Professor Kim Jong-hwan, president of FIRA and director of Intelligent Robot Research Center at KAIST, is not a native speaker of English, reminisced about his ten years as the father of robot soccer.
"I started the MiroSot game in 1995," he said, "and at that time it was a 3 versus 3 game. It was more than ten years ago, so at that time integration of the motor control system, computer vision technology, wireless communication, and multi-agent cooperation system was very difficult." The Professor said that the game has evolved quite a lot in the past years, so that now full integration of many complex systems allows twenty-two robots to play full 11 on 11 games.
Lateral expansion
The robot soccer tournament has also expanded far beyond that singular first idea of 3 on 3. There are now seven different categories in which organizations can compete.
The first, and oldest, is the MiroSot, which refers to micro robot soccer tournaments.
There is also RoboSot for larger robots and NaroSot for smaller sized robots.
SimuroSot is a competition for simulated software soccer games without actual robots at all. KhepraSot is a 1 on 1 match between self-contained autonomous robot models. AndroSot is a competition between robots that are controlled by people.
The HuroCup is, in theory, a competition between walking, autonomous robots that kick a soccer ball and play exactly like regular human players.
Because it is there
The professor spoke about the HuroCup by saying: "For example, humanoid robot soccer games, HiroSot, so people can easily imagine robots can easily run and jump and [hit with their] head. But the technology is still challenging, quite difficult to do such a nice performance. So still we are waiting for more technology to have a nice game." The goal, for HiroSot of course is to have an 11 robot team on each side, but for now the competition is limited to single robot feats of athletics such as a penalty kick, a robotic marathon and a weight lifting competition.
Since the HuroCup is the newest competition, Kim said that there are also some funding problems. Building a humanoid style robot can get expensive.
"Some rich labs can build these robots quite comfortably, but most labs around the world may not have enough funds to build humanoid robots," the professor explained. He further went on to say that soccercapable robots require a kind of technology that is simply beyond current commercial products. He pointed to Honda and Sony, Japanese companies who are producing bipedal robots for commercial entertainment use, and noted that while they are interesting from a commercial standpoint they cannot really do something as athletic as kick a ball.
"So I may say that we started this HuroCup," Professor Kim continued, "and as I mentioned we plan to follow the same strategy as the MiroSot." He referred to the ten year history of the small robot soccer tournament, and reminded that while in the beginning the matches of 3 on 3 robots were a striking technical achievement, now full 11 on 11 tournaments are routine.
When asked if it was better in the long run for robots to play soccer in humanoid form or a more conventional robotic form with wheels, the professor mentioned that the HuroSot was more technically challenging from a research point of view. "Even just to find the way to the ball and kick it towards the goal is quite difficult," he explained. "This is much more difficult than other types."
Familiarity breeds ease
After speaking about the difficulties of humanoid robots, the professor was asked which of the competitions was the easiest from a technical standpoint.
He responded to this by saying: "Well the simulations games are pretty easy. Also we started with MiroSot so it is the easiest." The technologies for the smaller, box robots with wheels are the most advanced, and the rules for their competition is the most mature. "For the people who have financial restrictions on developing hardware robot systems, the SimuroSot is open to everybody. The same system is in place, however the controlling strategies should be developed by participants.
So this one, somehow, is the easiest one compared to other categories, because they only have to build their own software."
Technical explanation
Kim explained the technical rules of the game in great detail.
The robots are usually controlled by a central computer which broadcasts commands to each robot wirelessly. While it is allowed in the rules for the robots to have their own sensors, normally feedback to the central computer that controls the robots is done from an overhead camera that is connected directly to the team's computer. "The information from the overhead vision camera is forwarded to the main PC, here," the professor explained while pointed to a diagram that he drew, "so while the game is played by the robots, the game is in essence a competition between this computer and this computer."
Kim was asked about the outcome of the games in the case that same robots and the same computer programs were run multiple times. "Even though the same robots and the same strategies are used, the outcome would not always be the same because of the Butterfly effect." However, he said that the computer programs and the robots were not yet able to develop their own strategies. Changing strategies would require the developers of the computer program would have to call a time out and change strategy manually.
"Maybe in twenty years we can expect those excellent learning systems, self-evolving systems," he admitted. He said that the main roles of the robots in his tournament are controlled by the PC, but the robots themselves only get direction and velocity commands from the PC, then the robots themselves carry out the goal of moving to the location that the PC commanded them. So the robots have a small bit of autonomy now, but not complete autonomy. "It depends on the architecture of organizing the system," he explained, "You can give all the intelligence into here [the PC], or some of the intelligence can be put into here [the robot]."
Waxing philosophical
Professor Kim also spoke about the future of robotics by referring to popular culture. "You might know the movie I, Robot -- NS5. This robot is quite smart, it has its own intelligence, it can make its own decisions, it has its own emotions and motivations.
Currently this robot is somehow a small mechanical or mechatronic system.
If we can put a more high-powered CPU into this small robot, then we can expect such a nice I, Robot robot. But now, no."
