0:00:11.544,0:00:14.067
So what does it mean[br]for a machine to be athletic?

0:00:15.059,0:00:17.884
We will demonstrate the concept[br]of machine athleticism

0:00:17.908,0:00:19.604
and the research to achieve it

0:00:19.628,0:00:22.486
with the help of these flying machines[br]called quadrocopters,

0:00:22.510,0:00:23.557
or quads, for short.

0:00:26.291,0:00:28.184
Quads have been around for a long time.

0:00:28.501,0:00:31.778
They're so popular these days[br]because they're mechanically simple.

0:00:32.255,0:00:34.644
By controlling[br]the speeds of these four propellers,

0:00:34.668,0:00:37.413
these machines can roll, pitch, yaw,

0:00:37.437,0:00:39.608
and accelerate[br]along their common orientation.

0:00:40.179,0:00:43.377
On board are also a battery, a computer,

0:00:43.401,0:00:45.354
various sensors and wireless radios.

0:00:47.298,0:00:50.607
Quads are extremely agile,[br]but this agility comes at a cost.

0:00:51.762,0:00:53.850
They are inherently unstable,

0:00:53.874,0:00:56.445
and they need some form[br]of automatic feedback control

0:00:56.469,0:00:57.842
in order to be able to fly.

0:01:04.259,0:01:06.512
So, how did it just do that?

0:01:07.497,0:01:09.291
Cameras on the ceiling and a laptop

0:01:09.315,0:01:11.745
serve as an indoor[br]global positioning system.

0:01:12.301,0:01:14.340
It's used to locate objects in the space

0:01:14.364,0:01:16.452
that have these reflective[br]markers on them.

0:01:16.476,0:01:18.514
This data is then sent to another laptop

0:01:18.538,0:01:20.896
that is running estimation[br]and control algorithms,

0:01:20.920,0:01:22.874
which in turn sends commands to the quad,

0:01:22.898,0:01:25.531
which is also running estimation[br]and control algorithms.

0:01:29.864,0:01:32.106
The bulk of our research is algorithms.

0:01:32.130,0:01:34.706
It's the magic that brings[br]these machines to life.

0:01:36.291,0:01:40.324
So how does one design the algorithms[br]that create a machine athlete?

0:01:40.792,0:01:43.379
We use something broadly[br]called model-based design.

0:01:43.839,0:01:45.205
We first capture the physics

0:01:45.229,0:01:48.212
with a mathematical model[br]of how the machines behave.

0:01:48.636,0:01:52.586
We then use a branch of mathematics[br]called control theory

0:01:52.610,0:01:54.267
to analyze these models

0:01:54.291,0:01:57.061
and also to synthesize[br]algorithms for controlling them.

0:01:57.820,0:02:00.844
For example, that's how we can[br]make the quad hover.

0:02:00.868,0:02:04.158
We first captured the dynamics[br]with a set of differential equations.

0:02:04.182,0:02:07.634
We then manipulate these equations[br]with the help of control theory

0:02:07.658,0:02:10.595
to create algorithms[br]that stabilize the quad.

0:02:11.130,0:02:13.423
Let me demonstrate[br]the strength of this approach.

0:02:17.198,0:02:19.831
Suppose that we want[br]this quad to not only hover

0:02:19.855,0:02:21.514
but to also balance this pole.

0:02:22.520,0:02:24.294
With a little bit of practice,

0:02:24.318,0:02:27.001
it's pretty straightforward[br]for a human being to do this,

0:02:27.025,0:02:30.216
although we do have the advantage[br]of having two feet on the ground

0:02:30.240,0:02:32.299
and the use of our very versatile hands.

0:02:32.804,0:02:35.542
It becomes a little bit more difficult

0:02:35.566,0:02:37.545
when I only have one foot on the ground

0:02:37.569,0:02:39.171
and when I don't use my hands.

0:02:40.733,0:02:43.295
Notice how this pole has[br]a reflective marker on top,

0:02:43.319,0:02:45.980
which means that it can[br]be located in the space.

0:02:52.276,0:02:53.343
(Audience) Oh!

0:02:53.367,0:02:54.626
(Applause)

0:02:57.695,0:02:59.363
(Applause ends)

0:02:59.387,0:03:02.028
You can notice that this quad[br]is making fine adjustments

0:03:02.052,0:03:03.426
to keep the pole balanced.

0:03:04.417,0:03:06.489
How did we design[br]the algorithms to do this?

0:03:07.027,0:03:09.108
We added the mathematical[br]model of the pole

0:03:09.132,0:03:10.579
to that of the quad.

0:03:10.603,0:03:13.581
Once we have a model[br]of the combined quad-pole system,

0:03:13.605,0:03:16.924
we can use control theory to create[br]algorithms for controlling it.

0:03:18.538,0:03:20.102
Here, you see that it's stable,

0:03:20.126,0:03:22.795
and even if I give it little nudges,

0:03:22.819,0:03:23.985
it goes back --

0:03:25.589,0:03:27.358
to the nice, balanced position.

0:03:28.087,0:03:29.477
We can also augment the model

0:03:29.501,0:03:31.872
to include where we want[br]the quad to be in space.

0:03:32.428,0:03:35.371
Using this pointer,[br]made out of reflective markers,

0:03:35.395,0:03:37.824
I can point to where I want[br]the quad to be in space

0:03:37.848,0:03:39.445
a fixed distance away from me.

0:03:47.785,0:03:49.665
(Laughter)

0:03:55.763,0:03:58.889
The key to these acrobatic[br]maneuvers is algorithms,

0:03:58.913,0:04:01.206
designed with the help[br]of mathematical models

0:04:01.230,0:04:02.415
and control theory.

0:04:03.246,0:04:05.231
Let's tell the quad to come back here

0:04:05.255,0:04:06.919
and let the pole drop,

0:04:06.943,0:04:09.074
and I will next demonstrate the importance

0:04:09.098,0:04:11.087
of understanding physical models

0:04:11.111,0:04:13.156
and the workings of the physical world.

0:04:25.462,0:04:29.290
Notice how the quad lost altitude[br]when I put this glass of water on it.

0:04:29.314,0:04:30.717
Unlike the balancing pole,

0:04:30.741,0:04:33.756
I did not include the mathematical[br]model of the glass

0:04:33.780,0:04:35.087
in the system.

0:04:35.111,0:04:38.124
In fact, the system doesn't even know[br]that the glass is there.

0:04:38.148,0:04:40.886
Like before, I could use[br]the pointer to tell the quad

0:04:40.910,0:04:42.481
where I want it to be in space.

0:04:45.684,0:04:47.483
(Applause)

0:04:51.531,0:04:53.071
(Applause ends)

0:04:53.515,0:04:55.254
Okay, you should be asking yourself,

0:04:55.278,0:04:57.340
why doesn't the water[br]fall out of the glass?

0:04:57.888,0:04:58.918
Two facts.

0:04:58.942,0:05:02.575
The first is that gravity acts[br]on all objects in the same way.

0:05:03.051,0:05:04.830
The second is that the propellers

0:05:04.854,0:05:07.927
are all pointing in the same direction[br]of the glass, pointing up.

0:05:08.642,0:05:10.289
You put these two things together,

0:05:10.313,0:05:13.868
the net result is that all side forces[br]on the glass are small

0:05:13.892,0:05:16.169
and are mainly dominated[br]by aerodynamic effects,

0:05:16.193,0:05:18.154
which at these speeds are negligible.

0:05:23.336,0:05:25.670
And that's why you don't need[br]to model the glass.

0:05:25.694,0:05:28.435
It naturally doesn't spill,[br]no matter what the quad does.

0:05:32.500,0:05:35.063
(Audience) Oh!

0:05:38.340,0:05:40.198
(Applause)

0:05:43.159,0:05:44.414
(Applause ends)

0:05:45.940,0:05:47.054
The lesson here

0:05:47.078,0:05:51.091
is that some high-performance tasks[br]are easier than others,

0:05:51.115,0:05:53.452
and that understanding[br]the physics of the problem

0:05:53.476,0:05:56.170
tells you which ones are easy[br]and which ones are hard.

0:05:56.194,0:05:58.704
In this instance, carrying[br]a glass of water is easy.

0:05:58.728,0:06:00.358
Balancing a pole is hard.

0:06:02.171,0:06:06.442
We've all heard stories of athletes[br]performing feats while physically injured.

0:06:06.466,0:06:09.651
Can a machine also perform[br]with extreme physical damage?

0:06:10.636,0:06:12.121
Conventional wisdom says

0:06:12.145,0:06:16.028
that you need at least four fixed motor[br]propeller pairs in order to fly,

0:06:16.052,0:06:18.576
because there are four degrees[br]of freedom to control:

0:06:18.600,0:06:20.565
roll, pitch, yaw and acceleration.

0:06:21.192,0:06:24.413
Hexacopters and octocopters,[br]with six and eight propellers,

0:06:24.437,0:06:25.896
can provide redundancy,

0:06:25.920,0:06:27.885
but quadrocopters are much more popular

0:06:27.909,0:06:31.595
because they have the minimum number[br]of fixed motor propeller pairs: four.

0:06:32.168,0:06:33.184
Or do they?

0:06:40.827,0:06:41.827
(Audience) Oh!

0:06:42.407,0:06:43.407
(Laughter)

0:06:49.334,0:06:52.052
If we analyze the mathematical[br]model of this machine

0:06:52.076,0:06:54.228
with only two working propellers,

0:06:54.252,0:06:56.992
we discover that there's[br]an unconventional way to fly it.

0:07:07.980,0:07:09.662
We relinquish control of yaw,

0:07:09.686,0:07:12.732
but roll, pitch and acceleration[br]can still be controlled

0:07:12.756,0:07:16.102
with algorithms that exploit[br]this new configuration.

0:07:21.668,0:07:25.957
Mathematical models tell us[br]exactly when and why this is possible.

0:07:25.981,0:07:28.759
In this instance, this knowledge[br]allows us to design

0:07:28.783,0:07:31.256
novel machine architectures

0:07:31.280,0:07:34.948
or to design clever algorithms[br]that gracefully handle damage,

0:07:34.972,0:07:36.672
just like human athletes do,

0:07:36.696,0:07:38.960
instead of building[br]machines with redundancy.

0:07:40.688,0:07:42.345
We can't help but hold our breath

0:07:42.369,0:07:45.044
when we watch a diver[br]somersaulting into the water,

0:07:45.068,0:07:47.021
or when a vaulter is twisting in the air,

0:07:47.045,0:07:48.552
the ground fast approaching.

0:07:48.576,0:07:51.120
Will the diver be able[br]to pull off a rip entry?

0:07:51.144,0:07:52.951
Will the vaulter stick the landing?

0:07:53.292,0:07:56.472
Suppose we want this quad here[br]to perform a triple flip

0:07:56.496,0:07:59.322
and finish off at the exact same[br]spot that it started.

0:07:59.736,0:08:01.855
This maneuver is going[br]to happen so quickly

0:08:01.879,0:08:05.724
that we can't use position feedback[br]to correct the motion during execution.

0:08:05.748,0:08:07.345
There simply isn't enough time.

0:08:07.782,0:08:11.452
Instead, what the quad can do[br]is perform the maneuver blindly,

0:08:11.476,0:08:13.684
observe how it finishes the maneuver,

0:08:13.708,0:08:16.185
and then use that information[br]to modify its behavior

0:08:16.209,0:08:17.847
so that the next flip is better.

0:08:18.421,0:08:20.209
Similar to the diver and the vaulter,

0:08:20.233,0:08:22.128
it is only through repeated practice

0:08:22.152,0:08:24.295
that the maneuver can[br]be learned and executed

0:08:24.319,0:08:25.561
to the highest standard.

0:08:33.100,0:08:34.369
(Laughter)

0:08:34.393,0:08:38.178
(Applause)

0:08:39.164,0:08:42.612
Striking a moving ball[br]is a necessary skill in many sports.

0:08:43.127,0:08:44.619
How do we make a machine do

0:08:44.643,0:08:47.231
what an athlete does[br]seemingly without effort?

0:08:57.547,0:08:59.054
(Laughter)

0:09:04.143,0:09:06.000
(Applause)

0:09:08.500,0:09:09.798
(Applause ends)

0:09:10.542,0:09:13.168
This quad has a racket[br]strapped onto its head

0:09:13.192,0:09:16.655
with a sweet spot roughly the size[br]of an apple, so not too large.

0:09:17.017,0:09:19.822
The following calculations[br]are made every 20 milliseconds,

0:09:19.846,0:09:21.163
or 50 times per second.

0:09:21.602,0:09:23.877
We first figure out where[br]the ball is going.

0:09:24.414,0:09:27.035
We then next calculate[br]how the quad should hit the ball

0:09:27.059,0:09:29.544
so that it flies[br]to where it was thrown from.

0:09:29.950,0:09:34.382
Third, a trajectory is planned[br]that carries the quad

0:09:34.406,0:09:37.406
from its current state[br]to the impact point with the ball.

0:09:37.430,0:09:41.009
Fourth, we only execute 20 milliseconds'[br]worth of that strategy.

0:09:41.366,0:09:44.159
Twenty milliseconds later,[br]the whole process is repeated

0:09:44.183,0:09:45.950
until the quad strikes the ball.

0:09:55.290,0:09:57.694
(Applause)

0:09:58.131,0:10:01.554
Machines can not only perform[br]dynamic maneuvers on their own,

0:10:01.578,0:10:03.417
they can do it collectively.

0:10:03.441,0:10:07.028
These three quads are cooperatively[br]carrying a sky net.

0:10:15.589,0:10:17.268
(Applause)

0:10:20.372,0:10:21.961
(Applause ends)

0:10:21.985,0:10:26.326
They perform an extremely dynamic[br]and collective maneuver

0:10:26.350,0:10:28.018
to launch the ball back to me.

0:10:28.042,0:10:31.188
Notice that, at full extension,[br]these quads are vertical.

0:10:36.486,0:10:38.234
(Applause)

0:10:38.258,0:10:40.401
In fact, when fully extended,

0:10:40.425,0:10:43.688
this is roughly five times greater[br]than what a bungee jumper feels

0:10:43.712,0:10:45.682
at the end of their launch.

0:10:51.203,0:10:53.662
The algorithms to do this are very similar

0:10:53.686,0:10:56.601
to what the single quad used[br]to hit the ball back to me.

0:10:57.070,0:10:59.910
Mathematical models are used[br]to continuously re-plan

0:10:59.934,0:11:02.765
a cooperative strategy[br]50 times per second.

0:11:04.506,0:11:08.765
Everything we have seen so far has been[br]about the machines and their capabilities.

0:11:08.789,0:11:11.781
What happens when we couple[br]this machine athleticism

0:11:11.805,0:11:13.288
with that of a human being?

0:11:13.765,0:11:17.361
What I have in front of me[br]is a commercial gesture sensor

0:11:17.385,0:11:18.670
mainly used in gaming.

0:11:18.694,0:11:20.742
It can recognize[br]what my various body parts

0:11:20.766,0:11:22.083
are doing in real time.

0:11:22.710,0:11:24.789
Similar to the pointer[br]that I used earlier,

0:11:24.813,0:11:27.035
we can use this as inputs to the system.

0:11:27.454,0:11:29.826
We now have a natural way of interacting

0:11:29.850,0:11:33.111
with the raw athleticism[br]of these quads with my gestures.

0:12:10.715,0:12:14.572
(Applause)

0:12:24.354,0:12:26.370
Interaction doesn't have to be virtual.

0:12:26.394,0:12:27.584
It can be physical.

0:12:27.989,0:12:29.504
Take this quad, for example.

0:12:29.893,0:12:32.116
It's trying to stay[br]at a fixed point in space.

0:12:32.933,0:12:36.622
If I try to move it[br]out of the way, it fights me,

0:12:36.646,0:12:38.520
and moves back to where it wants to be.

0:12:40.254,0:12:42.341
We can change this behavior, however.

0:12:43.373,0:12:45.048
We can use mathematical models

0:12:45.072,0:12:47.736
to estimate the force[br]that I'm applying to the quad.

0:12:48.205,0:12:51.419
Once we know this force,[br]we can also change the laws of physics,

0:12:51.443,0:12:53.530
as far as the quad[br]is concerned, of course.

0:12:55.942,0:12:59.251
Here, the quad is behaving[br]as if it were in a viscous fluid.

0:13:02.570,0:13:06.289
We now have an intimate way[br]of interacting with a machine.

0:13:06.593,0:13:09.075
I will use this new capability to position

0:13:09.099,0:13:11.652
this camera-carrying quad[br]to the appropriate location

0:13:11.676,0:13:14.104
for filming the remainder[br]of this demonstration.

0:13:24.583,0:13:26.966
So we can physically interact[br]with these quads

0:13:26.990,0:13:28.950
and we can change the laws of physics.

0:13:29.393,0:13:31.370
Let's have a little bit of fun with this.

0:13:31.539,0:13:32.908
For what you will see next,

0:13:32.932,0:13:36.226
these quads will initially behave[br]as if they were on Pluto.

0:13:36.906,0:13:39.194
As time goes on, gravity will be increased

0:13:39.218,0:13:41.243
until we're all back on planet Earth,

0:13:41.267,0:13:43.033
but I assure you we won't get there.

0:13:43.462,0:13:44.486
Okay, here goes.

0:13:52.897,0:13:54.448
(Laughter)

0:14:22.778,0:14:25.738
(Laughter)

0:14:25.858,0:14:29.619
(Applause)

0:14:29.760,0:14:30.826
Whew!

0:14:35.045,0:14:36.272
You're all thinking now,

0:14:36.296,0:14:38.247
these guys are having way too much fun,

0:14:38.271,0:14:40.326
and you're probably also asking yourself,

0:14:40.350,0:14:43.151
why exactly are they building[br]machine athletes?

0:14:44.286,0:14:47.096
Some conjecture that the role[br]of play in the animal kingdom

0:14:47.120,0:14:49.257
is to hone skills[br]and develop capabilities.

0:14:49.690,0:14:51.905
Others think that it has[br]more of a social role,

0:14:51.929,0:14:53.579
that it's used to bind the group.

0:14:53.603,0:14:56.939
Similarly, we use the analogy[br]of sports and athleticism

0:14:56.963,0:14:59.018
to create new algorithms for machines

0:14:59.042,0:15:00.485
to push them to their limits.

0:15:01.350,0:15:04.398
What impact will the speed[br]of machines have on our way of life?

0:15:04.866,0:15:07.231
Like all our past creations[br]and innovations,

0:15:07.255,0:15:10.058
they may be used to improve[br]the human condition

0:15:10.082,0:15:12.056
or they may be misused and abused.

0:15:12.595,0:15:14.897
This is not a technical choice[br]we are faced with;

0:15:14.921,0:15:15.936
it's a social one.

0:15:16.332,0:15:17.683
Let's make the right choice,

0:15:17.707,0:15:20.619
the choice that brings out the best[br]in the future of machines,

0:15:20.643,0:15:23.508
just like athleticism in sports[br]can bring out the best in us.

0:15:24.380,0:15:27.455
Let me introduce you to the wizards[br]behind the green curtain.

0:15:27.479,0:15:30.813
They're the current members[br]of the Flying Machine Arena research team.

0:15:30.837,0:15:35.054
(Applause)

0:15:35.267,0:15:37.545
Federico Augugliaro, Dario Brescianini,

0:15:37.569,0:15:40.994
Markus Hehn, Sergei Lupashin,[br]Mark Muller and Robin Ritz.

0:15:41.018,0:15:43.542
Look out for them.[br]They're destined for great things.

0:15:43.566,0:15:44.620
Thank you.

0:15:44.644,0:15:48.100
(Applause)