0:00:00.672,0:00:03.597
This is Pleurobot.

0:00:03.597,0:00:07.127
Pleurobot is a robot that we designed[br]to closely mimic a salamander species

0:00:07.127,0:00:09.124
called ??

0:00:09.124,0:00:11.492
Pleurobot can walk, as you can see here,

0:00:11.492,0:00:14.301
and as you'll see later, it can also swim.

0:00:14.301,0:00:16.948
So you might ask,[br]why did we design this robot?

0:00:16.948,0:00:21.639
And in fact, this robot has been designed[br]as a scientific tool for neuroscience.

0:00:21.639,0:00:24.100
Indeed, we designed it[br]together with neurobiologists

0:00:24.100,0:00:25.981
to understand how animals move,

0:00:25.981,0:00:29.533
and especially how the spinal cord[br]controls locomotion.

0:00:29.533,0:00:31.739
But the more I work in biorobotics,

0:00:31.739,0:00:34.177
the more I'm really impressed[br]by animal locomotion.

0:00:34.177,0:00:38.449
If you think of a dolphin swimming[br]or a cat running or jumping around,

0:00:38.449,0:00:40.492
or even us as humans,

0:00:40.492,0:00:41.978
when you go jogging or play tennis,

0:00:41.978,0:00:44.091
we do amazing things.

0:00:44.091,0:00:47.806
And in fact, our nervous system solves[br]a very, very complex control problem.

0:00:47.806,0:00:51.119
It has to coordinate more[br]or less 200 muscles perfectly,

0:00:51.374,0:00:55.809
because if the coordination is bad,[br]we fall over or we do bad locomotion.

0:00:55.809,0:00:59.454
And my goal is to understand[br]how this works.

0:00:59.454,0:01:02.705
There are four main components[br]behind animal locomotion.

0:01:02.705,0:01:04.957
The first component is just the body,

0:01:04.957,0:01:08.486
and in fact we should never underestimate[br]what extent the biomechanics

0:01:08.486,0:01:10.901
already simplify locomotion in animals.

0:01:10.901,0:01:12.480
Then you have the spinal cord,

0:01:12.480,0:01:14.523
and in the spinal cord you find reflexes,

0:01:14.523,0:01:18.029
like multiple reflexes that create[br]a sensory motor coordination loop

0:01:18.029,0:01:22.162
between neural activity in the spinal cord[br]and mechanical activity.

0:01:22.162,0:01:24.925
A third component[br]are central pattern generators.

0:01:24.925,0:01:29.012
These are very interesting circuits[br]in the spinal cord of vertebrate animals

0:01:29.012,0:01:33.400
that can generate, by themselves, very[br]coordinated rhythmic patterns of activity

0:01:33.400,0:01:35.978
while receiving only[br]very simple input signals.

0:01:35.978,0:01:38.601
And these input signals come from[br]descending modulation

0:01:38.601,0:01:41.434
from higher parts of the brain,[br]from the motor cortex, the cerebellum,

0:01:41.434,0:01:45.196
the basal ganglia, will all modulate[br]activity of the spinal cord

0:01:45.196,0:01:46.658
while we do locomotion.

0:01:46.658,0:01:49.654
But what's interesting is to what extent[br]just a low level component,

0:01:49.654,0:01:52.289
the spinal cord, together with the body,[br]already solves a big part

0:01:52.289,0:01:53.632
of the locomotion problem,

0:01:53.632,0:01:57.222
and you probably know it by the fact[br]that you can cut the head of the chicken,

0:01:57.222,0:01:58.663
it can still run for a while,

0:01:58.663,0:02:00.961
showing that just the lower part,[br]spinal cord and body,

0:02:00.961,0:02:03.492
already solves a big part of locomotion.

0:02:03.492,0:02:06.116
Now, understanding how this works[br]is very complex,

0:02:06.116,0:02:07.416
because first of all,

0:02:07.416,0:02:09.831
recording activity in the spinal cord[br]is very difficult.

0:02:09.831,0:02:12.710
It's much easier to implant electrodes[br]in the motor cortex

0:02:12.710,0:02:15.659
than in the spinal cord, because[br]it's protected by the vertebrae.

0:02:15.659,0:02:17.540
Especially in humans,[br]it's very hard to do.

0:02:17.540,0:02:21.162
A second difficulty is that locomotion[br]is really due to a very complex

0:02:21.162,0:02:24.436
and very dynamic interaction[br]between these four components.

0:02:24.436,0:02:28.011
So it's very hard to find out[br]what's the role of each over time.

0:02:28.011,0:02:32.655
This is where biorobots like Pleurobot[br]and mathematical models

0:02:32.655,0:02:34.559
can really help.

0:02:34.559,0:02:36.951
So what's biorobotics?

0:02:36.951,0:02:38.994
Biorobotics is a very active field[br]of research in robotics

0:02:38.994,0:02:42.802
where people want to take inspiration[br]from animals to make robots

0:02:42.802,0:02:44.567
to go outdoors,

0:02:44.567,0:02:47.353
like service robots[br]or search-and-rescue robots

0:02:47.353,0:02:48.885
or field robots,

0:02:48.885,0:02:51.138
and the big goal here is[br]to take inspiration from animals

0:02:51.138,0:02:53.483
to make robotics that can handle[br]complex terrain --

0:02:53.483,0:02:55.572
stairs, mountains, forests,

0:02:55.572,0:02:57.616
places where robots[br]still have difficulties

0:02:57.616,0:02:59.891
and where animals can do[br]a much better job.

0:02:59.891,0:03:02.468
The robot can be[br]a wonderful scientific tool as well.

0:03:02.468,0:03:04.581
There are some very nice projects[br]where robots are used

0:03:04.581,0:03:08.947
like a scientific tool for neuroscience,[br]for biomechanics, or for ?? dynamics.

0:03:08.947,0:03:11.617
And this is exactly[br]the purpose of Pleurobot.

0:03:11.617,0:03:14.473
So what we do in my lab[br]is to collaborate with neurobiologists

0:03:14.473,0:03:18.048
like Jean-Marie Cabelguen,[br]a neurobiologist in Bordeaux in France,

0:03:18.048,0:03:22.344
and we want to make spinal cord models[br]and validate them on robots.

0:03:22.344,0:03:23.992
And here we want to start simple.

0:03:23.992,0:03:26.036
So it's good to start with simple animals

0:03:26.036,0:03:28.404
like lampreys, which are[br]very primitive fish,

0:03:28.404,0:03:30.796
and then gradually go toward[br]more complex locomotion,

0:03:30.796,0:03:32.235
like in salamanders,

0:03:32.235,0:03:34.046
but also in cats and in humans,

0:03:34.046,0:03:35.718
in mammals.

0:03:35.718,0:03:37.877
And here, a robot becomes[br]an interesting tool

0:03:37.877,0:03:39.572
to validate our models,

0:03:39.572,0:03:42.614
and in fact, for me, Pleurobot[br]is a kind of dream becoming true.

0:03:42.614,0:03:46.770
Like, more or less 20 years ago[br]I was already working on a computer

0:03:46.770,0:03:49.510
making simulations of lamprey[br]and salamander locomotion

0:03:49.510,0:03:50.903
during my Ph.D.

0:03:50.903,0:03:54.270
But I always knew that my simulations[br]were just approximations.

0:03:54.270,0:03:58.171
Like, simulating the physics in water[br]or with mud or with complex ground,

0:03:58.171,0:04:00.887
it's very hard to simulate that[br]properly on a computer.

0:04:00.887,0:04:03.813
Why not have a real robot[br]and real physics?

0:04:03.813,0:04:06.831
So among all these animals,[br]one of my favorites is the salamander.

0:04:06.831,0:04:10.221
You might as why, and it's because[br]as an amphibian,

0:04:10.221,0:04:12.822
it's a really key animal[br]from an evolutionary point of view.

0:04:12.822,0:04:14.145
It makes a wonderful link

0:04:14.145,0:04:17.025
between swimming, as you find it[br]in eels or fish,

0:04:17.025,0:04:21.947
and quadruped motion, as you see[br]in mammals, in cats and humans.

0:04:21.947,0:04:23.805
And in fact, the modern salamander

0:04:23.805,0:04:25.894
is very close to the first[br]terrestrial vertebrate,

0:04:25.894,0:04:27.705
so it's almost a living fossil,

0:04:27.705,0:04:30.027
which gives us access to our ancestor,

0:04:30.027,0:04:33.255
the ancestor to all terrestrial tetrapods.

0:04:33.255,0:04:35.971
So the salamander swims by doing[br]what's using what's called

0:04:35.971,0:04:37.109
a ??? swimming gait,

0:04:37.109,0:04:40.917
so they propagate a nice traveling wave[br]of muscle activity from head to tail.

0:04:40.917,0:04:43.146
And if you place the salamander[br]on the ground,

0:04:43.146,0:04:45.607
it switches to what's called[br]a walking trot gait.

0:04:45.607,0:04:48.639
In this case, you have nice[br]activation of the limbs

0:04:48.639,0:04:52.052
which are very nicely coordinated[br]with this standing wave undulation

0:04:52.052,0:04:53.469
of the body,

0:04:53.469,0:04:57.068
and that's exactly the gait[br]that you are seeing here on Pleurobot.

0:04:57.068,0:04:59.831
Now, one thing which is very surprising[br]and fascinating in fact

0:04:59.831,0:05:03.964
is the fact that all this can be generated[br]just by the spinal cord and the body.

0:05:03.964,0:05:05.984
So if you take ?? salamander --

0:05:05.984,0:05:08.027
it's not so nice[br]but you remove the head --

0:05:08.027,0:05:10.488
and if you electrically stimulate[br]the spinal cord,

0:05:10.488,0:05:13.938
at low level of stimulation[br]this will use a walking-like gait.

0:05:13.938,0:05:16.516
If you stimulate a bit more,[br]the gait accelerates,

0:05:16.516,0:05:19.116
and at some point, there's a transfer,[br]and automatically,

0:05:19.116,0:05:21.322
the animal switches to swimming.

0:05:21.322,0:05:22.645
This is amazing,

0:05:22.645,0:05:25.501
just changing the global drive[br]as if you are pressing the gas pedal

0:05:25.501,0:05:27.916
of descending modulation[br]to your spinal cord,

0:05:27.916,0:05:32.072
makes a complete switch[br]between two very different gaits.

0:05:32.072,0:05:35.253
And in fact, the same[br]has been observed in cats.

0:05:35.253,0:05:36.856
If you stimulate the spinal cord of a cat,

0:05:36.856,0:05:39.201
you can switch between[br]walk, trot, and gallop.

0:05:39.201,0:05:42.080
Or in birds, you can make a bird[br]switch between walking,

0:05:42.080,0:05:43.543
at low levels of stimulation,

0:05:43.543,0:05:46.375
and flapping its wings[br]at high level stimulation.

0:05:46.375,0:05:49.417
And this really shows that the spinal cord[br]is a very sophisticated

0:05:49.417,0:05:51.089
locomotion controller.

0:05:51.089,0:05:53.341
So we studied salamander locomotion[br]in more detail,

0:05:53.341,0:05:56.638
and we had in fact access[br]to a very nice x-ray video machine

0:05:56.638,0:05:59.935
from Professor Martin Fischer[br]in Jena University in Germany.

0:05:59.935,0:06:03.000
And thanks to that, you really have[br]an amazing machine

0:06:03.000,0:06:05.438
to record all the bone motion[br]in great detail.

0:06:05.438,0:06:06.762
That's what we did.

0:06:06.762,0:06:09.873
So we basically figured out[br]which bones are important for us

0:06:09.873,0:06:12.520
and collected their motion in 3D.

0:06:12.520,0:06:15.399
And what we did is collect[br]a whole database of motions,

0:06:15.399,0:06:17.187
both on ground and in water,

0:06:17.187,0:06:19.346
to really collect the whole database[br]of motor behaviors

0:06:19.346,0:06:21.041
that a real animal can do,

0:06:21.041,0:06:23.642
and then our job as roboticists[br]was to replicate that in our robot.

0:06:23.642,0:06:27.450
So we did a whole optimization process[br]to find out the right structure,

0:06:27.450,0:06:30.120
where to place the motors,[br]how to connect them together,

0:06:30.120,0:06:33.688
to be able to replay[br]these motions as well as possible.

0:06:33.688,0:06:36.023
And this is how Pleurobot came to life.

0:06:37.184,0:06:40.063
So let's look how closely it is[br]to the real animal.

0:06:40.063,0:06:43.523
So what you see here is almost[br]a direct comparison

0:06:43.523,0:06:46.239
between the walking[br]of the real animal and the Pleurobot.

0:06:46.239,0:06:48.886
You can see that we have almost[br]A one-to-one exact replay

0:06:48.886,0:06:50.512
of the walking gait.

0:06:50.512,0:06:53.344
If you go backwards and slowly,[br]you see it even better.

0:06:55.782,0:06:58.151
But even better, we can do swimming.

0:06:58.151,0:07:00.682
So for that we have a dry suit[br]that we put all over the robot --

0:07:00.682,0:07:02.261
(Laughter) --

0:07:02.261,0:07:05.465
and then we can go in water[br]and start replaying the swimming gaits.

0:07:05.465,0:07:07.578
And here, we were very happy,[br]because this is difficult to do.

0:07:07.578,0:07:10.782
The physics of interaction are complex.

0:07:10.782,0:07:13.313
Our robot is much bigger[br]than a small animal,

0:07:13.313,0:07:16.099
so we had to do what's called[br]dynamical scaling of the frequencies

0:07:16.099,0:07:18.630
to make sure we had the same[br]interaction physics.

0:07:18.630,0:07:21.300
But you see at the end[br]we have a very close match,

0:07:21.300,0:07:23.715
and we were very, very happy with this.

0:07:23.715,0:07:25.851
So let's go do the spinal cord.

0:07:25.851,0:07:27.987
So here what we did[br]with Jean-Marie Cabelguen

0:07:27.987,0:07:30.472
is model the spinal cord circuits.

0:07:30.472,0:07:33.397
And what's interesting is that[br]the salamander has kept

0:07:33.397,0:07:34.930
a very primitive circuit

0:07:34.930,0:07:37.530
which is very similar to the one[br]we find in the lamprey,

0:07:37.530,0:07:39.504
pretty much this eel-like fish,

0:07:39.504,0:07:41.199
and it looks like during evolution,

0:07:41.199,0:07:44.496
new neuronal oscillators have been added[br]to control the limbs,

0:07:44.496,0:07:46.028
to do the leg locomotion.

0:07:46.028,0:07:48.002
And we know where[br]these neuronal oscillators are

0:07:48.002,0:07:49.999
but what we did was to make[br]a mathematical model

0:07:49.999,0:07:51.508
to see how they should be coupled

0:07:51.508,0:07:54.318
to allow this transition between[br]the two very different gaits.

0:07:54.318,0:07:57.940
And we tested that on board of a robot.

0:07:57.940,0:07:59.728
And this is how it looks.

0:08:06.716,0:08:08.249
So what you see here

0:08:08.249,0:08:10.176
is a previous version of Pleurobot

0:08:10.176,0:08:12.800
that's completely controlled[br]by our spinal cord model

0:08:12.800,0:08:14.890
programmed on board of the robot.

0:08:14.890,0:08:17.722
And the only thing we do[br]is send to the robot

0:08:17.722,0:08:19.046
through a remote control

0:08:19.046,0:08:21.135
the two descending signals[br]it normally should receive

0:08:21.135,0:08:23.272
from the upper part of the brain.

0:08:23.272,0:08:25.361
And what's interesting is,[br]by playing with these signals,

0:08:25.361,0:08:29.796
we can completely control[br]speed, heading, and type of gait.

0:08:29.796,0:08:31.050
For instance,

0:08:31.050,0:08:33.999
when we stimulate at a low level,[br]we have the walking gait,

0:08:33.999,0:08:37.361
and at some point, if we stimulate a lot,[br]very rapidly it switches

0:08:37.361,0:08:39.521
to the swimming gait.

0:08:39.521,0:08:41.262
And finally, we can also do turning

0:08:41.262,0:08:46.115
very nicely by just stimulating more one[br]side of the spinal cord than the other.

0:08:46.115,0:08:47.879
And I think it's really beautiful

0:08:47.879,0:08:50.201
how nature has distributed control

0:08:50.201,0:08:52.825
to really give a lot of responsibility[br]to the spinal cord

0:08:52.825,0:08:56.378
so that the upper part of the brain[br]doesn't need to worry about every muscle.

0:08:56.378,0:08:58.746
It just has to worry about[br]this high-level modulation,

0:08:58.746,0:09:03.088
and it's really the job of the spinal cord[br]to coordinate all the muscles.

0:09:03.088,0:09:06.942
So now let's go to cat locomotion,[br]and the importance of biomechanics.

0:09:06.942,0:09:08.591
So this is another project

0:09:08.591,0:09:10.820
where we studied cat biomechanics,

0:09:10.820,0:09:14.953
and we wanted to see how much[br]the morphology helps locomotion.

0:09:14.953,0:09:18.621
And we found three important[br]criteria in the properties,

0:09:18.621,0:09:20.293
basically, of the limbs.

0:09:20.293,0:09:22.244
The first one is that a cat limb

0:09:22.244,0:09:24.837
more or less looks[br]like a pantograph-like structure.

0:09:24.837,0:09:27.251
So a pantograph is a mechanical structure

0:09:27.251,0:09:31.733
which keeps the upper segment[br]and the lower segments always parallel.

0:09:31.733,0:09:34.449
So a simple geometrical system[br]that kind of coordinates a bit

0:09:34.449,0:09:36.423
the internal movement of the segments.

0:09:36.423,0:09:39.372
A second property of cat limbs[br]is that they are very lightweight.

0:09:39.372,0:09:41.508
Most of the muscles are in the trunk,

0:09:41.508,0:09:44.015
which is a good idea, because then[br]the limbs have low inertia

0:09:44.015,0:09:46.035
and can be moved very rapidly.

0:09:46.035,0:09:50.168
The last final important property is this[br]very elastic behavior of the cat limb,

0:09:50.168,0:09:52.769
so to handle impacts and forces.

0:09:52.769,0:09:55.021
And this is how we designed Cheetah-Cub.

0:09:55.021,0:09:57.366
So let's invite Cheetah-Cub onstage.

0:09:57.366,0:10:06.073
So this is Peter Eckert, who does[br]his Ph.D on this robot,

0:10:06.073,0:10:08.140
and as you see, it's a cute little robot.

0:10:08.140,0:10:09.278
It looks a bit like a toy,

0:10:09.278,0:10:11.367
But it was really used[br]as a scientific tool

0:10:11.367,0:10:14.595
to investigate these properties[br]of the legs of the cat.

0:10:14.595,0:10:16.777
So you see, it's very compliant,[br]very lightweight,

0:10:16.777,0:10:18.821
and also very elastic,

0:10:18.821,0:10:21.305
so you can easily press it down[br]and it will not break.

0:10:21.305,0:10:23.070
It will just jump, in fact.

0:10:23.070,0:10:26.622
And this very elastic property[br]is also very important.

0:10:26.622,0:10:29.826
And you also see a bit this properties[br]of these three segments

0:10:29.826,0:10:32.566
of the leg as pantograph.

0:10:32.566,0:10:35.167
Now, what's interesting is that[br]this quite dynamic gait

0:10:35.167,0:10:36.955
is obtained purely in open loop,

0:10:36.955,0:10:38.325
meaning no sensors,[br]no complex feedback loops.

0:10:38.325,0:10:42.458
And that's interesting, because it means

0:10:42.458,0:10:46.405
that just the mechanics already stabilized[br]this quite rapid gait,

0:10:46.405,0:10:51.095
and that really good mechanics[br]already basically simplify locomotion.

0:10:51.095,0:10:54.090
To the extent that we can even[br]disturb a bit locomotion,

0:10:54.090,0:10:55.832
as you will see in the next video,

0:10:55.832,0:10:59.593
where we can for instance do some exercise[br]where we have the robot go down a step,

0:10:59.593,0:11:01.474
and the robot will not fall over,

0:11:01.474,0:11:03.215
which was a surprise for us.

0:11:03.215,0:11:04.562
This is a small perturbation.

0:11:04.562,0:11:06.536
I was expecting the robot[br]to immediately fall over,

0:11:06.536,0:11:08.811
because there is no sensors,[br]no fast feedback loop.

0:11:08.811,0:11:11.203
But no, just the mechanics[br]stabilized the gait,

0:11:11.203,0:11:12.851
and the robot doesn't fall over.

0:11:12.851,0:11:15.916
Obviously, if you make the step bigger,[br]and if you have obstacles,

0:11:15.916,0:11:19.840
you need the full control loops[br]and reflexes and everything,

0:11:19.840,0:11:22.464
but what's important here[br]is that just for small perturbation,

0:11:22.464,0:11:24.461
the mechanics are right.

0:11:24.461,0:11:26.086
And I think this is[br]a very important message

0:11:26.086,0:11:28.803
from biomechanics and robotics[br]to neuroscience,

0:11:28.803,0:11:30.753
saying don't underestimate to what extent

0:11:30.753,0:11:34.445
the body already helps locomotion.

0:11:34.445,0:11:37.742
Now, how does this relate[br]to human locomotion?

0:11:37.742,0:11:42.177
Clearly, human locomotion is more complex[br]than cat and salamander locomotion,

0:11:42.177,0:11:45.172
but at the same time, the nervous system[br]of humans is very similar

0:11:45.172,0:11:47.355
to that of other vertebrates.

0:11:48.024,0:11:52.041
and especially the spinal cord is also the[br]key controller for locomotion in humans.

0:11:52.041,0:11:53.945
That's why, if there's a lesion[br]of the spinal cord,

0:11:53.945,0:11:55.803
this has dramatic effects.

0:11:55.803,0:11:57.475
The person can become [br]paraplegic or tetraplegic.

0:11:57.475,0:12:00.925
This is become the brain[br]loses its communication

0:12:00.925,0:12:02.225
with the spinal cord.

0:12:02.225,0:12:04.408
Especially, it loses[br]the descending modulation

0:12:04.408,0:12:07.496
to initiate and modulate locomotion.

0:12:07.496,0:12:09.307
So a big goal of neuroprosthetics

0:12:09.307,0:12:11.698
is to be able to reactivate[br]that communication

0:12:11.698,0:12:14.670
using electrical or chemical stimulations.

0:12:14.670,0:12:17.364
And there are several teams[br]in the world that do exactly that,

0:12:17.364,0:12:18.710
especially at EPFL.

0:12:18.710,0:12:21.218
My colleagues ?? and ??,

0:12:21.218,0:12:23.424
with whom I collaborate.

0:12:23.424,0:12:26.744
But to do this properly,[br]it's very important to understand

0:12:26.744,0:12:28.648
how the spinal cord works,

0:12:28.648,0:12:30.552
how it interacts with the body,

0:12:30.552,0:12:34.058
and how the brain communicates[br]with the spinal cord.

0:12:34.058,0:12:36.519
This is where the robots[br]and models that I've presented today

0:12:36.519,0:12:38.563
will hopefully play a key role

0:12:38.563,0:12:41.303
towards these very important goals.

0:12:41.303,0:12:42.765
Thank you.

0:12:42.765,0:12:48.826
(Applause)

0:12:51.914,0:12:54.305
Bruno Giussani: Okay, I've seen[br]in your lab other robots

0:12:54.305,0:12:57.533
that do things like swim in pollution[br]and measure the pollution

0:12:57.533,0:12:59.669
while they swim,

0:12:59.669,0:13:04.731
but for this one, you mentioned[br]in your talk, like a side project,

0:13:04.731,0:13:06.843
search and rescue,

0:13:06.843,0:13:09.096
and it does have a camera on its nose.

0:13:09.096,0:13:12.849
Auke Ijspeert: Absolutely. So the robot,[br]we have some spinoff projects

0:13:12.849,0:13:16.053
where we would like to use the robots[br]to do search and rescue inspection,

0:13:16.053,0:13:17.772
so this robot is now seeing you.

0:13:17.772,0:13:21.440
And the big dream is to,[br]if you have a difficult situation

0:13:21.440,0:13:24.877
like a collapsed building[br]or a building that is flooded,

0:13:24.877,0:13:28.545
and this is very dangerous[br]for a rescue team or even rescue dogs,

0:13:28.545,0:13:31.378
why not send in a robot[br]that can crawl around, swim, walk,

0:13:31.378,0:13:34.745
with a camera onboard[br]to do inspection and identify survivors

0:13:34.745,0:13:36.997
and possibly create a communication link[br]with the survivor.

0:13:36.997,0:13:41.246
BG: Of course, assuming the survivors[br]don't get scared by the shape of this.

0:13:41.246,0:13:44.427
AI: Yeah, we should probably change[br]the appearance quite a bit,

0:13:44.427,0:13:47.074
because here I guess a survivor[br]might die of a heart attack

0:13:47.074,0:13:49.744
just by being worried that this[br]would feed on you.

0:13:49.744,0:13:52.461
But by changing the appearance[br]and it making it more robust,

0:13:52.461,0:13:53.877
I'm sure we can make[br]a good tool out of it.

0:13:53.877,0:13:56.222
BG: All right, thank you very much.[br]Thank you and your team.

0:13:56.222,0:13:57.105
(Applause)