czech english

Robotem rovně 2010

An outdoor competition of autonomous robots in Písek

The second year of "robotem rovně" (freely translated as go straightforward with robots) took part on May 22nd 2010 in Palackého sady, Písek. The event is organized by Radioklub Písek, aiming to increase robotics awareness in general public and to attract youth to technology. Entrants compete in two categories: robots and cars (robots without any sensors). The goal is to get as far as you can on a straight path, but it’s not nearly as easy to do, as it appears to be.

Rules

The goal of the competition: “To let the robot drive through the track without any interaction with his owner or anyone else and to stay on the pathway at all times. The road is about 3 meters wide and 300 meters long.” That’s just an excerpt from the rules, which you can find at Radioklub Pisek official website.
The robots competed in three rounds and in a given order, which has been determined by their velocity, measured at the approval (a test if the robot can pass at least a meter long track). The faster robots went first and the slower later, so they wouldn’t have to redundantly pass each other. The time difference between the starts was two minutes, so they went relatively quickly one after another.
A point was given for every meter the robot successfully drove, and the total from all the rounds was the final score.

First impressions

When I saw Radioklub Pisek had this competition last summer, I thought of it as a Robotour’s rival. But I see it way differently now – “Robotem rovně” (RR) could be an ideal preparation race for Robotour. RR gives a chance to complete beginners and mainly, lots of experience that you can use at Robotour. It was not a coincidence, that many teams from Robotour came to this event, to dust off their robots and train.
The event was very well organized and you could see that this wasn’t the first one, that Radioklub Pisek held. Even the relatively fast tempo of the competition was acceptable – observers weren’t bored, something was going on all the time, and the competitors also had a good time those three rounds.

Competition difficulty

Even though it doesn’t seem at first sight, the competition difficulty is pretty high. To be more precise, the further you want to go, the harder it is. Almost all the cars and robots make it 10 meters. The width to length ratio of 3m to 10m gives you about 17 degrees range, so it’s no real problem to point your vehicle that way.
If your ambitions are a little bit higher, you need a bit of luck, or some sensors. If you were thinking about GPS, you can give up already. There are some beautiful, huge trees in the park and even though you have signal from up to 9 satellites, the error isn’t going to be below 5 meters. I think that a demonstration sample from NMEA at Google maps (see pic.) will give you an idea. Necessary to say, the worst GPS position was on the start (then at least led the same way as the road), but you can forget about telling on which side of the road you are. After the import to OpenStreetMap, you can clearly see the uncertainty level (circles).
Probably the next sensor, that comes to your mind, is a compass. That will propel you way further, but some surprise will show up also. Firstly, you’ll most likely have just a 2D compass, because they’re cheaper and more available. Which wouldn’t matter much, but you have to know what tilting does to the sensor. The road starts sloping downwards and, even if your compass can handle this, is like any other road slightly tilted to one side because of the rain.
Aside the tilt of the compass, which can be compensated, pay attention to the cables and the sewers hidden underneath the road, they will turn your compass into a roulette. So, you can use one, but it would be best to combine with odometry and gyroscopes.
The only remaining thing from commonly used sensors, is camera. The race consisted of three rounds, and each one of them was unique. The weather at 10 a.m. was cloudy, at 11 a.m. rainy and at 12.a.m. the sun was shining sharp. How’s Radioklub Pisek doing that, so you could train in all possible weather conditions, I don’t know .
About 100 meters from the start, there’s a past for camera too. You will arrive on a small, open atrium, and now search, kiddo. There was a little unclarity in the rules, whether the robot can use the side ways: path is path, but I guess the organizers meant to stay on the main path. Well, it isn’t simple.
But a proof, that you can reach the end, brought us Roboauto, that got the maximum possible score. The road was 300m last year, but this time it was cut down to 200m because of excavation works.

Results

The table of results is taken from the article robotem-rovne-2010-uspesne-za-nami

R1 — Quido

Tomáš Ondráček/Roboauto

We took the competition in Písek as an opportunity to test our new platform: robot Quido based on an RC model wheelframe. It used to be a second car following a target on the lead “Karlík”, but we have upgraded it to work autonomously.
From cooperation with FIT VUT in Brno, our team received lidar SICK LMS 100 for the Sick Robot Day 2010 competition. We attached it to Quido not statically, but to a turning hinge, so that it could turn during the ride, and cover most of the area this way. Other than a camera we also used an odometer and a compass. We turned off the GPS for the race, because it wasn’t necessary and it would probably just complicate everything. For the same reasons, the front sonar was turned off.
The driving software was the same we used at Robotour. Lidar and camera took care of way detection, and the information was aggregated into a 2D map, in which the robot planned his way using a reactivating algorithm. No global map was necessary, considering the spirit of the competition.
To make the robot not to turn any undesired way on crossroads and large paved areas, we added an option to set a preferred stable direction in the way planner (which otherwise changes dynamically along the position in the RNDF map).
Last Friday night before the competition we arrived to the park in Písek with an intention to try the road once and to “smooth tune constants” from the recording whole night long. To our surprise Quido went forth and back the whole way without any doubts, so we could go to check the city square and some local restaurants :). On Saturday the same thing repeated, Quido had no malfunctions, so we ended up getting the maximum number of points.
The competition is, in my opinion, a good preparation for this year’s Robotour in autumn and can be a nice first public test for beginning roboticians. Radioklub Pisek did a great job in organizing this year, took care of nice weather, and a big thanks belongs to them for that.

R3 — Irena

Kamil Řezáč/Sirael

It might sound defeatistly, but I didn’t come to Písek to win. I took the competition primarily as something that will make me finish the robot in time. Which I had, with a couple of reductions, been able to do, but I didn’t have time to put together some advanced sensors or algorithms. So I decided to create a very simple algorithm – to control the driving unit (turning of the front wheels) according to the compass – basically a P regulator. To avoid problems with the absolute value of azimuth, the error caused by compass tilting and so on, I sampled and averaged the data from after the start from the compass (50 samples) and I drove by that value.
Originally I planned to not change the algorithm (which was tested one evening on the sidewalk in front of my house) at all. That didn’t go very well, the regulation loop was oscillating more and more. Despite all that, I started the first round with the original program, a poor result (25 m) was corresponding. So I reduced the “strengthening” of the regulation loop, the robot calmed down and the distance passed almost doubled (42 m). Because I saw the robot trying to turn right a little, I set it it turned slightly left the last round and the result was another increase in score (80m). The robot could go a couple meters beyond that, but his destiny has been cut off by a collision with the 80m sign…
I think this is the best (or close to it), that a “blind” robot can do, and it surpassed my expectations by far (best test try result was 42m).

R4 — Eduro Maxi

Martin Dlouhý/Eduro Team

This was the first of the outdoor competitions this year for “Maxík”. It’s still a baby, which had driven his first meter on Monday in a kitchen, then a couple more meters after closing the bottom on Wednesday, and finally on Saturday, it went on his first race.
Robot was homologed using version 0 – go straight using the odometry. Version 1 then combined compass and version 2 added data from a camera and GPS. In the end, the robot was using a customized version 2, he drove straight for 1m, used a combined data from camera and the compass, turned the right way, and went straight for another meter. No rush, but you could easily see when the sensors started to feel, that the robot is going off the road.
The “Robotem rovně” was a great competition. It provoked us to make the robot work, so it will hopefully have a chance to score some points at Field Robot Event. A week after that is RoboOrienteering, and well, Robotour 2010 in Bratislava in autumn and SICK Robot Day in Germany. Once again, huge thanks belong to the organizers for a great event.

R7 — ARbot

Aleš Ruda/ARbot

The robot went through a big changeover after last year’s Robotour. New, more accurate GPS, AHRS instead of a compass, new optics on the camera, writing parts of the SW into assembly, a new barrel holder and a lot of little things. Unfortunately, the MD23 motor units started to stiffen. Subsequently I found out, that the cause were high voltage peaks in the 5V supply branch. Two months of suffering came to an end in the beginning of May, when I upgraded to MD25. Last day before the competition, the robot started moving again, some errors in SW had been fixed and he even drove through a curvy way towards my house, so “Let’s go!” to Písek in the morning.
Little of time, no space to be a hero. The strategy was, that a simple camera will keep the robot on the track and sonar will assist with obstacles. It worked for Robotour, anyways.
Quickly test the robot in local ambience and to the start of the first round. 3, 2, 1, start. The robot started moving, but instead of going straight on the way, he “drunkishly” went from side to side, but he made it to the end. Greeeeat. What’s with the weaving, though? Oh, I forgot the camera lens cover.
I tried to make it go with the camera between rounds, but the results weren’t good. Robot went off the road very soon. I couldn’t fix the road detection algorithm. I tried to make the AHRS work, but didn’t face much success. So, if it worked once, why not twice more? Robot will go according to the sonar. Next round 90 and 184 meters. Good job. I wouldn’t believe this if someone told it to me, but it’s true. I saw it with my own two eyes.
I think that this competition was a great start of the outdoor robot contests. Thanks to organizers for the attitude.

R9 — Brimstone

Michal Eliáš and Monika Svědirohová

The “Robotem rovně” competition presented to us the first experience with outdoor navigation. The wheelcase of the robot is Monika’s Žluťásek, I was left with the algorithms and track planning. I used a convolution principle from the webcamera, according to which our robot drove, no other sensors were used. The principle is to drive through the park manually, when the camera takes pictures at a given rate, while the picture isn’t saved as a whole, just as vertical columns of information calculated from the picture. Then on the competition the robots takes pictures at the same rate, and compares them to ones from before. This method is very memory and processor demanding, it needs at least an ARM7 level processor. We have been able to go 152 meters, but testing before, we always passed the whole track. The reason for not passing the track was a vulnerability of the algorithm to spectators standing alongside the track. We came fifth in the end, which is a great success, considering the fact, that the first test of the algorithm was made in Písek.

R11 — Ambra, R13 - UTrooper

Jiří Iša/Cogito MART

Our faculty (MFF UK) has bought a new outdoor robotic wheelcase this year – Utrooper. The goal (achieved) for Písek’s “Robotem rovně” was to test it in a competition and to find some of it’s attributes and flaws. Flaws were the bigger piece, but that can be expected with a new untested wheelcase.
The other robot taking part in the competition was Ambra – a customized RC Hummer, originally built for Robotou 2009. Ambra went through a big hardware change this year. So it’s basically a new robot’s premiere. Even this robot showed a nice flaw, after two pretty successful rounds. But altogether Ambra drove pretty well, plus got the vote of the 2nd vicemiss from the spectators!
We recommend the “Robotem rovně” competition to teams, that find Robotour too difficult, but also teams, that have bigger ambition, but have a new robot. Thanks to the organizers for a well held event in a beautiful neighbourhood

Final words

We recommend the “Robotem rovně a.k.a Autíčka v parku” competition warmly to everybody. Other than a good event it is a great place to meet robotics and roboticians, and full afternoon is free, so you can chat a bit. If you missed this event, the next outdoor autonomous robots contest is RoboOrienteering in Rychnov nad Kněžnou. Finally, if you’re from Písek or its surroundings and robots are one of your interests or hobbies, don’t be shy to contact Radioklub Písek, they’ll be glad to assist you, or take you into their team.

References: