1
00:00:00,000 --> 00:00:00,280

2
00:00:00,280 --> 00:00:02,029
>> RADHIKA NAGPAL: My
name is Radhika Nagpal.

3
00:00:02,029 --> 00:00:03,950
I am a professor at Harvard University.

4
00:00:03,950 --> 00:00:06,370
And I do bio inspired robotics.

5
00:00:06,370 --> 00:00:09,919
I'm really interested in how groups
can work together really well.

6
00:00:09,919 --> 00:00:11,710
ERIK SCHLUNTZ: So my
name is Erik Schluntz.

7
00:00:11,710 --> 00:00:14,126
I'm studying electrical
engineering at Harvard University.

8
00:00:14,126 --> 00:00:19,890
KATE DONAHUE: I'm part of RFC Cambridge.

9
00:00:19,890 --> 00:00:21,760
It's a joint Harvard-MIT team.

10
00:00:21,760 --> 00:00:23,921
We basically build
robots to play soccer.

11
00:00:23,921 --> 00:00:26,420
RADHIKA NAGPAL: So the challenge
of the Robocup competition,

12
00:00:26,420 --> 00:00:28,190
and it's an international
competition that

13
00:00:28,190 --> 00:00:31,490
was started by people in
the U.S. and Japan actually,

14
00:00:31,490 --> 00:00:36,300
is to create a team of robots
that can play and possibly

15
00:00:36,300 --> 00:00:38,607
win against the world
champions of the World Cup.

16
00:00:38,607 --> 00:00:40,940
KATE DONAHUE: There are teams
from all around the world,

17
00:00:40,940 --> 00:00:43,440
and because we're so far spread
out, it's really hard for us

18
00:00:43,440 --> 00:00:45,320
to get to test the robots
against each other.

19
00:00:45,320 --> 00:00:48,130
So it's basically just sort
of like a soccer match.

20
00:00:48,130 --> 00:00:51,040
There'll be some round robin games
and then a final competition.

21
00:00:51,040 --> 00:00:55,160
And it's a chance to see how our robots
stack up against others, and just

22
00:00:55,160 --> 00:00:56,240
exchange ideas.

23
00:00:56,240 --> 00:01:00,010
>> RADHIKA NAGPAL: If you think about
little kids playing soccer and older

24
00:01:00,010 --> 00:01:02,600
people playing soccer, and
think about the difference,

25
00:01:02,600 --> 00:01:07,580
like all the things that a little kid
cannot do that an older person can.

26
00:01:07,580 --> 00:01:09,820
Computer science has
to fill that whole gap.

27
00:01:09,820 --> 00:01:11,910
>> ERIK SCHLUNTZ: So you
could say, I want the robot

28
00:01:11,910 --> 00:01:13,920
to go behind the ball
to set up for a shot.

29
00:01:13,920 --> 00:01:16,045
But what you really have
to tell the computer to do

30
00:01:16,045 --> 00:01:18,600
is, find the vector between
the ball and the goal,

31
00:01:18,600 --> 00:01:21,004
and go there, minus an offset.

32
00:01:21,004 --> 00:01:24,170
And you really need the computer science
people to express yourself in a way

33
00:01:24,170 --> 00:01:25,620
that the robots can understand.

34
00:01:25,620 --> 00:01:28,244
>> KATE DONAHUE: We all work on very
different parts of the robot,

35
00:01:28,244 --> 00:01:29,810
but we really need to coordinate.

36
00:01:29,810 --> 00:01:34,260
It's a big challenge, and it's great
when we all work on something together.

37
00:01:34,260 --> 00:01:36,380
So we'll build mechanical
part of it, and then

38
00:01:36,380 --> 00:01:38,982
the electrical engineers
will make the circuit boards,

39
00:01:38,982 --> 00:01:41,690
and the computer science people
will have done their simulations,

40
00:01:41,690 --> 00:01:44,390
and then try and figure out how
they can combine that and make

41
00:01:44,390 --> 00:01:46,010
the robot actually move.

42
00:01:46,010 --> 00:01:49,400
>> RADHIKA NAGPAL: I think, really,
robotics is often about iteration.

43
00:01:49,400 --> 00:01:52,670
You do take one step forward,
you take two steps back.

44
00:01:52,670 --> 00:01:55,719
You try to make one thing better,
it makes something else worse.

45
00:01:55,719 --> 00:01:58,760
ERIK SCHLUNTZ: On the computer science
team, we've made a lot of progress

46
00:01:58,760 --> 00:02:00,380
on our artificial intelligence.

47
00:02:00,380 --> 00:02:02,300
We threw out all the old strategy code.

48
00:02:02,300 --> 00:02:04,091
And we're making
something that's much more

49
00:02:04,091 --> 00:02:06,800
adaptive to how the
other team is playing.

50
00:02:06,800 --> 00:02:09,930
S on defense, we do things like
ranking the most dangerous players

51
00:02:09,930 --> 00:02:13,370
and covering them in a man to
man defense, based on that.

52
00:02:13,370 --> 00:02:16,430
And our offense, we generate these
maps over the entire field of how good

53
00:02:16,430 --> 00:02:19,557
the spot is, and then assign
a robot to use dynamically.

54
00:02:19,557 --> 00:02:21,390
RADHIKA NAGPAL: They
need to be able to see,

55
00:02:21,390 --> 00:02:24,100
they need people to
understand the world.

56
00:02:24,100 --> 00:02:27,250
They need to be able to move
fast and turn and manipulate.

57
00:02:27,250 --> 00:02:29,280
They need to be able to
notice their teammates

58
00:02:29,280 --> 00:02:31,810
and understand what they're doing.

59
00:02:31,810 --> 00:02:35,690
They need to have a strategy, and
they need to adapt their strategy,

60
00:02:35,690 --> 00:02:38,120
because their opponent will
be doing things all the time.

61
00:02:38,120 --> 00:02:40,372
And so you can't have
a pre-determined plan.

62
00:02:40,372 --> 00:02:41,580
You have to be able to adapt.

63
00:02:41,580 --> 00:02:42,640
>> KATE DONAHUE: Since
last year's competition,

64
00:02:42,640 --> 00:02:45,830
we've done a lot of advances,
especially in the wheel design.

65
00:02:45,830 --> 00:02:49,390
We've shifted the motors down, and
made everything much more compact,

66
00:02:49,390 --> 00:02:52,140
has allowed us to move our
center of gravity down, which

67
00:02:52,140 --> 00:02:53,110
allows us to go faster.

68
00:02:53,110 --> 00:02:56,276
And also put in a dribbler, which is
something we've wanted for a long time,

69
00:02:56,276 --> 00:02:58,443
but just haven't been able
to make it fit until now.

70
00:02:58,443 --> 00:03:00,817
ERIK SCHLUNTZ: So each of the
circuit boards on the robot

71
00:03:00,817 --> 00:03:02,040
has a different purpose.

72
00:03:02,040 --> 00:03:08,020
The four big ones, there and there and
there, each control one of the motors.

73
00:03:08,020 --> 00:03:11,660
So that it basically takes
a signal from the computer,

74
00:03:11,660 --> 00:03:14,370
decides how fast the wheels
should spin, and sends

75
00:03:14,370 --> 00:03:17,030
the correct voltage to
the wheels to do that.

76
00:03:17,030 --> 00:03:19,350
Like this board here
controls this motor,

77
00:03:19,350 --> 00:03:22,000
and this one here controls this motor.

78
00:03:22,000 --> 00:03:26,360
We also have these two boards in the
middle--this one right here controls

79
00:03:26,360 --> 00:03:30,680
charging these big
capacitors, for the kicker.

80
00:03:30,680 --> 00:03:34,430
This board here controls
when the robot kicks

81
00:03:34,430 --> 00:03:36,350
by using this light sensor right here.

82
00:03:36,350 --> 00:03:37,860
See when the ball is in front of it.

83
00:03:37,860 --> 00:03:39,568
It also controls the
dribbler here, which

84
00:03:39,568 --> 00:03:41,696
we use to put back spin on the ball.

85
00:03:41,696 --> 00:03:43,070
So we can move backwards with it.

86
00:03:43,070 --> 00:03:44,605
>> KATE DONAHUE: It's not
just you're trying to win,

87
00:03:44,605 --> 00:03:46,480
it's that you're trying
to advance knowledge.

88
00:03:46,480 --> 00:03:48,670
And so you work on whatever you work on.

89
00:03:48,670 --> 00:03:51,530
And then you have to release a
paper saying exactly what you

90
00:03:51,530 --> 00:03:53,750
did that's so cool, and how
other people can do that.

91
00:03:53,750 --> 00:03:55,530
And they can build on your work.

92
00:03:55,530 --> 00:03:58,710
And so if there's something so
advanced someone comes up with,

93
00:03:58,710 --> 00:04:01,430
everybody can sort of draw on that.

94
00:04:01,430 --> 00:04:05,290
You can have that for one year, and
maybe nobody else can use it that year,

95
00:04:05,290 --> 00:04:09,420
but then the next year everybody else
will modify it and done the same thing.

96
00:04:09,420 --> 00:04:13,670
And so it's really just moving forward
and not staying in the same place.

97
00:04:13,670 --> 00:04:17,249
>> ERIK SCHLUNTZ: I definitely want to
be working on robotics in my career.

98
00:04:17,249 --> 00:04:19,290
I think that there's just
a lot of amazing things

99
00:04:19,290 --> 00:04:22,350
that can be done to automate
things, to make cars safer,

100
00:04:22,350 --> 00:04:25,190
to make just really
everything work automatically,

101
00:04:25,190 --> 00:04:29,480
so that people don't have to do
things that are dangerous or dull.

102
00:04:29,480 --> 00:04:32,650
There's just so many things that
robots can do better than people.

103
00:04:32,650 --> 00:04:36,260
And I think that, as a society, we need
to start doing those to free people

104
00:04:36,260 --> 00:04:38,560
to do more interesting things.

105
00:04:38,560 --> 00:04:40,086