RoboCup 2000 Diary

Prologue:

The following team members set out on Wednesday morning for Melbourne, and meet up with D'Andrea Wednesday night in Los Angeles:

Bryan Audiffred
Michael Babish
Josh Pollak
Saeed Saeed
Nicole Schlegel
Mark Schwager
Will Stokes
Tobias Welge-Luessen

The Cornell team safely arrives in Melbourne on Friday, the only casualty a broken toolbox with some non-critical components missing. Immediately after check-in, the computers and the main vision camera are checked for damage in one of the hotel rooms. The system is put through some preliminary tests, with no obvious signs of damage; there is substantial relief in the RoboCup camp. The team stays up as long as possible to get in sync with Melbourne time, with some help from The Matrix and other DVDs brought along for the long flight.

Setup begins officially on Saturday at noon. The venue this year is the Melbourne Exhibition Center, located in the heart of the city, and alongsize the Crown Casino and the Yarra river. Melbourne is a very clean and modern city; it is not quite like an American city, and is in fact very similar in character to Toronto, especially with the street cars in the downtown core. Almost being hit by a car traveling on the "wrong side of the road" quickly reminds us, however, that we are not in North America anymore.

Sunday finds the Cornell team well rested and ready to put the system through its paces. The first setback is encountered by the Cornell team: the venue lighting is not to specifications (the light intensity is almost two times over the max spec), which adverseley affects the team's new infra-red communication system. This new system would have allowed Big Red to reduce system latency by approximately 20 percent (useful for defending against hard shots on net, and reaction time in general), and more importantly, would have allowed the team to not use the wireless communication system used by many of the other teams. Big Red could, in fact, request a change in the lighting conditions, but this request could adverseley affect the other teams that have already calibrated their vision systems with the more poweful lights.

We are seeded 1 st overall. While this may seem like an advantage, it turns out that we have a tough draw:

Cornell 1st
France 8th
Melbourne, Australia 9th
New Zealand 16th

The French team participated in the 97 and 98 World Cups, but did not compete in 99. The team from Melbourne is new, but is well balanced, and shares something in common with Cornell: they have also implemented a dribbling mechanism! Unlike the Cornell team, however, only one of their robots possesses this new feature, and none of their robots have kicking mechanisms.

As the defending champions, we attract sizeable crowds during warmup. We do not disappoint, with effective displays of our three new robot features: omni-directional drive, dribbling, and a goalkeeper ball trapping mechanism. This is clearly not what people were expecting: last year's team was strong, quick, and played a hard, fast game. It was clearly the expectation of the competition and crowds that we would build upon that strength to retain our title. This year's team, however, has traded strength and quickness for maneuvaribility and control, and the Cornell game has progressed from one of pure athleticism to one of finesse and grace. In other words, in 99 we played like England; this year, we play like Brazil.

There is immediate controversy, instigated by the French and German teams, over our dribbling mechanism. It is their premise that we are violating the rule by holding the ball. After a demonstration from the Australians and Cornell of our dribbler, however, it only takes the rules committee 10 minutes to decide that not only is the dribbler allowed (to be specific, it does not "remove all of the ball's degrees of freedom" and "prevent other robots from stripping the ball"), we are praised for the new innovation and the new dimension that it brings to the game.

Another controversy arises over our goalkeeper's ball trapping mechanism. The French and Germans feel that it is illegal, since it occludes the ball from view. They base this complaint on a rather ackward interpretation of the rules (I will not go into details, but I now appreciate why there are so many lawyers in the world). The rule committee again decides in our favour, for two main reasons: common sense interpretation of the rules, and more convincingly, precedent in 1998 Paris: the English team fielded a golie with a trapping mechanism that year.

After a brief discussion with the French team, we decide that we will disable our keeper's trapping mechanism when we play them in the tournament opener on Monday; if their system loses sight of the ball for periods longer than several seconds, their robots begin to chase phantom balls on AND OFF the playing field, with obvious catastrophic results. It is clearly not to our advantage to make this voluntary concession (uncontrolled behavior inevitably results in red cards and expulsion); we none the less make this decision for the sake of a good tournament opener and fair play.

As Sunday night approaches, we begin to notice (along with other teams), that there is a serious problem with one of the frequencies alotted for wireless communication: there seems to be large amounts of noise in the 433 MHz band, which results in intermittent losses of data. We only notice this problem late in the evening, as we have been testing our system with 418 MHz (the other commonly used frequency). Unfortunately, we are scheduled to use 433 MHz for the tournament opener against the French at 11 AM on Monday...

Monday August 28th

10:00 AM. We start setting up for the tournament opener in front of a crowd of spectators and some internet film crews. We quickly discover that the noisy 433 MHz frequency is posing a serious problem to robot control; every minute or so, we lose the ability to control our robots for periods of up to 10 seconds. This is by far the worst interference encountered to date, and it is supected that some consumer electronic devices must be causing this inteference. The fact that it is a work day, near noon, does not help the matter. The organizers make an attempt to locate the source of the interference, with no success. We discuss our predicament with the referee and the French captain, and it is agreed that we will swap frequencies at half-time in the interest of a fair game.

11:00 AM. FIRST HALF. The game finally begins. The French play an extremely aggressive game; their robots are quite fast, but somewhat out of control. There is repeated contact with our robots, and the French are given a yellow card. Unfortunately, we are completely thrown off our game by the radio interference and the loss of control; we cannot execute our coordinated plays, and the half finishes 0-0.

11:20 AM. HALF TIME. After replacing the batteries, we begin replacing the wireless modules with the 418 MHz receivers. The French team, however, states that it will take them over 45 minutes to change their modules. We are somewhat taken aback, as we had agreed upon before the game to change our modules at half time. The rules committee, however, overrules the referee and states that there is nothing in the rules that states that teams must swap frequencies. Given the long time it would take for the French to swap their modules, we are told that we must finish the game at 433 MHz. We a re not very happy with this decision; it is clear that the fair course of action is a swap of frequencies, but due to time constraints and the inability of the French team to change their modules, we must continue to play a sub-standard game. The obvious question arises in the Cornell camp: should we enable our keeper's ball trapping mechanism? It is quickly decided that we will not go back on our word, irrespective of recent events; we will finish the game with the mechanism disabled.

11:30. SECOND HALF. Not surprisingly, the second half starts off in a similar manner to the first half. The French robots take advantage of our lapses in control, and push the ball into our end. In the course of 5 minutes, the French are awarded 4 penalty shots; this is due to our errant defender wandering into our defense zone while out of control, an infringement that carries the ultimate penalty of a penalty shot when it occurs during an offensive opportunity. Miraculously, the French miss all of their chances; our goalkeeper remains in control long enough to make the saves. It is only a matter of time, however, before the intermittent loss of control takes its toll; the French team "scores" goal with approximately 3 minutes remaining in the game (we score the goals ourselves, in fact, when our defender randomly decides to move backwards in the goal area and push the ball into the net). With time running out, however, Cornell gains the equalizer. We settle for the tie.

FINAL SCORE: France 1 - Cornell 1.

6:00 PM. EMERGENCY TEAM LEADER MEETING. An emergency team leader meeting is held to address the interference issue. The Cornell team proposes an addition to the current rules and regulations, effective immediately: a team has the right to request a frequency change at half time. The motion passes by a large majority. If a team cannot change their frequency modules within a 20 minute time window at half time, they must give the other team the choice of freqency for the entire game. The majority feel that it is the only fair course of action. The new resolution is immediately put into effect; the teams playing at 7:00 PM switch their frequency modules at half time, a procedure that in fact does not add any extra time to the break. The Cornell team continues to make changes to their system based on observations from the first game, and awaits the 8:00 PM kickoff against Melbourne. The Australian team will be using a custom spread-spectrum wireless communication system, allowing the Cornell team to use 418 MHz for the duration of the game.

8:00 PM. FIRST HALF. The whistle blows to begin one of the best matches in Robot Soccer history; the Cornell team can finally take advantage of the radical new design and associated control strategies, and quickly scores its first of four goals. The Australian team is strong, with a solid defence, but Cornell's passing and dribbling game cannot be stopped. The first half finishes 4-0.

8:30 PM. SECOND HALF. The passing game continues, with similar results. The Cornell team is able to score another 3 goals, and the Cornell goalie, with its trapping mechanism enabled, shuts out the opposition.

FINAL SCORE: Cornell 7 - Melbourne 0.

Tuesday August 29th.

The next match for Cornell is against the team from New Zealand. The Kiwis were beaten the previous day by the Australians 14-0 and by the French 11-0. They are relying solely on local vision, a technology which currently cannot effectively compete against the more commonly used global vision (on a separate note, however, in group C a local vision system from Australia eaked out a 2-0 victory over a team from Denmark which used global vision). The Cornell team gets ready for the 6 PM kickoff; this will be a great test for the new passing and kicking algorithms that were developed since the tournament started. Cornell is again free to use the reliable 418 MHz wireless communication system.

6:00 PM. Cornell scores the fastest goal in RoboCup history: 6 seconds. The coordination of Passing, Dribbling, and Shooting, along with the smooth transition between plays, entertains the crowds. Goals come in quick succession. In order to allow the Kiwis to better test their local vision system, and to speed up the game, the Cornell team proposes to not stop play and the timer after a goal, but rather to reposition the ball on the fly at various locations on the playing field. The New Zealanders embrace this concession which will result in fewer Cornell goals, and the game finishes in approximately 25 minutes.

FINAL SCORE: Cornell 40 - New Zealand 0

The Cornell team is guaranteed to advance to the quarter finals.

Wednesday August 30th.

A match between France and Australia decides whether Cornell finishes first or second in group A. France needs to beat Australia and score at least 36 goals to finish the group in the top spot. A tie will guarantee them a place in the quarterfinals. The Australians squander a penalty shot opportunity, and the French capitalize on a defensive lapse by the Australians to score their only goal, and the game finishes France 1 - Australia 0.

Due to the tie, Cornell finishes the Round Robin tournament seeded 4th, while the French finish seeded 5th. This results in a rematch in the quarterfinals on Friday, at 10:00 AM. We are looking forward to the opportunity to replay the French...

In the other groups, Lucky Star from Singapore (the team that Cornell beat 6-2 in the semifinals in '99) will probably finish the Round Robin seeded first, due to a 66-0 win. This large score is deceiving, however: their Danish opponents forfeited the game due to a gross system malfunction, and Lucky Star was left to play stationary objects. They immediately proceeded to score as many goals as possible to ensure the top spot in the standings. The Fu-Fighters from Germany (the RoboCup '99 runners up, defeated by Cornell 15-0 in the championship final) will most likely finish the Round Robin seeded 2nd, while another team from Singapore (the Field Rangers) will finish first in group D and capture the number 3 seed.

Friday , September 1

Much of the day is spent reviewing tapes and fine tuning the system. The robots are holding up well mechanically; in particular, we are relieved to discover that the motors and gear-boxes are still in tip-top shape, despite of the performance level that they have been pushed to in the competition. The dribbling mechanisms are also holding out well. Preventive maintenance and fine tuning the robots occupies most of MechE's time, with the exception of reinforcements to the kicking bars, in preparation for the rough and tough French team.

The review of the tapes suggests one major flaw in our system; our kicks are deemed too weak and sporadic to effectively cope with the fast teams that we will be encountering. In particular, more goals could have been scored against the French team if we had capitalized on kicking opportunities. This is puzzling to the team, as the kicks were observed to be much faster back in the Cornell lab. To remedy the problem, parts are purchased to modify the circuit boards and add 4 NiCd batteries to our robots for boosting kicking power. The first prototype with the modifications is completed Friday evening; the kick is effectively increased by a factor of two, and performs well when tested with other robots in various system level tests. The team begins to make the modifications to all of the robots immediately.

A major addition to the software code Friday includes the addition of specialized penalty shot modes. For defending penalty shots, game states are coded in that will launch the goalie to the far posts (the equivalent of guessing and diving for the ball in human soccer). For offensive penalty shot opportunities, a new algorithm is coded: the player taking the penalty shot will engage the dribbler, wait for the goalie to commit, and with cool precision turn and shoot to the open side (the dribbler makes this possible, action is a swap of frequencies, but due to time constraints and the inability of the French team to change their modules, we must continue to play a sub-standard game. The obvious question arises in the Cornell camp: should we enable our keeper's ball trapping mechanism? It is quickly decided that we will not go back on our word, irrespective of recent events; we will finish the game with the mechanism disabled.

11:30. SECOND HALF. Not surprisingly, the second half starts off in a similar manner to the first half. The French robots take advantage of our lapses in control, and push the ball into our end. In the course of 5 minutes, the French are awarded 4 penalty shots; this is due to our errant defender wandering into our defense zone while out of control, an infringement that carries the ultimate penalty of a penalty shot when it occurs during an offensive opportunity. Miraculously, the French miss all of their chances; our goalkeeper remains in control long enough to make the saves. It is only a matter of time, however, before the intermittent loss of control takes its toll; the French team "scores" goal with approximately 3 minutes remaining in the game (we score the goals ourselves, in fact, when our defender randomly decides to move backwards in the goal area and push the ball into the net). With time running out, however, Cornell gains the equalizer. We settle for the tie.

FINAL SCORE: France 1 - Cornell 1.

6:00 PM. EMERGENCY TEAM LEADER MEETING. An emergency team leader meeting is held to address the interference issue. The Cornell team proposes an addition to the current rules and regulations, effective immediately: a team has the right to request a frequency change at half time. The motion passes by a large majority. If a team cannot change their frequency modules within a 20 minute time window at half time, they must give the other team the choice of freqency for the entire game. The majority feel that it is the only fair course of action. The new resolution is immediately put into effect; the teams playing at 7:00 PM switch their frequency modules at half time, a procedure that in fact does not add any extra time to the break. The Cornell team continues to make changes to their system based on observations from the first game, and awaits the 8:00 PM kickoff against Melbourne. The Australian team will be using a custom spread-spectrum wireless communication system, allowing the Cornell team to use 418 MHz for the duration of the game.

8:00 PM. FIRST HALF. The whistle blows to begin one of the best matches in Robot Soccer history; the Cornell team can finally take advantage of the radical new design and associated control strategies, and quickly scores its first of four goals. The Australian team is strong, with a solid defence, but Cornell's passing and dribbling game cannot be stopped. The first half finishes 4-0.

8:30 PM. SECOND HALF. The passing game continues, with similar results. The Cornell team is able to score another 3 goals, and the Cornell goalie, with its trapping mechanism enabled, shuts out the opposition.

FINAL SCORE: Cornell 7 - Melbourne 0.

Tuesday August 29th.

The next match for Cornell is against the team from New Zealand. The Kiwis were beaten the previous day by the Australians 14-0 and by the French 11-0. They are relying solely on local vision, a technology which currently cannot effectively compete against the more commonly used global vision (on a separate note, however, in group C a local vision system from Australia eaked out a 2-0 victory over a team from Denmark which used global vision). The Cornell team gets ready for the 6 PM kickoff; this will be a great test for the new passing and kicking algorithms that were developed since the tournament started. Cornell is again free to use the reliable 418 MHz wireless communication system.

6:00 PM. Cornell scores the fastest goal in RoboCup history: 6 seconds. The coordination of Passing, Dribbling, and Shooting, along with the smooth transition between plays, entertains the crowds. Goals come in quick succession. In order to allow the Kiwis to better test their local vision system, and to speed up the game, the Cornell team proposes to not stop play and the timer after a goal, but rather to reposition the ball on the fly at various locations on the playing field. The New Zealanders embrace this concession which will result in fewer Cornell goals, and the game finishes in approximately 25 minutes.

FINAL SCORE: Cornell 40 - New Zealand 0

The Cornell team is guaranteed to advance to the quarter finals.

Wednesday August 30th.

A match between France and Australia decides whether Cornell finishes first or second in group A. France needs to beat Australia and score at least 36 goals to finish the group in the top spot. A tie will guarantee them a place in the quarterfinals. The Australians squander a penalty shot opportunity, and the French capitalize on a defensive lapse by the Australians to score their only goal, and the game finishes France 1 - Australia 0.

Due to the tie, Cornell finishes the Round Robin tournament seeded 4th, while the French finish seeded 5th. This results in a rematch in the quarterfinals on Friday, at 10:00 AM. We are looking forward to the opportunity to replay the French...

In the other groups, Lucky Star from Singapore (the team that Cornell beat 6-2 in the semifinals in '99) will probably finish the Round Robin seeded first, due to a 66-0 win. This large score is deceiving, however: their Danish opponents forfeited the game due to a gross system malfunction, and Lucky Star was left to play stationary objects. They immediately proceeded to score as many goals as possible to ensure the top spot in the standings. The Fu-Fighters from Germany (the RoboCup '99 runners up, defeated by Cornell 15-0 in the championship final) will most likely finish the Round Robin seeded 2nd, while another team from Singapore (the Field Rangers) will finish first in group D and capture the number 3 seed.

Friday , September 1

Much of the day is spent reviewing tapes and fine tuning the system. The robots are holding up well mechanically; in particular, we are relieved to discover that the motors and gear-boxes are still in tip-top shape, despite of the performance level that they have been pushed to in the competition. The dribbling mechanisms are also holding out well. Preventive maintenance and fine tuning the robots occupies most of MechE's time, with the exception of reinforcements to the kicking bars, in preparation for the rough and tough French team.

The review of the tapes suggests one major flaw in our system; our kicks are deemed too weak and sporadic to effectively cope with the fast teams that we will be encountering. In particular, more goals could have been scored against the French team if we had capitalized on kicking opportunities. This is puzzling to the team, as the kicks were observed to be much faster back in the Cornell lab. To remedy the problem, parts are purchased to modify the circuit boards and add 4 NiCd batteries to our robots for boosting kicking power. The first prototype with the modifications is completed Friday evening; the kick is effectively increased by a factor of two, and performs well when tested with other robots in various system level tests. The team begins to make the modifications to all of the robots immediately.

A major addition to the software code Friday includes the addition of specialized penalty shot modes. For defending penalty shots, game states are coded in that will launch the goalie to the far posts (the equivalent of guessing and diving for the ball in human soccer). For offensive penalty shot opportunities, a new algorithm is coded: the player taking the penalty shot will engage the dribbler, wait for the goalie to commit, and with cool precision turn and shoot to the open side (the dribbler makes this possible, due to the extra control). The risk with this strategy is that taking too long to shoot may result in defensive players rushing in (they are required to start 15 cm behind the penalty shot taker) and blocking the shot.

A major drawback is encountered in the wee hours of Friday night/ Saturday morning: the prototype kicking circuitry cannot be effectively duplicated on the other robots. This is eventually traced to a faulty design borrowed from the '99 team that went unnoticed until now, and that unfortunately resulted in intermittent problems and gradual degradation of kicking power. At 4 AM on Saturday morning, it is decided that all attempts to add faster kicks to the other robots will be put on hold until Saturday evening, provided that we make it to the finals; making changes that we know are not reliable to all of the robots

seems suicidal at this stage. In particular, the circuitry fails in the ON state, resulting in damage to the solenoid, battery, and freezing the robot during game play. The working prototype kicking circuitry, which we now realize may fail at any time, is kept as the spare and denoted "SuperKick" it will be our ace in the hole on Saturday in the event that we fall behind in our games.

Saturday, September 2

Our quarter final game against the French is scheduled for 11:00 AM. In the two games before ours, the FU-Fighters from Germany easily defeat Crimson from Korea (Crimson is essentially Robotis, the team that Cornell beat 2-1 in the tournament opener, and the team that was eliminated by Germany in the semi-final in '99), and the RoboRoos from Australia edge out the Field Rangers from Singapore.

CORNELL VS. FRANCE, FIRST HALF. The game begins, and it becomes immediately clear to the Cornell team that the French have not been sitting idly by in the last two days. The French defense is extremely strong, and foils the Cornell offense on numerous occasions. Cornell dominates the game play, and no real French scoring chances occur, but the score remains 0-0 after the end of the first half.

CORNELL VS. FRANCE, SECOND HALF. Some minor modifications to the system are made in preparation for the second half. "SuperKick" is put in the game. Cornell again dominates the game, and is in possession for most the game. The inevitable occurs with approximately 5 minutes to go in the game, and "SuperKick" takes advantage of an opening and scores for Cornell to take the lead. The game finishes 1-0, and Cornell advances to the semi-finals. Once again, Cornell is kept to a low score by the French due to Cornell's inability to capitalize on scoring opportunities, and the effective French strategy of jostling our robots and preventing quality passes and shots to take place.

In the other semi-final game, Lucky Star II from Singapore (the team that Cornell beat 6-2 in '99 to advance to the finals) handily beats Rogi from Spain; the long awaited rematch between the two tournament favorites is scheduled for 3:00 PM.

CORNELL VS. SINGAPORE, FIRST HALF. It is immediately clear that this will go down as the best match of Robot Soccer to date. The two teams play very differently, but are evenly matched; the Cornell team is deliberate and precise, while the team from Singapore is quick and relentless. Our scoring chances consist of passing plays, while the Singapore scoring chances come from quick reactions and numerous kicks. For both teams, the defense is more than able to reject the scoring opportunities. The half ends at 0-0.

CORNELL VS. SINGAPORE, SECOND HALF. The beginning of the second half is a repeat of the first half. Both teams set up scoring opportunities, only to be rejected by strong defense. With approximately 2 minutes to go in the game, the unthinkable happens; the Cornell goalie accidentally traps the ball outside of the goal area (we are still not sure why). The rules for this offense are strict; the Cornell team is given a yellow card, and Singapore is awarded a free kick. Free kicks occur frequently during the game, but typically the players are in position when this happens. Unfortunately for Cornell, the goalie is out of position, and no repositioning of the goalie is allowed before the kick. As a result, Singapore has a clear line to the open goal, with only 30 cm separating the ball and a Cornell defeat. The goalie is approximately 15 cm away from being in a position to intercept the ball; unfortunately, the Lucky Star kick is strong (faster than "SuperKick", at greater than 1 m/s), and it is doubtful that the goalie can make the save when the whistle is blown and Singapore takes the shot on the open net. The whistle blows. The ball is kicked by Singapore, while the Cornell goalie scrambles to intercept it. Once again, the unthinkable happens; the Cornell goalie makes contact with the ball, just enough to deflect it wide of the goal. A loud "YEAH!!!" erupts from the Cornell side, while Singapore is left to look on in disbelief. The game finishes 0-0, with the crowd on the edge of their seats.

CORNELL VS. SINGAPORE, 5 MINUTE OVERTIME. The golden goal rule applies to overtime; if a team scores in the 5 minute period, they will secure a win. Once again, however, the Cornell and Singapore defense are up to the task, and the overtime ends in a draw. The game will be decided on penalty shots.

CORNELL VS. SINGAPORE, PENALTY SHOTS. Cornell wins the toss, and requests to shoot last. The penalty shootout is fashioned after FIFA rules. The teams alternate taking 5 penalty shots (with a victor being declared when a win becomes a mathematical certainty), and in the event of a tie after 5 shots, the teams continue to alternate taking shots until a victor emerges.

SINGAPORE, FIRST SHOT. At the whistle, the Cornell goalie rushes the ball (possible only due to our omni-directional drive), but the Singapore kick is executed with surgical precision, and the ball finds the corner of the goal. SINGAPORE 1 - CORNELL 0.

CORNELL, FIRST SHOT. "SuperKick" is selected to take the Cornell shot; it's kick speed is essentially that of Singapore's, but unfortunately for Cornell, the Singapore goalie is extremely quick. If the goalie guesses the right direction, it could potentially save the shot. Instead of simply aiming the ball towards the corner and kicking at the whistle (the Singapore strategy, and the strategy of all of the other teams), the Cornell player lines up straight to the ball. The Singapore teams (the humans, that is), look confused: what is Cornell up to? The whistle blows. The Cornell robot slowly rotates around the ball, and begins to move the ball AWAY from the goalie. The shot is immediately called a miss. What happened?!? The player was trying to score on the opposite goal! In the excitement, the Cornell team does not take into account the fact that all of the penalty shots are taking place in one end of the field, and that offence must shoot in the same direction as defense.

SINGAPORE 1- CORNELL 0.

SINGAPORE, SECOND SHOT. Once again, the Singapore kick is executed perfectly. SINGAPORE 2 - CORNELL 0.

CORNELL, SECOND KICK. After fixing the "wrong direction" problem, the Cornell players is once again lined up straight towards the ball. The whistle blows. The Singapore goalie guesses to the left, while the Cornell player cooly waits for the dive, aims to the right, and shoots in the open goal. SINGAPORE 2 - CORNELL 1.

SINGAPORE, THIRD KICK. The Singapore team member responsible for setting up his robot for the penalty shot, clearly affected by our goalie's ability to rush the ball, tries too hard to line up the robot to place the ball into the open corner, with the result a miss to the right. SINGAPORE 2 - CORNELL 1.

CORNELL, THIRD KICK. Repeat of the second kick; after the dive, the Cornell kicker aims and shoots towards the open net. SINGAPORE 2 - CORNELL 2.

SINGAPORE, FOURTH KICK. Goal. SINGAPORE 3 - CORNELL 2 .

CORNELL FOURTH KICK. Goal. SINGAPORE 3 - CORNELL 3 .

SINGAPORE, FIFTH KICK. Goal.

SINGAPORE 4 - CORNELL 3 .

CORNELL FIFTH KICK. The Cornell player must sco re to tie, and keep the shootout alive. It does. SINGAPORE 4 - CORNELL 4 .

The penalty shootout now advances to immediate elimination. A goal/miss combination will end the game, and decide who will advance to the finals.

SINGAPORE, SIXTH KICK. The Cornell goalie makes a splendid save to reject Singapore!

CORNELL, SIXTH KICK. The Cornell player is aimed to the right side of the net. At the whistle, the Singapore goalie darts to the right to make the save. Once again, however, the Cornell player makes the turn and shoots into the open side. The place erupts. Cornell advances to the finals!

We cannot believe what has just happened. We all felt that our tournament was over in regulation, when Singapore had the open shot on net. In the penalty shoot out, we also felt that it was over for Cornell, and that we would watch the final as spectators, when we fell behind 2-0. Miraculously, however, we were able to hold Singapore back, and edge them out in the most exciting match of Robot Soccer to date.

In the other semi final, Germany handily beat the Australia 6-0. The final will be a rematch of '99.

Sunday, September 3.

In the wee hours of Sunday morning, modifications to all of the robots take place; a new component that will allow us to handle the extra power to the kickers is incorporated into the boards, and the system is finally ready to go at 6:00 AM. All of the robots are now able to reliably kick the ball at over 1 m/s. For passing the ball, it is found that enabling the dribbler during kicking slows the ball down enough to prevent the ball from leaving the field. The system is tested with the new capabilities, and the Cornell team makes its way back to the hotel to grab 3 hours of sleep before the final. Game time is set for noon.

CORNELL VS. GERMANY, FIRST HALF. It is immediately obvious that this is not the same team that Cornell beat last year in the final 14-0. The German team is much more accurate, and its defense is like a brick wall. The Germans still have their powerful kicking devices, d which impart velocities in excess of 3 m/s to the ball. The Cornell team makes a similar change to its defense that it made the previous year; one defensive player will track the goalie, and prevent it from scoring with a hard kick. The game is very even; the Germans play very similar to the team from Singapore, an effective strategy against the deliberate passing game of the Cornell squad. No real scoring opportunities take place, and the first half finishes 0-0.

CORNELL VS. GERMANY, SECOND HALF. Cornell starts the game with some modifications; the defender is no longer restricted to shadow the goalie and the ball. It was observed that this last minute change resulted in weaker transitions. It was also felt that the Cornell keeper could handle the fast German shots. The change proves pivotal; Cornell scores with 5 minutes to go in the game. The Germans make some modifications before resuming play. As a result, their offensive chances lead to a combination of attackers and ball position that excite a strange instability in our keeper; it rapidly begins to oscillate between two defensive plays, starts to shake, and leaves wide gaps in the near post. The Germans come close to the equalizer on several occasions, but cannot capitalize on our strange system bug. On the other end of the pitch, their recent changes also have a detrimental effect on their own system; one of their defenders makes contact with the ball in the goalie zone (the equivalent of a hand ball in the goal box in human soccer), and Cornell is awarded a penalty shot with 2 minutes to go in the game. The first reaction of the Cornell team is to be conservative, aim the robot towards the corner of the net, and take advantage of the new kicking capabilities to bury the ball in for a goal; the German goalie is slower than that of Singapore, and it is unlikely that it will be able to rush to the side to make the save, even if it guesses correctly. We decide, however, to go with style: we will use our slick penalty shot code instead, even though it is riskier, and the game is by no means over; the Germans are coming closer and closer to scoring. The Cornell penalty shot taker is lined up facing the left post. The whistle blows to start the penalty shot. The German goalie rushes to the left, as its defenders swarm towards the ball. Our player keeps its cool, turns to the right, and releases the ball just before the German defenders can block the ball; the ball finds the empty right side of the goal, and Cornell goes up 2-0, the final score in the game. Cornell is once again the RoboCup champions!

EPILOGUE

In retrospect, our game play was not optimized to win the tournament; we probably tried to play too much of a finesse game, a strategy that was impeccable against most teams, but did not hold up against the faster ones. In particular, a combination of very fast opponents and physical play greatly diminished the effectiveness of our "Brazil" like game play.

At the end of the competition, the team leaders met to discuss the tournament. By all measures, it was a great success: the games were exciting, the final was a nail biter, and more importantly, the game was taken to a new level. Cornell showed that passing and fine control was possible, and that team must adopt this strategy in order to be successful. Many teams approached us after the final and took pictures and videos of the inner workings of our robots, in particular dribbling and omni-drive. Several asked for CAD drawings of our robots, and for the algorithms used for controlling these devices in such a smooth fashion.

We once again raised the bar, as we did in '99. It was interesting to see how many of the strong teams this year played very similar to our '99 team. It will be interesting to see how many teams in '01 look like our '00 team. We will probably make all of our information available to our competitors once again this year; as long as they base their system on our previous work, we will always be one step ahead...