track sections continues..
|
20/02/05
|
A transient recorder has been set up and has proven very helpful in finding errors and in optimising the controller parameters. Christine is now able to communicate using an lcd display, which will be mounted in the windscreen. The arm for grabbing the tennis ball is also under construction. And then there are some minor set backs.. Unfortunately our sonar is faulty due to a loose connection somewhere in the receiver. Moreover, Nana's motherboard experienced a meltdown a few days ago.
|
17/01/05
|
The communication between the motor controller, the line sensor, and the main processor is now working.
|
04/01/05
|
Today our new homepage has been uploaded.
|
30/04/04
|
This year I decided only to go for the part without tape and the tunnel due to the physical limitations of the robot. The part without tape is not really difficult. All you have to do is go right for a while and then go left until you find the line again. But due to the large turning-circle of the car, it would come disquietingly close to the port before finding the line. A video of the port and the tunnel can be found in the gallery section.
To open the tunnel door I mounted a spring with a magnet on a curtain rail. (The function can be seen on the video). As I made the design and the preliminary tests at home I assumed that the steel door would be conductive. (it looked shining on the videos) Therefor I mounted two electrodes on the magnet to detect the door. Of course the real door turned out to be lacquered. So instead of the electrodes I used the output of the speed controller to detect the door.
The ramp has a gradient of approximately 13.6 deg. As I did the first tests, the motor controller would block one wheel making the car slide on its way down and eventually leave the ramp. The solution of that problem was to mount all the batteries and a little extra ballast on the back of the car. It didn't look pretty, but it enabled the car to stabilize it self. A car with 4WD would not have been bad for this purpose.
The night before the contest Nana and I stayed up until 2 am finishing and testing the software. As we had a stable version, which was able to complete the track with the selected obstacles more successive times, we went home to sleep and to charge the batteries. At that time there was still a lot of activity on the course.
On the great day I made two complete runs in order to test if the light or the fresh batteries would have influenced the function. That didn't seem to be the case. So we kept the software from the night before.
As earlier mentioned the robot did not make the finishing line. As it had passed half the distance to the first port a photographer took a picture with flash light directly into the distance sensor. After that the robot lowered the speed a little, which it was meant to do after the first port. At that point I knew that the race was over. After the first port it raised the speed again and took the left branch towards the wall thinking that it was on the way to the part without tape.
In Formula 1 they use to say "If You want to finish first, you have to finish first". I almost did finish first but unfortunately not at the finishing line.....
If the robot had taken the course like in the test it would have become 9 points and rank 10. But then again it wasn't the only robot that had difficulties with the light on the track.
If we can find the time this winter we will construct a new robot from scratch using the gained knowledge. The following points should at least be taken into account:
- The turning-circle should be as small as possible
- The steering servo should work as fast as possible
- The car should be a little narrower than the latter
- The line sensor should be made with modulated infra red light and have a better resolution as the latter
- The car should have a display and a keyboard for better debugging possibilities
- If a distance sensor is required for the next course an ultra sound sensor should be used. (not for detecting the ports)
|