00:00:01
My name is Josh.
00:00:02
I've been at Boston Dynamics
about four years now.
00:00:05
I'm a mechanical engineer
on the Atlas team.
00:00:07
I think what first got me
interested in high school,
00:00:10
I'd build a drone,
a little quadcopter,
00:00:12
and it was it was
absolutely terrible.
00:00:14
I could barely fly,
but we did everything.
00:00:17
We did all the software,
the electronics, and,
00:00:20
just going through the process
of seeing how we're tuning
00:00:23
variables and it's responding
to to inputs and trying to
00:00:25
stabilize was truly empowering.
00:00:28
And I think from there,
I just felt, you know,
00:00:30
a desire to do more.
00:00:31
And in college, I kept on
learning and building more
00:00:33
robots, did all the
projects I could,
00:00:35
and eventually that
led to working here.
00:00:44
As a mechanical engineer,
00:00:45
we design all the physical
parts on the robot,
00:00:47
the arms and the
legs, structures,
00:00:49
the motors and
actuation, batteries,
00:00:51
anything on the robot that's
physical and you can see there,
00:00:54
that's where we specialize.
00:00:55
We also do a lot of analysis
and math to make sure that the
00:00:58
robot will work, that
the joints are strong enough,
00:01:00
that if it face plants
while doing a back flip,
00:01:02
things aren't gonna
break and fall apart.
00:01:04
You know, know, it kinda
goes through phases.
00:01:05
Some weeks, we're
doing a lot of testing,
00:01:07
and we're breaking robots
intentionally or or unintentionally.
00:01:11
Testing mechanisms, if they
perform the way we expect to.
00:01:13
Other weeks are
really design heavy.
00:01:15
And like everyone here,
00:01:16
we coordinate a lot with
other teams and disciplines,
00:01:18
electrical engineers,
software engineers,
00:01:21
to make sure that the
robot will will work well.
00:01:24
So how do you build a robot?
00:01:27
So, yeah, for, the robots
here in Boston Dynamics,
00:01:30
we really have to approach
it from a systems level.
00:01:33
You can't sort of design each
part individually in a bubble.
00:01:36
It'll never work that way.
00:01:37
You can think of it sort of as
a a picture coming into focus
00:01:41
where there's a lot of blurred
lines and and things aren't
00:01:45
really don't have
much definition,
00:01:46
but we're trying to think
about everything holistically.
00:01:48
And then throughout stages,
00:01:50
we're detailing out parts kind
of in parallel at the same time.
00:01:53
So we take all these
people, you know,
00:01:55
and as we're building the robot,
00:01:56
everyone's voicing
concerns, contributing,
00:01:59
and then kinda through
that large team effort,
00:02:00
you end up with a robot.
00:02:02
You're never the smartest
person in the room and that's
00:02:05
the best place to be.
00:02:07
There's always so much to learn.
00:02:09
What does Atlas do? That
question gets asked a lot.
00:02:13
So Atlas is our R&D robot, our
research and development robot.
00:02:17
And you can think of it sort of
as a technology demonstrator.
00:02:21
So we're really trying to push
the limits of hardware and
00:02:24
software and
controls with Atlas,
00:02:26
thinking about years and years
into the into the future I mean,
00:02:29
how we can learn from that technology
and spread it to all of our robots.
00:02:33
How fast can Atlas go?
00:02:36
So the fastest we've gotten it
to go in lab is about two and a
00:02:39
half meters per second.
00:02:41
It's about five and a
half miles per hour.
00:02:43
But there's sort of this
running joke around here how we
00:02:46
keep on thinking we've
capped out the robot,
00:02:48
and then the software team
develops some new behavior that
00:02:52
is more athletic and coordinated
than anything we've done before.
00:02:56
So we think there's quite a
long ways to go before we've
00:02:58
capped out the hardware.
00:03:00
What was the first
prototype of Atlas?
00:03:03
So the first was
way before my time.
00:03:05
It was called Petman Proto.
00:03:06
It was actually
our robot, BigDog,
00:03:06
which you may have seen.
00:03:07
It's one of the classic ones where
we kick it and it stabilizes.
00:03:09
It was that robot.
00:03:11
We just took
00:03:13
took two legs from it and
stood it on its hind legs,
00:03:15
and then we attached
two arms to it.
00:03:18
And that was the predecessor to our
first full humanoid robot called Petman,
00:03:22
and then that led to over
a decade of development and
00:03:26
iteration to the robot
you see nowadays,
00:03:28
which we refer to
as HD internally,
00:03:31
which is humanoid version d,
00:03:33
and that's what you see on our
YouTube videos doing the parkour.
00:03:36
What hardware do you
specifically focus on?
00:03:39
So on Atlas, that robot was more or less
finished when by the time I got here.
00:03:44
So I just did a few upgrades.
We upgraded the battery.
00:03:47
So working on how that
battery gets installed and,
00:03:50
the roll cage that
protects that.
00:03:52
On stretch, I focused
on the distal arm,
00:03:55
which is the end of the robot by
the wrist and the gripper as well.
00:03:59
How does manufacturing
differ between robots?
00:04:03
It's a good question.
00:04:04
So we have a really broad range of
of manufacturing here at the company.
00:04:08
So on Atlas, we design a lot of metal
three d printed parts that are super
00:04:12
lightweight and very,
very strong and stiff.
00:04:15
Whereas on stretch and spot,
00:04:17
we're we're designing for very
specific processes that scale
00:04:21
well, towards production.
00:04:24
Apart from software
and electronics,
00:04:26
how much of mechanics is needed
to study to approach robotics?
00:04:30
Nowadays, robotics is super
approachable from all majors.
00:04:34
You don't kind of have to know
everything in every field to do robotics,
00:04:38
which is really exciting because
it it didn't used to be that way.
00:04:41
If you're really
interested in software,
00:04:43
you can buy pre built kits
that let you spend ninety five
00:04:46
percent of your time coding,
00:04:47
and you don't have to
worry about the mechanisms.
00:04:50
And likewise,
00:04:51
if you're really into
machining and designing parts,
00:04:53
there's a lot of
microcontrollers that make
00:04:56
programming really simple, And,
00:04:58
then you can focus on
building cool machines.
00:05:00
There's a lot you can still
do to interact with robots,
00:05:03
without an engineering degree.
00:05:04
So, the applications of robotics are
only getting larger every single year,
00:05:09
and there's a lot of roles
where you can introduce
00:05:12
robotics into new fields.
00:05:14
How can I get
started with robots?
00:05:17
There's a ton of ways to get
started with robotics nowadays.
00:05:21
In in most schools, there's
now robotics programs.
00:05:24
A lot of the principles there
really truly apply to the types
00:05:27
of things we do here.
00:05:28
If you don't have access to
that or really even in general,
00:05:31
I would encourage you just to
take apart everything that your
00:05:34
parents won't be sad
if it gets destroyed.
00:05:36
If you get excited about
it and if you're learning,
00:05:38
you're doing the right thing.
00:05:39
So just keep on
going down that path.
00:05:41
But a lot of people here,
00:05:43
that's kind of the
path that we took.
00:05:45
We've we fell in love with
robotics. We do it more.
00:05:48
We're learning as we're going,
00:05:49
and eventually that leads
to doing exciting work here.
