Robosezon 2013: from under the water into the sky
So, the results of the 2013 robot season:
Autonomous Submarine Robots. Singapore.Mart.2013 = 1 place; // The Singapore AUV Challenge - 2013
Telecontrolled Underwater Robots. Seattle. June 2013 = 7th place; // 12th Annual MATE International ROV Competition - 2013 Autonomous Underwater Robots. San Diego. July 2013 = 3rd place; // AUVSI Foundation and ONR's 16th International RoboSub Competition - 2013 Autonomous Flying Robots. Moscow. August 2013 = finalist diploma :) // CROC - Flying Robots - 2013
A small amount of organizational and technical information is under the cut.
The Singapore AUV Challenge was held for the first time this year ( results ). The tasks were simple, but most of the participants were not very well prepared for them. Among the favorites was the FEFU team (because in 2012 it took 5th place on Robosub, and everyone knew about it) and one of the Singapore teams (they were going to take part in Robosub in this, 2013, year and in July received a prize for Robosub 2013 for the best debut). The task can be roughly described as follows: you need to go over the black stripe on the bottom of the pool and at a certain moment emerge. That is, such a line-following robot. The rules, however, somewhat unclear, describe the whole thing. There are some gates through which to go. But if it is good to go over the strip, then you will pass through them automatically. There is also a rack on which the ball stands, and it must be kicked. Again, if you go well above the strip and at the correct height, then at the next turn you kick the ball automatically. And you need to emerge in the right place - when the black strip ends. In general, the whole competition was spinning around the black strip at the bottom :-).
In these competitions there is a limit on the cost of the device (not more than 10,000 Singapore dollars) and on its weight (not more than 20 kg). The team for Singapore produced the device from last year's Robo-sab apparatus by discarding the Doppler lag. It turned out, perhaps, a slightly ugly, but working apparatus.
As always, the test-dives of the apparatus took place at night (this time the pool was kindly provided by the Maritime University):
')
The real speed of the robot is not very high:
The results were very good.
At these competitions, the FEFU team took the 1st place. Singaporeans took second place (their robot was at a loss after dropping the ball from the stand). The Russian robot neatly completed all the tasks, thanks to a good software study of the logic of following the line and switching at the points of its break. The video is clearly seen:
Despite the fact that the competitions were quite simple, they were useful and interesting. Firstly, contacts were established with the Singapore team. People there are hospitable and open. Secondly, we managed to debug many things in the apparatus at different levels of the control system. For example, they reduced delays in the control system at the lower level and began to more accurately link the navigation information to the frames from the camera, which ultimately affected the quality of control.
In 2013, the MATE ROV International Competition tele-controlled underwater vehicles championship was held 12 times. The FEFU team was considered the favorite, because in 2012 won first place at these competitions. For technical details, refer to the technical report and the poster team.
This year the team participated in the following composition:
Angelina Borovskaya - design, engineering, assembly;
Anton Tolstonogov - lead programmer, co-pilot;
Yaroslav Volkov - leading electronics engineer;
Max Kapshiter - assembly and mechatronics;
Maxim Fursov - the designer of mechanisms;
Roman Babayev - electronics;
Vitaly Nechaev - electronics;
Vitaly Storozhenko - electronics;
Vladislav Bolotov - software development, multimedia part, first pilot.
Sergey Moon - mentor and trainer.
When the FEFU team arrived at the competition site and assembled the device, it was fully operational, only during the flight one buoyancy broke. But it was repaired in two days. She looked like new.
On the first day of competition during the 20-minute practice, the machine worked perfectly. But right at the presentation the device stopped working. Fortunately, the jury did not notice.
Upon arrival at the hotel, for a long time they found out the cause of the breakdown, it turned out that the flash drive on the on-board computer was broken. Something happened literally with a pair of bytes, and the OS did not boot on it. The team spent the rest of the night preparing a new flash drive, finished at about 9 am, and the first attempt at 11. No one slept, because the maximum was able to sleep for 2 hours, and someone didn’t sleep at all the whole night. But, nevertheless, all systems checked, everything worked before the first attempt.
However, due to overwork or excitement when connecting to a power source during deployment, the guys confused the terminals and burned the voltage converter for the router in the switching unit, the connection was lost and the team could not begin to carry out the mission. As a result, only 35 points were obtained (for passing the security control and developing the transmissometer). The next attempt was made the next day, and again all the systems were repaired for this attempt. All repaired already before 10 pm. We rehearsed the fulfillment of the mission several times in order to synchronize the work of the team once again, since there are a lot of tasks and it was also necessary to check the apparatus for workability. We finished at 12 at night and went to bed, so that we could be vigorous in the morning. Judging by the training, the team was ready to complete all the tasks and score 300 points to get around the best result that other teams showed.
In the morning everything was ready, everything worked. The attempt began at 9:00. The guys quietly turned around in the 5-minute time, but suddenly the STM-card did not start, they rebooted the device once, the second time, the third time, and only the 4th time the board earned. This has never happened before. Lost 6 minutes from mission time. We started to carry out the mission, skipping cheap tasks, stopping only on expensive (high-grade), but during the mission, STM had to be restarted again, and as a result, the guys were able to score only 130 points (7th place). Baumana scored 125 points (8th place).
The maximum result was shown by Sea-tech 4-h club, 260 points.
Jesuits (Jesuit High School) and Canadians scored approximately 225 points.
And about 4-5 teams scored about 200 points.
As a result, at the awarding the chief designer of the team, Angelina, received an individual prize (For creativity and ingenuity) for the best design of the design and the design of the poster.
The third place was taken by the Sea-tech 4-h club team, although everyone expected that they would win, but apparently they did not have enough good ones. report and poster.
The second place was taken by the Canadian team, and the Jesuits won again (Jesuit High School).
That's how the competition went. The device was too complex and not reliable enough, despite the fact that all the systems were debugged during the training and testing (during the training and the trip everything that could break, but everything was repaired), there were still a lot of thin points in electronics and software. provision of
After hard work at the competitions, a part of the team, along with the robot of 2013, Alien 3fd and the robot of 2008, Junior-1 went on a cruise to the Novosibirsk islands on the Arctic Ocean on the research vessel "Professor Khlyustin" :
In 2013, the International RoboSub Competition was held 16 times. In 2012, the FEFU team was a debut at these competitions and won the prize for the best debut (The Best New Entry) and took the 5th place.
Robosub is much more difficult than the Singapore competition for autonomous robots. Robusub is much more than that. There are restrictions on the weight and size of the device, but they are not so easy to break. There are no restrictions on the cost of the device, so the teams try to use the maximum of what they can (many teams, including FEFU teams, have Doppler lag, and this is not a cheap thing: from 0.5 to 2 million rubles). There are no restrictions on the number of participants from one team. Suffice it to say that the winners (Cornel University) have a team of 43 people, 14 of whom came to San Diego. They say that they somehow introduced this competition into the educational process and people get academic credit for participating. Interested in technical details can see the description of the missions and the technical report of the team .
Preparation for RoboSab began in March, after returning from Singapore. Slowly she went to the finish line. First debugged the logic of the mission on the model, and the recognizers on last year's images. All this happened, so to speak, in his spare time. The students had an educational process, the mentors had to work on dissertations.
By the way, about the composition of the team:
Alexander Scherbatyuk - our head (member of the Corr. Of RAS). Denis Mikhailov - "technical instructor", developed a lot of electronics on this device
Andrey Kushnerik - last year he developed the mechanics of the device.
Vlad Goy - worked on the details of the apparatus (throwers, shooters, buoyancy, changes in mechanics) and on mock-ups (this year, color-changing LED buoys were needed to work nontrivially).
Kolya Sergeenko - worked together with Vlad on mechanics and electronics, and in fact carried out the support of the device: he rewritten and reworked something, if necessary, changed batteries, etc. This person was literally “assigned to the apparatus” and was always there when the apparatus was in the water.
Yaroslav Volkov - joined us at the end of June as participated in the ROVs. He gained wisdom from Kolya.
Fedor Dubrovin - hydroacoustics and navigation. In addition, Fedor this year was engaged in almost all org. questions.
Igor Tufanov - was the captain of the team and someone like the team lead on programming. Occasionally he wrote code (a utility for visualizing logs, a task with baskets), but mostly coordinated the activities of programmers, formalized tasks, determined the direction of movement, so to speak.
Maxim Sporyshev - the main module of the video. In early July, he budded and went to Yandex for an internship.
Andrei Gatsenko - became the main video module after Max left. Also engaged in the operator's console, and assisted in setting the regulators.
Anton Tolstonogov - was engaged in finalizing the video. He wrote the function of classifying numbers in baskets.
Andrei Sakharov - the lower level of the control system. I wrote some tasks at the top level (shooting torpedoes).
Gleb Shestopalov - top level control (balls, buoys, gates).
Mark Gulyaev - the upper level of management (walking on the tracks).
They also took part in Singapore (no longer on RoboSab):
Anton Zeleev (administration, setup of onboard Linux),
Dmitry Cherkasov (mechanic, electrician). In general, by the beginning of July there was a ready-made apparatus and Andrei Sakharov and Andrei Gatsenko during the entire debugging of the lower level of control in a small pool at the IMTI.
On July 5, the robot was brought to the swimming pool on campus on Fr. Russian. They took the robot for customs procedures from there on July 15th.
The team members returned home from the island at night (they spent the night by the pool only once), but these were days devoted entirely to preparation. And then there was another day and one night in the pool of Moscow State University. Admiral G.I. Nevelsky, because the FEFU pool was closed for cleaning.
In total, there are 12 days of intensive training. It is about 150 hours of test dives in the pool (this is not much; for example, last year the ETS team spent 500 hours in the pool, they said). During this time, the robot was trained: poking at balls and buoys, passing the gate, fixing orange tracks at the bottom, throwing loads into the necessary baskets, getting torpedoes into small hexagons and going out to pinger.
True, some of the elements were crashed from time to time and the computer of vision was sometimes reloaded.
Not everyone went to the competition. Andrei Gatsenko, Mark Gulyaev and Gleb Shestopalov were not given a visa. Rather, the US Consulate General in Vladivostok said that there would be an additional check, and it was delayed for more than a month. So 7 people went to the competition - those who already had American visas before. And the three guys stayed home.
Upon arrival in San Diego, one bag with the Doppler lag (the most expensive part of the device) and the main container with electronics were missing. In San Francisco she was. So, she crossed the border and stayed somewhere in San Francisco. Appealed to the airline, to the airport, etc. In general, where the bag - is unknown. Appealed to the organizers of the competition. They turned out to be responsive - a special person several times a day contacted the airport about the missing bag.
Robosub competitions are traditionally held in a huge outdoor pool. This is a special oval acoustic test pool - it can be found on the San Diego map at the Navy base at Point Loma. The pool is huge and there are symmetrical layouts on both sides. The first three days are held in the format of training swims. Upon arrival at the pool, the team can enroll in a queue on either of the two sides of the pool. When the team's turn comes, their apparatus is immersed in water, and they are free to do anything, including walking on the remote control. If they pass through the gate, the very first model, then they are considered qualified. Usually, teams are qualified and after that they collect data on the approximate layout of the models, shoot a lot of videos in order to get rid of the individual elements of the mission on the sign boards. Run lasts 20 minutes. All this time, the team helps military diver. He indicates the location of the device, and can do anything with it at the request of the command: turn it off, free it from the layout, drag it to another place in the pool, struggling with working engines, and direct it to the desired object. Divers are also looking for and pick up weights and torpedoes, dropped by the device. When the trial swim is over, the team can again enter the test swim queue. Since there are a lot of teams, the one who comes to the pool at 5-6 am gets the most practice. They manage to make 3-4 swims per day. Arriving at 8, usually have time to make 2 swims. After three days in this mode, then there are two days of semi-finals (each team has two attempts) and the last day is the final (teams with good results in the semi-finals participate in it).
All this was not about the FEFU team this year. The guys assembled the frame of the apparatus, screwed the engines and just sat under the awning by the pool, and waited for the missing bag with electronics. They ate the burritos, played the ChGK, in general they were bored ...
The bag was found unexpectedly on the third day. She was really lost by some of the airport services in San Francisco. The bag was immediately taken to the pool, the guys assembled the device on site, and managed to spend one practice swim on the third day. Then there was the semi-final, which for the first time failed because of poorly tuned recognizers. In the second semi-final, they did intelligible things and reached the final.
In the final, the guys really wanted to shoot torpedoes, but somewhere there was a malfunction in the recognition algorithm, and in the course of the mission the unit lost the board on which it was necessary to shoot. In the end, it still turned out to be a good launch and as a result - the third place.
To participate in the competition for CROC in FEFU, the following team organized itself:
Boris Notkin - the main ideologist and generator of ideas, inertial navigation system, stabilization algorithms, regulators, trajectory planning, director of video presentation and installation of video clips;
Dmitry Egorov - the embodiment of crazy ideas of the main ideologist in hardware (electronics, programming, transfer of algorithms from MATLAB to C ++), work with kilometer wires; Andrey Vara - design, 3D model making and embodiment in hardware, video signal processing, frame polygons ;
Maxim Nevmerzhitsky - a mathematical model of the movement of a helicopter, help in 3D modeling, construction, tools, frame models of polygons, photo and video operator;
Kostya Borisov - videographer, stage director, humorist and inspirer;
Julia Genze - stabilization algorithms, frame models of polygons.
The decision to take part in the competition was finally made somewhere by the beginning of January, although the guys had to go back in the fall. It was originally acquired by AR.Drone 1.0, but it quickly became clear that its capabilities for carrying out missions would not be enough. It was decided to build a drone from scratch.
The general idea was as follows:
We decided to make a “sandwich” of two quad-core single-board computers on Android (One single-board device was fully planned to be loaded by video signal processing, polygon recognition and position correction from camera data. The second single-board device had to process data continuously coming from an inertial stabilization system, run a pilot program and provide maximum possible accelerations on the desired trajectory), Multi Wii (it was planned to use to collect data from telemetry systems and transfer information to single boards iki) and four knobs (had to provide feedback and control motors).
In March, based on the intention, electronics was ordered from online stores, which arrived only at the end of the first week of August (ie, 3 weeks before the final).
But even without the necessary electronics, we were able to complete all the qualifying tasks and go through 28 finalist teams.
As a result, the final performance on August 25 is described by the guys as follows:
“The team got together early in the morning. The latest news was disappointing: to launch a pilot program in the copter, both monolabels were needed — the hemispheres of our drone. But alas. The cable connecting them was lost during the flight from Vladivostok to Moscow. Wireless connection, already as 7 The clock on which Dima fought was not created. Both companies worked on the curves of Chinese drivers, and could not work stably together for 2 minutes. Combining them through a Wi-Fi router was not successful - they both also conflicted with him. But they came is fresh Heads started. Strengthened troubleshooting started 4 hours before the start of the 2nd day of the competition. The solution was found quickly enough. We tried to create an access point to the local Wi-Fi network for the hemispheres via a smartphone. So simple! And it worked ... After 7 hours of unsuccessful loss of time. After 20 minutes, Dima connected all the hardware to the single-sided ones and synchronized their work. But unfortunately, we have not added, tested or debugged a pilot program :( "
The guys until the last minute before the final performance tried to finish the algorithms, debug the work of the programs, but the miracle did not happen. As a result, in order not to break the robot finally, it was decided to buzz propellers and disperse ...
As the results of the competition of flying robots:
received a diploma finalist;
experience gained in performing competitions in flying robots;
There is a ready-made iron and algorithmic basis for participation in the next competition in a year.
Finally, I want to talk about the common difficulties encountered in robotic projects.
Everyone knows that Vladivostok (marked A on the map) is far, but this is the city of Nashensky (V.I. Lenin).
The closest point to which the guys flew was Singapore (about 5500 km - indicated in the picture by the letter B). The distance that had to be covered to Seattle (indicated on the map by the letter C) is about 12,400 km. Slightly less had to get to San Diego (marked on the map by the letter D) - about 11,200 km. Well, to Moscow (letter E) only about 6400 km by plane.
In addition to the distance, there is another very unpleasant thing called jetlag (jet lag):
Vlidivostok - UTC +11;
Singapore - UTC +8;
Seattle - UTC -7;
San Diego - UTC -7;
Moscow - UTC +4.
If there is almost no difference at 3 o'clock when flying to Singapore, then a 7–8 o'clock time difference and a short adaptation period before competitions in the United States and Moscow have a serious impact on our well-being, and ultimately on the results of our performances.
Difficulties with logistics result from large distances: starting from ordering spare parts and ending with organizing a flight across the ocean with customs. , , 7 ( , ).
, , .
94 223 .
. , - 1.5 . …
: (, , ), ( — , ), , , , .
10000 USD. 100000 USD. — 2000 USD. . , - " " , .
. 100-500 .
: , .
, 7-12 , . . , , , 4 .
, .
Autonomous Submarine Robots. Singapore.Mart.2013 = 1 place; // The Singapore AUV Challenge - 2013
Telecontrolled Underwater Robots. Seattle. June 2013 = 7th place; // 12th Annual MATE International ROV Competition - 2013 Autonomous Underwater Robots. San Diego. July 2013 = 3rd place; // AUVSI Foundation and ONR's 16th International RoboSub Competition - 2013 Autonomous Flying Robots. Moscow. August 2013 = finalist diploma :) // CROC - Flying Robots - 2013
A small amount of organizational and technical information is under the cut.
Singapore
The Singapore AUV Challenge was held for the first time this year ( results ). The tasks were simple, but most of the participants were not very well prepared for them. Among the favorites was the FEFU team (because in 2012 it took 5th place on Robosub, and everyone knew about it) and one of the Singapore teams (they were going to take part in Robosub in this, 2013, year and in July received a prize for Robosub 2013 for the best debut). The task can be roughly described as follows: you need to go over the black stripe on the bottom of the pool and at a certain moment emerge. That is, such a line-following robot. The rules, however, somewhat unclear, describe the whole thing. There are some gates through which to go. But if it is good to go over the strip, then you will pass through them automatically. There is also a rack on which the ball stands, and it must be kicked. Again, if you go well above the strip and at the correct height, then at the next turn you kick the ball automatically. And you need to emerge in the right place - when the black strip ends. In general, the whole competition was spinning around the black strip at the bottom :-).
In these competitions there is a limit on the cost of the device (not more than 10,000 Singapore dollars) and on its weight (not more than 20 kg). The team for Singapore produced the device from last year's Robo-sab apparatus by discarding the Doppler lag. It turned out, perhaps, a slightly ugly, but working apparatus.
As always, the test-dives of the apparatus took place at night (this time the pool was kindly provided by the Maritime University):
')
The real speed of the robot is not very high:
The results were very good.
At these competitions, the FEFU team took the 1st place. Singaporeans took second place (their robot was at a loss after dropping the ball from the stand). The Russian robot neatly completed all the tasks, thanks to a good software study of the logic of following the line and switching at the points of its break. The video is clearly seen:Despite the fact that the competitions were quite simple, they were useful and interesting. Firstly, contacts were established with the Singapore team. People there are hospitable and open. Secondly, we managed to debug many things in the apparatus at different levels of the control system. For example, they reduced delays in the control system at the lower level and began to more accurately link the navigation information to the frames from the camera, which ultimately affected the quality of control.
Seattle
In 2013, the MATE ROV International Competition tele-controlled underwater vehicles championship was held 12 times. The FEFU team was considered the favorite, because in 2012 won first place at these competitions. For technical details, refer to the technical report and the poster team.
This year the team participated in the following composition:
Angelina Borovskaya - design, engineering, assembly;
Anton Tolstonogov - lead programmer, co-pilot;
Yaroslav Volkov - leading electronics engineer;
Max Kapshiter - assembly and mechatronics;
Maxim Fursov - the designer of mechanisms;
Roman Babayev - electronics;
Vitaly Nechaev - electronics;
Vitaly Storozhenko - electronics;
Vladislav Bolotov - software development, multimedia part, first pilot.
Sergey Moon - mentor and trainer.
When the FEFU team arrived at the competition site and assembled the device, it was fully operational, only during the flight one buoyancy broke. But it was repaired in two days. She looked like new.
On the first day of competition during the 20-minute practice, the machine worked perfectly. But right at the presentation the device stopped working. Fortunately, the jury did not notice.
Upon arrival at the hotel, for a long time they found out the cause of the breakdown, it turned out that the flash drive on the on-board computer was broken. Something happened literally with a pair of bytes, and the OS did not boot on it. The team spent the rest of the night preparing a new flash drive, finished at about 9 am, and the first attempt at 11. No one slept, because the maximum was able to sleep for 2 hours, and someone didn’t sleep at all the whole night. But, nevertheless, all systems checked, everything worked before the first attempt.
However, due to overwork or excitement when connecting to a power source during deployment, the guys confused the terminals and burned the voltage converter for the router in the switching unit, the connection was lost and the team could not begin to carry out the mission. As a result, only 35 points were obtained (for passing the security control and developing the transmissometer). The next attempt was made the next day, and again all the systems were repaired for this attempt. All repaired already before 10 pm. We rehearsed the fulfillment of the mission several times in order to synchronize the work of the team once again, since there are a lot of tasks and it was also necessary to check the apparatus for workability. We finished at 12 at night and went to bed, so that we could be vigorous in the morning. Judging by the training, the team was ready to complete all the tasks and score 300 points to get around the best result that other teams showed.
In the morning everything was ready, everything worked. The attempt began at 9:00. The guys quietly turned around in the 5-minute time, but suddenly the STM-card did not start, they rebooted the device once, the second time, the third time, and only the 4th time the board earned. This has never happened before. Lost 6 minutes from mission time. We started to carry out the mission, skipping cheap tasks, stopping only on expensive (high-grade), but during the mission, STM had to be restarted again, and as a result, the guys were able to score only 130 points (7th place). Baumana scored 125 points (8th place).
The maximum result was shown by Sea-tech 4-h club, 260 points.
Jesuits (Jesuit High School) and Canadians scored approximately 225 points.
And about 4-5 teams scored about 200 points.
As a result, at the awarding the chief designer of the team, Angelina, received an individual prize (For creativity and ingenuity) for the best design of the design and the design of the poster.
The third place was taken by the Sea-tech 4-h club team, although everyone expected that they would win, but apparently they did not have enough good ones. report and poster.
The second place was taken by the Canadian team, and the Jesuits won again (Jesuit High School).
That's how the competition went. The device was too complex and not reliable enough, despite the fact that all the systems were debugged during the training and testing (during the training and the trip everything that could break, but everything was repaired), there were still a lot of thin points in electronics and software. provision of
After hard work at the competitions, a part of the team, along with the robot of 2013, Alien 3fd and the robot of 2008, Junior-1 went on a cruise to the Novosibirsk islands on the Arctic Ocean on the research vessel "Professor Khlyustin" :
San diego
In 2013, the International RoboSub Competition was held 16 times. In 2012, the FEFU team was a debut at these competitions and won the prize for the best debut (The Best New Entry) and took the 5th place.
Robosub is much more difficult than the Singapore competition for autonomous robots. Robusub is much more than that. There are restrictions on the weight and size of the device, but they are not so easy to break. There are no restrictions on the cost of the device, so the teams try to use the maximum of what they can (many teams, including FEFU teams, have Doppler lag, and this is not a cheap thing: from 0.5 to 2 million rubles). There are no restrictions on the number of participants from one team. Suffice it to say that the winners (Cornel University) have a team of 43 people, 14 of whom came to San Diego. They say that they somehow introduced this competition into the educational process and people get academic credit for participating. Interested in technical details can see the description of the missions and the technical report of the team .
Preparation for RoboSab began in March, after returning from Singapore. Slowly she went to the finish line. First debugged the logic of the mission on the model, and the recognizers on last year's images. All this happened, so to speak, in his spare time. The students had an educational process, the mentors had to work on dissertations.
By the way, about the composition of the team:
Alexander Scherbatyuk - our head (member of the Corr. Of RAS). Denis Mikhailov - "technical instructor", developed a lot of electronics on this device
Andrey Kushnerik - last year he developed the mechanics of the device.
Vlad Goy - worked on the details of the apparatus (throwers, shooters, buoyancy, changes in mechanics) and on mock-ups (this year, color-changing LED buoys were needed to work nontrivially).
Kolya Sergeenko - worked together with Vlad on mechanics and electronics, and in fact carried out the support of the device: he rewritten and reworked something, if necessary, changed batteries, etc. This person was literally “assigned to the apparatus” and was always there when the apparatus was in the water.
Yaroslav Volkov - joined us at the end of June as participated in the ROVs. He gained wisdom from Kolya.
Fedor Dubrovin - hydroacoustics and navigation. In addition, Fedor this year was engaged in almost all org. questions.
Igor Tufanov - was the captain of the team and someone like the team lead on programming. Occasionally he wrote code (a utility for visualizing logs, a task with baskets), but mostly coordinated the activities of programmers, formalized tasks, determined the direction of movement, so to speak.
Maxim Sporyshev - the main module of the video. In early July, he budded and went to Yandex for an internship.
Andrei Gatsenko - became the main video module after Max left. Also engaged in the operator's console, and assisted in setting the regulators.
Anton Tolstonogov - was engaged in finalizing the video. He wrote the function of classifying numbers in baskets.
Andrei Sakharov - the lower level of the control system. I wrote some tasks at the top level (shooting torpedoes).
Gleb Shestopalov - top level control (balls, buoys, gates).
Mark Gulyaev - the upper level of management (walking on the tracks).
They also took part in Singapore (no longer on RoboSab):
Anton Zeleev (administration, setup of onboard Linux),
Dmitry Cherkasov (mechanic, electrician). In general, by the beginning of July there was a ready-made apparatus and Andrei Sakharov and Andrei Gatsenko during the entire debugging of the lower level of control in a small pool at the IMTI.
On July 5, the robot was brought to the swimming pool on campus on Fr. Russian. They took the robot for customs procedures from there on July 15th.The team members returned home from the island at night (they spent the night by the pool only once), but these were days devoted entirely to preparation. And then there was another day and one night in the pool of Moscow State University. Admiral G.I. Nevelsky, because the FEFU pool was closed for cleaning.
In total, there are 12 days of intensive training. It is about 150 hours of test dives in the pool (this is not much; for example, last year the ETS team spent 500 hours in the pool, they said). During this time, the robot was trained: poking at balls and buoys, passing the gate, fixing orange tracks at the bottom, throwing loads into the necessary baskets, getting torpedoes into small hexagons and going out to pinger.
True, some of the elements were crashed from time to time and the computer of vision was sometimes reloaded.
Not everyone went to the competition. Andrei Gatsenko, Mark Gulyaev and Gleb Shestopalov were not given a visa. Rather, the US Consulate General in Vladivostok said that there would be an additional check, and it was delayed for more than a month. So 7 people went to the competition - those who already had American visas before. And the three guys stayed home.
Upon arrival in San Diego, one bag with the Doppler lag (the most expensive part of the device) and the main container with electronics were missing. In San Francisco she was. So, she crossed the border and stayed somewhere in San Francisco. Appealed to the airline, to the airport, etc. In general, where the bag - is unknown. Appealed to the organizers of the competition. They turned out to be responsive - a special person several times a day contacted the airport about the missing bag.
Robosub competitions are traditionally held in a huge outdoor pool. This is a special oval acoustic test pool - it can be found on the San Diego map at the Navy base at Point Loma. The pool is huge and there are symmetrical layouts on both sides. The first three days are held in the format of training swims. Upon arrival at the pool, the team can enroll in a queue on either of the two sides of the pool. When the team's turn comes, their apparatus is immersed in water, and they are free to do anything, including walking on the remote control. If they pass through the gate, the very first model, then they are considered qualified. Usually, teams are qualified and after that they collect data on the approximate layout of the models, shoot a lot of videos in order to get rid of the individual elements of the mission on the sign boards. Run lasts 20 minutes. All this time, the team helps military diver. He indicates the location of the device, and can do anything with it at the request of the command: turn it off, free it from the layout, drag it to another place in the pool, struggling with working engines, and direct it to the desired object. Divers are also looking for and pick up weights and torpedoes, dropped by the device. When the trial swim is over, the team can again enter the test swim queue. Since there are a lot of teams, the one who comes to the pool at 5-6 am gets the most practice. They manage to make 3-4 swims per day. Arriving at 8, usually have time to make 2 swims. After three days in this mode, then there are two days of semi-finals (each team has two attempts) and the last day is the final (teams with good results in the semi-finals participate in it).
All this was not about the FEFU team this year. The guys assembled the frame of the apparatus, screwed the engines and just sat under the awning by the pool, and waited for the missing bag with electronics. They ate the burritos, played the ChGK, in general they were bored ...
The bag was found unexpectedly on the third day. She was really lost by some of the airport services in San Francisco. The bag was immediately taken to the pool, the guys assembled the device on site, and managed to spend one practice swim on the third day. Then there was the semi-final, which for the first time failed because of poorly tuned recognizers. In the second semi-final, they did intelligible things and reached the final.
In the final, the guys really wanted to shoot torpedoes, but somewhere there was a malfunction in the recognition algorithm, and in the course of the mission the unit lost the board on which it was necessary to shoot. In the end, it still turned out to be a good launch and as a result - the third place.
Moscow
To participate in the competition for CROC in FEFU, the following team organized itself:
Boris Notkin - the main ideologist and generator of ideas, inertial navigation system, stabilization algorithms, regulators, trajectory planning, director of video presentation and installation of video clips;
Dmitry Egorov - the embodiment of crazy ideas of the main ideologist in hardware (electronics, programming, transfer of algorithms from MATLAB to C ++), work with kilometer wires; Andrey Vara - design, 3D model making and embodiment in hardware, video signal processing, frame polygons ;
Maxim Nevmerzhitsky - a mathematical model of the movement of a helicopter, help in 3D modeling, construction, tools, frame models of polygons, photo and video operator;
Kostya Borisov - videographer, stage director, humorist and inspirer;
Julia Genze - stabilization algorithms, frame models of polygons.
The decision to take part in the competition was finally made somewhere by the beginning of January, although the guys had to go back in the fall. It was originally acquired by AR.Drone 1.0, but it quickly became clear that its capabilities for carrying out missions would not be enough. It was decided to build a drone from scratch.
The general idea was as follows:
We decided to make a “sandwich” of two quad-core single-board computers on Android (One single-board device was fully planned to be loaded by video signal processing, polygon recognition and position correction from camera data. The second single-board device had to process data continuously coming from an inertial stabilization system, run a pilot program and provide maximum possible accelerations on the desired trajectory), Multi Wii (it was planned to use to collect data from telemetry systems and transfer information to single boards iki) and four knobs (had to provide feedback and control motors).
In March, based on the intention, electronics was ordered from online stores, which arrived only at the end of the first week of August (ie, 3 weeks before the final).
But even without the necessary electronics, we were able to complete all the qualifying tasks and go through 28 finalist teams.
As a result, the final performance on August 25 is described by the guys as follows:
“The team got together early in the morning. The latest news was disappointing: to launch a pilot program in the copter, both monolabels were needed — the hemispheres of our drone. But alas. The cable connecting them was lost during the flight from Vladivostok to Moscow. Wireless connection, already as 7 The clock on which Dima fought was not created. Both companies worked on the curves of Chinese drivers, and could not work stably together for 2 minutes. Combining them through a Wi-Fi router was not successful - they both also conflicted with him. But they came is fresh Heads started. Strengthened troubleshooting started 4 hours before the start of the 2nd day of the competition. The solution was found quickly enough. We tried to create an access point to the local Wi-Fi network for the hemispheres via a smartphone. So simple! And it worked ... After 7 hours of unsuccessful loss of time. After 20 minutes, Dima connected all the hardware to the single-sided ones and synchronized their work. But unfortunately, we have not added, tested or debugged a pilot program :( "
The guys until the last minute before the final performance tried to finish the algorithms, debug the work of the programs, but the miracle did not happen. As a result, in order not to break the robot finally, it was decided to buzz propellers and disperse ...
As the results of the competition of flying robots:
received a diploma finalist;
experience gained in performing competitions in flying robots;
There is a ready-made iron and algorithmic basis for participation in the next competition in a year.
Afterword
Finally, I want to talk about the common difficulties encountered in robotic projects.
1. Distances
Everyone knows that Vladivostok (marked A on the map) is far, but this is the city of Nashensky (V.I. Lenin).
The closest point to which the guys flew was Singapore (about 5500 km - indicated in the picture by the letter B). The distance that had to be covered to Seattle (indicated on the map by the letter C) is about 12,400 km. Slightly less had to get to San Diego (marked on the map by the letter D) - about 11,200 km. Well, to Moscow (letter E) only about 6400 km by plane.
In addition to the distance, there is another very unpleasant thing called jetlag (jet lag):
Vlidivostok - UTC +11;
Singapore - UTC +8;
Seattle - UTC -7;
San Diego - UTC -7;
Moscow - UTC +4.
If there is almost no difference at 3 o'clock when flying to Singapore, then a 7–8 o'clock time difference and a short adaptation period before competitions in the United States and Moscow have a serious impact on our well-being, and ultimately on the results of our performances.
2. Logistics
Difficulties with logistics result from large distances: starting from ordering spare parts and ending with organizing a flight across the ocean with customs. , , 7 ( , ).
, , .
94 223 .
. , - 1.5 . …
3.
: (, , ), ( — , ), , , , .
10000 USD. 100000 USD. — 2000 USD. . , - " " , .
. 100-500 .
4.
: , .
, 7-12 , . . , , , 4 .
, .
Source: https://habr.com/ru/post/192326/
All Articles