June 17, 2015 • 41 mins
In early June, teams competed to control a robot that could respond to a simulated emergency situation. What did the robots have to do and who won?
Learn more about your ad-choices at https://www.iheartpodcastnetwork.com
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Brought to you by Toyota. Let's go places. Welcome to
Forward Thinking. Hey there, and welcome to Forward Thinking, the
podcast that looks at the future and says, try to
be a robot king and settled for a robot boy.
I'm Jonathan Strickland, I'm Lauren bolck Obam, and our other
(00:22):
co host, Joe McCormick is not with us today. He's
perfectly fine. He's on an excellent vacation and we'll be
back with us soon, but perhaps even on a on
a on a bogus journey. Thinking of the whole bill
and ted that I caught it. So today we're talking
about robots. Yes, we are sorry. That was so awkward.
(00:45):
I inject awkwardness wherever I go. Oh me too. It's
one of the things that I really pride myself on. Um.
One of the things that robots pride themselves on is
being in affiliation with dart Yeah, or at least DARPA
is prides themselves on their affiliation with robots one or another. Yes,
So we've done a bunch of episodes about DARPA and
(01:08):
their connections to robots the past. Um. I mean, we
mentioned them all the time, but just this May, we
did a whole episode about how the organization works, end
of the projects that they're sponsoring right now, and uh
that that one was called What's Up with DARPA if
you would like to check it out. And back in November,
we did an episode about one of their other robot challenges,
(01:30):
the Grand Challenge, which is an autonomous car obstacle course
sort of thing, and that one's called robot you Can
Drive My Car? Yeah, so, uh, and we'll talk more
a little bit about the Grand Challenge towards the end
to kind of tie in what the more recent DARPA
Challenge it was all about, because as of the recording
(01:52):
of this podcast, we're recording it on June nine, it's
just after the most recent DARPA Challenge finals concluded. Yes,
the I mean DARPA, of course, being the Defense Advanced
Research Projects Agency. I am never going to get used
to saying that out loud. It's so horrible. That's why. Yes,
(02:13):
DARPA's Robotics Challenge, or the robo Olympics, as my friend
Laslow has called it, that's a fitting term. It is
because because they have to do so much, they do
they well, so I guess would be an octathlon. Octathlon
tasks for for walking robots. Yeah, that are beautiful and
(02:35):
fearsome and fall down so much. Yes, there they are
glorious and fearsome and vulnerable. It's like a Pixar character.
They are. They are so So what the challenge is
the Robotics Challenges is it's a competition for teams from
around the world to build these savvy, adaptable robots that
(02:55):
are capable of helping with disaster relief related tasks. It
was announced and has been since then going through various stages.
First a virtual challenge to guide a simulated robot through
three tasks in a virtual environment, and then physical trials
and now the finals. There were prizes for the finals
two million for the winner, one million for the first
(03:18):
runner up, and five hundred thousand for the next runners. Yeah. Yeah,
so so we're talking some serious money and this is
typical of DARPA with their challenges, and you could you know,
that does sound like it's a ton of money, and
it is, I mean, not to put not to dismiss
it is a lot of money, but some of these
teams are investing lots into building these robos. Oh yeah,
(03:40):
way more than that. Much money. Yeah, So really, the
the challenge is exactly what it sounds like. It is
a challenge to engineers to develop machines capable of overcoming
or performing some sort of task in uh, usually some
trying circumstances. Yes, right, So, because the whole thing was
(04:00):
spurred by the Fukushima disaster back in afterwards, a lot
of officials started realizing that, you know, there are these
situations in disaster relief which put human lives in danger,
not just of the victims of the disaster, of course,
but the but the human responders. Right sure, and you know,
we live in this incredible future and can make hardy
(04:24):
robots that can go into places or if we don't
quite live in that incredible future yet, DARPA was like, hey,
maybe we should build that incredible future. It sounds like
a nice place to be. So when did they announce
this specifically, that was April first, which seems like a
really bad day to announce things if you ask me. Yeah,
just imagine if at the conclusion this past weekend, when
(04:47):
the winning team steps forward and DARPA's like, here's your
two million psych April fools. Three years later, that would
be that would be a great April Fools. That's pretty
epic April Fool's joke. One that didn't not sure that
Korea has a pro fools and so spoiler alert the
Korean team one. So I don't know that it would
have gone over so well. But yeah, so so the
(05:10):
the This is exactly what you were saying, Lauren. It
was an attempt to say, here are some situations that
we might need to have a robot be able to
respond to. Here are the various scenarios that the robot
would encounter. Your job is to go out and build
a robot that can complete these tasks under this scenario,
(05:32):
or at the very least, to build software for a
pre existing robot. Right, yes, so you didn't have to
go out and build a brand new robot from scratch.
You could design the software, the brains, the programming that
would allow the robot to complete the tasks. So as
long as the robots form factor already allowed it to
do each of the tasks that were assigned, then you know,
(05:52):
you didn't have to go out and build brand new hardware. Yeah. Yeah,
nothing from from scratch if you didn't want to. And
in fact, a lot of the teams have used Boston
Dynamics at Liss Robot. Boston Dynamics, of course, being the
uh eccentric company uh that that builds uh robo dogs,
Big Dog and they and they kick adorable robotic dogs that.
(06:13):
Oh I just thought about that again and it's still upsets.
The video left a lasting impression on me. I know,
I know the robot can't feel anything, right, I don't
know that. I don't know that, but it's still is
hard to watch um for me at any rate. So
let's talk about Let's talk about the stages. You mentioned
(06:34):
that the first one was a virtual challenge. What was
that all about? Yeah, so this virtual challenge took place
back in June, and twenty six teams built software to
control a virtual Atlas robot through three disaster related tasks
getting into a vehicle, driving it, and getting out that
that's a single one, Yeah, which, by the way, sounds
(06:56):
really basic, but when you think about it, it's natural
for us, but not natural for a robot. I've also
I can't do that all the time without like bumping
my knees and falling down, or I stand up before
my head is clear of the car, like once a week. Man,
if you ever wonder why I'm a little bit fuzzy sometimes?
Do you ever wonder why I'm bleeding from the scalp
(07:17):
because I don't have hair, So it's just there. That's
that's why, all right. So so that was the first
task at then walking through an area of uneven ground
that had mud and debris to contend with a right
that the ground didn't have the mudd and debris to
contend with the robot, yes, um. And then the third
was attaching a hose to a spigott and turning the
(07:37):
valve alright, So these are things that it might have
to do in a response situation, for example, putting out
a fire in an area that had suffered earthquake damage
or something. Yeah. To make it a little bit more fun,
the teams also had to deal with simulated communication lags
and limits, like like limits to the amount of data
(07:58):
that they could transfer back and forth to the robot
and uh a lag time of of maybe like five
dred milliseconds or something like, which again makes sense because
in a disaster environment some of the infrastructure has probably
suffered damage as well, so your ability to communicate could
be limited. Um. In the end, they were judged on
how well the simulated robots completed the simulated tasks and
(08:21):
how much control the operators commanded. And from that UH
one team team I h m C. That's the Institute
for Human and Machine Cognition out of PENNSI Colla, Florida
one and today, along with eight other teams, qualified for
the trials in December. All right, so that moves us
(08:43):
on to those trials. Yes, a couple of the pairs
of teams that made it through merged. In the end,
six teams went on to compete in the trials. That
was at Homestead Miami Speedway in Florida on December one,
and the robots were set eight tasks. They had to
(09:03):
man a vehicle and I wish that there was a
better word for that than robot a vehicle, Yeah, that thing. Um.
They had to navigate uneven terrain. They had to climb
a ladder, remove debris from a door, and then open
the door, breakthrough a wall like kool aid Man. They
did not, however, have to yell oh yeah, I really
hope that they all did though too, okay, and handle
(09:27):
some valves and again attached a host with spigott. So
the winning team was called Shaft sc H A F T,
which I love. Uh they're they're they're out of Japan
and they had just been bought by Google prior to
the trials. Now, wait a minute, I'm sure you're gonna
get to this, but I don't remember seeing that team
(09:48):
name when I was looking over the finals. Story about
that just a moment, Yes, yes, So Shaft really just
looked robo butt. In these trials, of thirty two possible points,
they earned twenty seven, which was an entire seven more
points than even their closest runner up, which was our
friends I H n C Robotics. Third place, at nine
(10:10):
points behind Shaft was a team called the Tartan Rescue
out of Carnegie Melon. Now I've heard about that one.
So what happened with the Google acquisition of this Japanese company? Uh? Well,
in June, Google announced that Shaft would be dropping out
of the finals in order to pursue work on a
(10:32):
commercial Google project. So, in other words, they said, you
know this, this research work is fantastic, it's great, but
we need to dedicate these people to building something that's
going to become a real world product and not a
proof of concept kind of tasting out things in a
scenario environment, and I suspect it also had something to
(10:53):
do with Google's prior decisions to stay out of military
related projects, which they have not infrequently said, you know,
like thanks guys, but no thanks, We're doing good not
funding military. We've seen we've seen it go the other
way where people from projects that were part of the
military ended up at Google later. But that's different. Yes, absolutely,
(11:14):
so they dropped out in June, and the finals took
place this past weekend as of when we're recording this
podcast on June five and six. Yeah, over in California
that they had a dartbed had a contest as well
for high school students to create a video explaining what
they how they felt robots fit into the future of culture, society,
(11:38):
how they were going to integrate, and how they would
be an important part of our lives. And the winners
got to go and witness the finals. So that's kind
of interesting. There was a big audience there there there
have been at the past couple I don't I'm not
sure if there was an audience for the virtual trials,
but definitely for the preliminary the physical trials. Yeah. Yeah,
(12:00):
what's interesting also is that if you watch the videos
from this, I mean that crowd is not shy, they're psyched.
It's so beautiful. It's like a sporting event. Yeah, it
really is. And spoiler alert, a lot of these robots
fell down a lot, and hearing the crowd react to
the robots falling is heartbreaking. Is they're they're also invested in.
They're not like laughing at the robots, which honestly I
(12:22):
tend to do when I'm just sitting on my computer.
Especially you see, like the animated gifts and like the
anime gifts make it look hilarious. Like there's a robot
that it takes a walk, it takes one step forward,
immediately collapses and and the gift makes it hilarious. But
you know when you kind of way, yeah, just like
it's kind of like like, I mean, it's it's that
(12:43):
ultimate in being taken down a peg right where you're right,
we're ready to compete, one step falls apart. Like it's
it's like almost like a national Lampoon vacation kind of moment.
You figured the Griswolds where the team behind Yeah, yeah,
but but the but the audience who was there? What
would own? Yeah, every every time a robot fell down,
(13:05):
like because everyone wanted everything to go Obviously, the the
performance here meant pushing the field of robotics forwards. So
it was more than just cheering on your favorite It
was about cheering on the the advancement of robotics as
a discipline. Yeah. And and these tasks were not easy
(13:27):
tasks that were set to the robots. They were they
were challenging. Yeah. They Now when we start saying them,
they're gonna sound easy because you're going to be thinking
about it in the terms of a human being completing
these tasks. And they're very simple. But keep in mind
that all these things are tasks that robots have to
be engineered to do. And only that, but the robot
has to be engineered to do all of these tasks root. Yeah,
(13:51):
and you think about most robots that you would see
in your day to day life, Like if you have
a rumba, that rumba is designed to clean the floor.
It's not designed to clean the floor plus do the dishes,
plus or watch your kids, or walk your dog or whatever.
It can't do those things and you wouldn't expect it to.
Yet these are robots that have to complete these eight
(14:12):
different tasks and have one robot to do it. Oh
yeah yeah, and with minimal supervision from the from the team,
right yeah, because again your communication is limited. And the
idea was build a robot that can autonomously do as
much of this as possible, not not expecting any robot
to be like the Terminator and not need any further
(14:33):
direction like go that way and save the people, like
there's nothing like that, but limited, you know, the more
limited than the human interaction, the better. So Task number
one drive. So so here's how this worked. They had
to drive a Polaris Ranger XP nine hundred, which is
essentially like a little not quite an altering vehicle, but
(14:56):
similar STV kind of thing. Yeah, not even in su
be more like a little almost like think of like
a souped up golf cart, so a little bit more
advanced than the golf cart. Um, and the teams had
five minutes to alter the vehicle without the use of tools,
so that their robots could operate it. So if the
robot was at a form factor that did not easily
(15:17):
fit into this, they could alter the vehicle, but they
couldn't use tools to do it. So this made me
think of like cutthroat kitchen. I'm picturing robotics teams just
like smashing their hands into panics to reform them or something.
You can alter this vehicle, but you can only use
gummy worms. Yeah, it is a little bit of a
of a fun twist on that. So the way the
(15:40):
challenge would start is the robots would already be behind
the wheel. They did not have to get into the vehicle.
They could be placed in the vehicle. The vehicle was
already running and in high gear. They said that they
chose high gear because it would move the most smoothly
out of the gears. But the teams were welcome to
design a robot that could shift gears if need be,
(16:01):
but they didn't have to. Uh. So then the robot
had to drive the vehicle from the starting position to
a different position through an obstacle course, not really an
obstacle course, but a driving course. Uh. If the robots
was to uh collide with one of the barriers and
move it, then it would not be eligible for a
(16:22):
point for the task. And we should say these eight
tasks represent eight points. So teams were judged on two criteria,
the number of points that they were able to accumulate
and the time it took the robot to complete the
series of tasks. So if you were able to do
all eight and you had the lowest time, you win.
Uh So if you if you ran ran to an obstacle,
(16:46):
you've got no point. You could also choose for your
robot to walk from the starting point to the next task,
but then you also time penalty share. Yeah, time penalty,
and you don't get a point because the robots not
operating the vehicle. Um. So really the robot bare minimum
had to be able to operate the accelerator and the
steering wheel, although you would hope it could also operate
(17:06):
the brake um because otherwise the robot would just be
easing off the accelerator so that it would essentially coast
into the final position. And then it had to uh
to get completely past a cut off point that said,
you know here, once the vehicle has passed this point,
you have completed the task without going so far as
(17:28):
to smash into the next obstacle. All right, So that's
that was the first task. Second task was called egress,
which I love. It's that's terrific vocabulary. Yeah, it's it's
it's a little nicer than get out of the DIRN car.
So that that's the southern version, right. So egress is
(17:50):
essentially what I just said. The robot had to get
itself out of the vehicle. It could exit out of
either side, so it was not required to exit from
one side versus the other. It had to be able
to stand up or otherwise get into its you know,
mobile format, because you didn't have to have a bipedal robot.
Most of the designs were, but you didn't have to
(18:11):
design one that way. It's just that when you're talking
about operating stuff that humans operate bipedal, since that's how
we are, that tends to beat. Yeah. So then it
had to move itself to the next uh task position.
So the whole judging, the whole thing is get is
allowing itself to get out and walk over to the
(18:32):
next point, and that would be another You would get
a point for the egress part of the task if
you were able to do that, or rather if your
robot were able to do that. After that was door. Okay,
so here's one of those things that sounds incredibly simple.
The robot had to open a door and walk through
the doorway. That's it that's the challenge. So so drive
(18:57):
a car is on equal playing field with open and
walk through it through a door. Yeah. However, spoiler alert,
this was one of the challenges that was the most
difficult for the robots. Uh So. It was a push door,
and it was a it was a central, a lever
handled door. It kind of like the one we have
(19:17):
to the studio where you have to either push down
or lift up on the levers and then push the
door to simultaneously push. There was no threshold, so the
robot didn't have to step over anything. That's very kind. Still,
a lot of robots fell over in this point. And
part of the reason why the robots were having so
much trouble because that the doors with when you include
(19:37):
the door jam, was only thirty three and a half inches,
and some of these robots were wider than that. So
they had to turn themselves, they had to pivot in,
or they had to fold their limbs inward and then
just kind of, you know, wiggle on through. I've got
all these great images of robots attempting to get through
a doorway. So yeah, that's the thing is that, in fact,
(19:59):
it prompted one guy in the audience to say, when
the robot uprising inevitably happens, just close your door and
you'll probably be safe. If you have a door and
you're at the top of a set of stairs, you're
doubly safe. Um. So, next was Valve, which is where
a robot had to buy a game on Steam. Okay, yeah,
(20:20):
preferably something like uh, well there's the new Fallout, so
pre ordering Fallout. Oh yeah yeah, actually no, no, it's
it was actually physical Valve. It was not not the
not the game company Valve. It was a terrible joke
that I wrote in our notes. Actually in there, I
was willing to go with it. You know, I think
that I think that navigating complex internet software is really
(20:43):
part of your future of robots. Yeah, you know where
your robot could just sit at home and buy all
those games for you. Yeah, exactly. I mean you're not
gonna play him anyway, You're just buying them because they're
on sale. Right, Really hate putting my credit card information
into Yeah, this is going to be valuable. You just
do it once with the robot and then you're done,
maybe less with maybe less with disaster situations. So in
this case, in this case, they were operating an actual valve.
(21:07):
It was a circular handle. They had to turn in
a counter clockwise direction at least three sixty degrees that
would be considering opening the valve. So this was another
one that was a little tough. Was you know, design
a robot that can grip a circular valve and then
turn it in this way. Also, the teams only knew
that the valves size would be somewhere between four and
(21:29):
sixteen inches in diameter, so they couldn't a gap. Yeah,
they couldn't just plan in advance, like all right, well,
we'll just design the robots so that it's hands are
you know, are precisely molded to this particular valve, right, Like,
let's just assume it's twelve inches. It'll be fine. That's
not necessarily the case. The next step was probably my
favorite out of all of them. Oh this is this
(21:51):
is the kool aid Man challenge. Yeah. Wall. This was wall,
but it wasn't just breaking through a wall, which is
the way it was always described to me when I
heard about, well it's really cutting through a wall, and
and that's and I and I said breaking through in
the in the previous trial challenge. But but it's really
cutting through a wall using a drill actually, and the
wall was dry walls half inch thick drywall, and there
(22:15):
was a shape that you your robot had to cut
the shape out of the wall, so I think it
was like a giant circle, at least the one I
saw it was a giant circle. Where the idea was
that the robot had to go pick up a drill.
The drill was an off position, so I had to
squeeze the trigger to have it go on. Turn it on, yeah,
and then insert the drill uh inside the line of
(22:38):
the shape, trace around that line on the inside, and
remove the rubble. The idea being that there might be
a time where a robot would have to have to
respond to an emergency get access to controls, but the
entrance to the control room is otherwise blocked, so this
would be a way of getting access to stuff without
being without having to go through the doorway. So it's
(23:00):
kind of a cool thing, and the pictures and video
from this section are really interesting. One robot got a
little over excited, dropped the drill, then immediately turned around
and walked to the next task, gave up. There were
two drills by the way, so if you dropped one,
you could pick up the other one you being the
robot um. But yeah, that was my favorite. Uh. The
(23:20):
next one was just called surprise, And this was one
of those things DARPA like DARPA told all the teams,
there will be a challenge that your robot will have
to do, but we're not going to tell you what
it is. Yeah, oh yeah, and if we didn't mention
it earlier, the teams had all of these tasks and
specifications for all of these tasks, like rough specifications way
(23:44):
in advance, like at least a year in advantage. They
had to know what the robot was going to have
to be capable of doing in order to engineer the robot.
Really of course, of course, so surprise what what wound
up being our surprise? Well, at least on one day,
and I have to look into it to find out
if it was the same for both, because they said
that they had the option of changing out the surprise
task each day, so that way, no to make it
(24:07):
more fair, So the teams that go on day two
don't have an advantage over the ones that day one.
But at least on one day it was to pick
up a plug and plug it into an electric outlet,
which apparently one robot tried like for fourteen minutes and
then just gave up one of those things. Again. I
just have you ever seen the video of the one
little robot arm that picks up like I think it's
(24:29):
like it's supposed to be a little piece of food
and it's supposed to put it at the mannequin's mouth,
but just starts slamming the fist into the mannez. That's
what I imagine. Um, But you know, so some of
the robots had trouble with this task of plugging, And honestly,
I've had issues where I have been unable to plug,
like my computer into an outlet. I can't blame the robots. Yeah, no,
(24:50):
not at all. Uh. Then we've got we've got two
more left. Yeah, so we've we've got we've got rubble. Yes,
this was just traversing debris or a terrain yield. Uh.
Some people said it either involved walking on top of
the debris or walking through the debris, but the again,
whichever works for you, I suppose. Yeah, and considering the
(25:12):
scenarios um uh inspiration the Fokushima event, than having rubble
to get through is is a realistic possibility. So some
of these robots just would walk over them, like the
Boston Dynamics approach is to maintain balance while walking over
uncertain terrain. Some of them were designed so that they
could go onto four limbs and they had like treads
(25:35):
on four limbs, I could roll over it. In fact,
the Carnegie Melon one had that, and the the Um treadmills.
The treads rather were essentially on the shins and foe
arms of this robot, so not on the hands and
feet on so it would get down on its shins
and its foe arms and roll across. Uh. So that
(25:57):
was kind of cool, fascinating. The last one was a
set of stairs that had a rail on the left side,
no rail on the right. And as we all know
from Dr Who not that the doaleks are robots, but
they have robotic elements yea, and robots hate stairs. Robots
hate stairs. You might have heard about Asimo, which is
capable of walking up and downstairs, but only if you
(26:17):
program exactly how many stairs are in the sequence and
how high they are. It is not good at just
you can't just put it in front of any staircase
and be like, go suck at yeah and be like
I'll see you on the top, and it doesn't happen
that way. So stairs are tricky, and so you know,
you hear a lot of crowds, Like if you watch
the last few minutes of any of these runs where
(26:39):
the robots are climbing the stairs, you hear the crowd
going nuts as the robot very slowly and carefully traverses
the stairs, like like there's a four minute video of
the of the winning team just slowly going four minute
might be a little much, but it's you seeing it
slowly going up each step and everyone the crowd is
(26:59):
just going the entire uh. And then, of course, because
this is a darker challenge, there was a twist. Yeah,
it's like a m night shya milan. Yeah. Yeah, you know,
you gotta gotta keep those robots on their proverbial or
literal toes as the case. Maybe. So this year, each
of the eight challenges had been practiced with a time
limit of thirty minutes per challenge, but on the day
(27:22):
of the competition, the robots were given an hour to
complete all eight. That's incredible together, so ha, that like
gets my blood pressure going, like just thinking about what
those robotics teams on site must have been feeling. Yeah.
I mean you're talking about you thought you were going
to have four hours and you have one hour. That's
(27:44):
a big change. So let's talk about the folks who
actually won. Uh So, Like I said that, the first
place team came from Korea, South Korea specifically, and that
was Team k A I S T KIST KIST and
they had DRC Hubo as their robot. The winning time
(28:04):
was Remember they had one hour to complete all these
four minutes twenty eight seconds. They did complete all eight tasks,
so uh more than fifteen minutes to spare. And like
I said, I was watching the this one and the
crowd went totally bonkers. To keep in mind, now, they
were not on the first day's events, so at the
(28:26):
end of the first day there was a different team
that was in the lead. So this was one of
those things where you yeah, um. And there's also a
preparation video for this particular robot that was hilarious. Did
you see this. It's like watching one of the movies
from the Rocky franchise because the robot is like in
(28:46):
a training montage. Uh So, it's showing the team working
on various things for the robot to do, like the
kind of precise movements that needs to do to do
things like turn valves to plug a hose into evalve,
that kind of stuff. And it also showed the robot
doing push ups to show the robot doing a fighting pose. Yeah,
(29:08):
something like like okay, this is this is cute. Also
doing stuff like climbing a ladder. It climbed the ladder backwards,
like like the part of the torso you would identify
as the front. Keep in mind that robots don't necessarily
have a front and back. That could be it's only
it's only what you have told it is. But to them,
it looks like it's backing up, and it's the way
(29:30):
it's arms bend. It looks like it's backwards because elbows
are going up and it's lifting itself up backwards up
a ladder. I suppose, I suppose that the balance spots.
That makes way more sense. We should all be going
up ladders backwards. Really, when you think about it now,
and this ladder, I should say it's more like a
step ladder than not a ladder ladder, not a rung ladder,
more like a step ladder. So I had a little
rail and everything so the robot could hold on to
(29:53):
the sides, so they won first. Second place was our
friends team I h M C out of Pensacola. They
were using an altered Boston Dynamics Atlas bought called Running Man,
and upon completing the course, it struck this amazing victory
post at the top of the stairs and then promptly
fell right over like it also completed all eight tasks
(30:17):
and did so in fifty minutes twenty six seconds, so
it was just about six minutes slower than the other
one in a competition. Third place was Carnegie Melon's Team
Tart and Rescue and it's robot Chimp. So Chimp is
a robot that has really long arms. Oh that one okay.
It's also the one that had the treads on its
(30:39):
forearms and shins. Now, Chimp was able to complete all
eight tasks in fifty five minutes fifteen seconds. Chimp was
the leading contender at the end of day one. Oh
Chimp also had fallen over and was the only robot
to fall over and right itself. All the other robots
had to have the team's come over and look at
(31:02):
the robot and try and get it back up on
its feet. Yeah, it's it's it's legopendages its ponds. Yeah.
So but but you know, chimp was in a shoot
all help and got up on itself. And I watched
the video of this, and this is freaky to watch
the video because it's essentially twisting its legs and in
(31:25):
ways that organic lifeform legs do not twist, and using
those treads to help get it back up to a
four m base. Yeah. Yeah, that that one, um, that
that would kind of creeped me out. It's it's a
little bit creepy looking. I think it's reminded me of
like maybe like the Wheelies from Return to Oz or something. Yeah,
(31:45):
I don't know, I can see that. So those three
teams were also the only three to complete all eight tests.
None of the other teams managed to do that. Four
teams completed seven, one team completed six tasks, two teams
complete five tasks, two teams completed four tasks, four teams
completed three tasks to complete to one completed one, and
(32:08):
then four teams didn't complete any successfully. Yeah. But they
it shows how like even even as we said these challenges,
like you clemoss stairs and for most people, that's not
a big issue. You open a door for most people,
not a big issue. Um, the walling across debris could
be a little tricky depending upon the type of debris.
But a lot of these are tasks that most people
(32:31):
would find incredibly simple. But it demonstrates how these simple
tasks are really tough engineering challenges. Like when you figured
that only three teams were able to successfully complete all
eight and only just under an hour, right, And these
are yeah, sensibly some of the some of the best
teams in the world in robotics right now, incredibly talented,
(32:53):
intelligent and and motivated people. So the cool thing is
that it shows us while these tasks have been really difficult,
they are ones that can actually be achieved through robotics,
and um, it's it's pretty pretty cool stuff. The whole
point was to show that robotics could be pushed to
(33:17):
completely specific tests, and it was taking them out of
the lab. That's one of the big one of the
big criticisms about technology in general is that we often
see it in the R and D phase in the laboratory,
where it's a very controlled environment. And in that controlled environment,
of course, you would expect success rates to be much
higher because you have control over everything. But but most
(33:41):
robots are not going to be in laboratories if they're
ever going to be useful. Yeah. Yeah, For if we're
talking about a first respond to robot, it has to
be able to adapt and respond to real world situations.
And this is what that challenge was really about, was
how well could you design a robot that could, uh,
you know, you know ahead of time what it's going
(34:02):
to have to do. And to some extent, that's going
to be the truth for at least the foreseeable future
where robots. Robots are going to be designed to do
specific things and anything outside of that they may or
may not be able to do based upon the stuff
you've programmed into that robot. But you know, it all
depends upon the design and programming of the robot with
(34:25):
foresight thinking these are the things this robot is typically
going to encounter, so here's how the robot needs to
be able to respond, and anything that goes outside of
that is really hard to design in a robot right now. Yeah, Yeah,
it's that old old problem of how do you tell
how do you tell a machine to specifically do something
when you don't know what that thing is. Yeah, and
(34:48):
so the special you know, surprise was an example that
it was. It was a variation on something that the
robots were already doing. So it wasn't like it was
completely lf field. It wasn't like, uh, you know, um,
you've you've got a cook an omelet, wasn't something like that,
you know it right, yeah, go buy games on Steam
(35:11):
and wasn't anything like that. So uh, you know, but
even then you see that some robots had real problems
with that. But this does mean that we are pushing
the field of robotics further as a result, And I
think that by identifying specific challenges, it gives engineers the
focus they need to be able to design the systems
(35:33):
that are capable of meeting those challenges. Right, So instead
of just having a wide open field where you say,
there's going to be an emergency situation, design a robot
to respond to it without any more direction. That's hard
to do. So the other point being that that this
also helped advance the the idea of how much communication
(35:55):
to humans need to be able to have with the
robot for it to be able to complete these challenges.
So a lot of these challenges the DARPA was saying, well,
we expect that some of the robots at least will
be able to respond to a fairly simple command and
carry it out, like go from this point to that point,
don't fall over. Not all the robots were able to
(36:17):
do that, obviously, but the dark but also the organization
I was saying, we don't expect any robot to be
able to go from start to finish with no human
contact contact at all, because to complete a single task
is one thing, but to chain them together in a
logical formation is another. So you wouldn't tell the robot, Hey,
(36:38):
I need you to drive to this point, Get out
of the vehicle, walk up to this door, open it.
Go through the door. You're going to see a wall
with a shape on it. Cut the shape out of
the wall, remove the rubble. Then you're gonna have something
we don't know what it is. Then you're going to
have to like you can't. You can't just tell a
robot that. It's much easier for right to to be
(36:59):
able to put that that input in every once in
a while to go like, oh, here's the part where
they're stairs, exactly, this is the stairs part. And because
otherwise it's trying to climb the vehicle and it's just
not you know, yeah, it's identified you know, a little
Timmy as the first step, and that's not going to
go well, uh, at least not for Timmy. So the
(37:20):
comparison I wanted to draw here was one of the
earlier DARPA challenges, the Grand Challenge. So with the Grand Challenge,
we saw all these teams trying to build automas cars
that could go through various obstacle courses. The first one
was a desert course. Then they had a simulated urban
environment complete with you know, traffic and other things estrians
(37:42):
that the cars had to contend with. And we saw
simulated pedestrians. Yes, yes, we saw no let me no
little Timmy, We're gonna need another Timmy. I think that
was from Dinosaurs, the television series. Um. But the uh,
the takeaway I wanted to give was that while that
(38:04):
wasn't intended to make automous cars reality the very next day,
what it did was it pushed the discipline forward, literally
pushed it forward, and we saw lots of people from
the various winning teams go into other organizations, including Google,
to develop the autonomous cars that are on the brink
of coming out on the consumer market. Now, within the
(38:27):
next few years, we we will probably see the first
steps and they'll probably be limited it first. But it's
because of that grand challenge that we're as far along
as we are. Sure. Yeah, and all these challenges, of
course also just raise public interest and excitement in the fields.
And you know, watching video of families that were out
(38:49):
watching the finals was so amazing, like getting to watch
the looks on kids faces and and yeah, and hearing
the cheers and the and the gasps from the crowd
whenever the robots were doing anything. It really does seem
to serve as an inspiration for people to get into
the stem part of education, right, to look into science, technology, engineering,
(39:09):
and uh, mathematics. Yeah. I was like, what's that fourth one?
That that one that we're all terrible at in this
podcast room English lit. Major If you listen to that
other episode where I suddenly reverted to my primal self.
Uh yeah. This this is one of those things where
you really you've got to admire it from multiple reasons. Um,
(39:30):
the idea of we're trying to solve a real world problem.
What happens in the face of an event like Fukushima.
How can we design something that could potentially save countless lives? Also,
how can we inspire the next generation of engineers to
want to get into that, to do that, and to
go beyond when we've even envisioned so far. So it's
(39:52):
really been great to look into this. I'm glad we
were able to do an episode about it and talk
about all the fun and foibles of robots, um whether
they're upright or falling over. Uh. It was really a
lot of fun to talk about this. So if you
guys have suggestions for future topics that we can cover
(40:12):
here on Forward Thinking, please write in and let us
know we love hearing from you. You guys have been
sending us some great ones, and we'll be doing another
listener mail episode before too long because we we accumulate
those and some of them, some of them are short
enough where we can do a couple in a single episode.
And then some of you guys have really complicated questions
that you want answered, and we're going to tackle those
(40:34):
as well. But if you have one you haven't voiced
it yet, send us an email FW thinking at how
stuff works dot com, or drop us a line on Twitter,
Google Plus, or Facebook. At Twitter and Google Plus, we
are f w thinking at Facebook. Just search fw thinking
in the search bar, will pop right up and leave
us a message. We'll talk to you again really soon.
(41:00):
For more on this topic and the future of technology,
visit forward thinking dot com, brought to you by Toyota.
Let's Go Places,
Brought to you by Toyota. Let's go places. Welcome to
Forward Thinking. Hey there, and welcome to Forward Thinking, the
podcast that looks at the future and says, try to
be a robot king and settled for a robot boy.
I'm Jonathan Strickland, I'm Lauren bolck Obam, and our other
(00:22):
co host, Joe McCormick is not with us today. He's
perfectly fine. He's on an excellent vacation and we'll be
back with us soon, but perhaps even on a on
a on a bogus journey. Thinking of the whole bill
and ted that I caught it. So today we're talking
about robots. Yes, we are sorry. That was so awkward.
(00:45):
I inject awkwardness wherever I go. Oh me too. It's
one of the things that I really pride myself on. Um.
One of the things that robots pride themselves on is
being in affiliation with dart Yeah, or at least DARPA
is prides themselves on their affiliation with robots one or another. Yes,
So we've done a bunch of episodes about DARPA and
(01:08):
their connections to robots the past. Um. I mean, we
mentioned them all the time, but just this May, we
did a whole episode about how the organization works, end
of the projects that they're sponsoring right now, and uh
that that one was called What's Up with DARPA if
you would like to check it out. And back in November,
we did an episode about one of their other robot challenges,
(01:30):
the Grand Challenge, which is an autonomous car obstacle course
sort of thing, and that one's called robot you Can
Drive My Car? Yeah, so, uh, and we'll talk more
a little bit about the Grand Challenge towards the end
to kind of tie in what the more recent DARPA
Challenge it was all about, because as of the recording
(01:52):
of this podcast, we're recording it on June nine, it's
just after the most recent DARPA Challenge finals concluded. Yes,
the I mean DARPA, of course, being the Defense Advanced
Research Projects Agency. I am never going to get used
to saying that out loud. It's so horrible. That's why. Yes,
(02:13):
DARPA's Robotics Challenge, or the robo Olympics, as my friend
Laslow has called it, that's a fitting term. It is
because because they have to do so much, they do
they well, so I guess would be an octathlon. Octathlon
tasks for for walking robots. Yeah, that are beautiful and
(02:35):
fearsome and fall down so much. Yes, there they are
glorious and fearsome and vulnerable. It's like a Pixar character.
They are. They are so So what the challenge is
the Robotics Challenges is it's a competition for teams from
around the world to build these savvy, adaptable robots that
(02:55):
are capable of helping with disaster relief related tasks. It
was announced and has been since then going through various stages.
First a virtual challenge to guide a simulated robot through
three tasks in a virtual environment, and then physical trials
and now the finals. There were prizes for the finals
two million for the winner, one million for the first
(03:18):
runner up, and five hundred thousand for the next runners. Yeah. Yeah,
so so we're talking some serious money and this is
typical of DARPA with their challenges, and you could you know,
that does sound like it's a ton of money, and
it is, I mean, not to put not to dismiss
it is a lot of money, but some of these
teams are investing lots into building these robos. Oh yeah,
(03:40):
way more than that. Much money. Yeah, So really, the
the challenge is exactly what it sounds like. It is
a challenge to engineers to develop machines capable of overcoming
or performing some sort of task in uh, usually some
trying circumstances. Yes, right, So, because the whole thing was
(04:00):
spurred by the Fukushima disaster back in afterwards, a lot
of officials started realizing that, you know, there are these
situations in disaster relief which put human lives in danger,
not just of the victims of the disaster, of course,
but the but the human responders. Right sure, and you know,
we live in this incredible future and can make hardy
(04:24):
robots that can go into places or if we don't
quite live in that incredible future yet, DARPA was like, hey,
maybe we should build that incredible future. It sounds like
a nice place to be. So when did they announce
this specifically, that was April first, which seems like a
really bad day to announce things if you ask me. Yeah,
just imagine if at the conclusion this past weekend, when
(04:47):
the winning team steps forward and DARPA's like, here's your
two million psych April fools. Three years later, that would
be that would be a great April Fools. That's pretty
epic April Fool's joke. One that didn't not sure that
Korea has a pro fools and so spoiler alert the
Korean team one. So I don't know that it would
have gone over so well. But yeah, so so the
(05:10):
the This is exactly what you were saying, Lauren. It
was an attempt to say, here are some situations that
we might need to have a robot be able to
respond to. Here are the various scenarios that the robot
would encounter. Your job is to go out and build
a robot that can complete these tasks under this scenario,
(05:32):
or at the very least, to build software for a
pre existing robot. Right, yes, so you didn't have to
go out and build a brand new robot from scratch.
You could design the software, the brains, the programming that
would allow the robot to complete the tasks. So as
long as the robots form factor already allowed it to
do each of the tasks that were assigned, then you know,
(05:52):
you didn't have to go out and build brand new hardware. Yeah. Yeah,
nothing from from scratch if you didn't want to. And
in fact, a lot of the teams have used Boston
Dynamics at Liss Robot. Boston Dynamics, of course, being the
uh eccentric company uh that that builds uh robo dogs,
Big Dog and they and they kick adorable robotic dogs that.
(06:13):
Oh I just thought about that again and it's still upsets.
The video left a lasting impression on me. I know,
I know the robot can't feel anything, right, I don't
know that. I don't know that, but it's still is
hard to watch um for me at any rate. So
let's talk about Let's talk about the stages. You mentioned
(06:34):
that the first one was a virtual challenge. What was
that all about? Yeah, so this virtual challenge took place
back in June, and twenty six teams built software to
control a virtual Atlas robot through three disaster related tasks
getting into a vehicle, driving it, and getting out that
that's a single one, Yeah, which, by the way, sounds
(06:56):
really basic, but when you think about it, it's natural
for us, but not natural for a robot. I've also
I can't do that all the time without like bumping
my knees and falling down, or I stand up before
my head is clear of the car, like once a week. Man,
if you ever wonder why I'm a little bit fuzzy sometimes?
Do you ever wonder why I'm bleeding from the scalp
(07:17):
because I don't have hair, So it's just there. That's
that's why, all right. So so that was the first
task at then walking through an area of uneven ground
that had mud and debris to contend with a right
that the ground didn't have the mudd and debris to
contend with the robot, yes, um. And then the third
was attaching a hose to a spigott and turning the
(07:37):
valve alright, So these are things that it might have
to do in a response situation, for example, putting out
a fire in an area that had suffered earthquake damage
or something. Yeah. To make it a little bit more fun,
the teams also had to deal with simulated communication lags
and limits, like like limits to the amount of data
(07:58):
that they could transfer back and forth to the robot
and uh a lag time of of maybe like five
dred milliseconds or something like, which again makes sense because
in a disaster environment some of the infrastructure has probably
suffered damage as well, so your ability to communicate could
be limited. Um. In the end, they were judged on
how well the simulated robots completed the simulated tasks and
(08:21):
how much control the operators commanded. And from that UH
one team team I h m C. That's the Institute
for Human and Machine Cognition out of PENNSI Colla, Florida
one and today, along with eight other teams, qualified for
the trials in December. All right, so that moves us
(08:43):
on to those trials. Yes, a couple of the pairs
of teams that made it through merged. In the end,
six teams went on to compete in the trials. That
was at Homestead Miami Speedway in Florida on December one,
and the robots were set eight tasks. They had to
(09:03):
man a vehicle and I wish that there was a
better word for that than robot a vehicle, Yeah, that thing. Um.
They had to navigate uneven terrain. They had to climb
a ladder, remove debris from a door, and then open
the door, breakthrough a wall like kool aid Man. They
did not, however, have to yell oh yeah, I really
hope that they all did though too, okay, and handle
(09:27):
some valves and again attached a host with spigott. So
the winning team was called Shaft sc H A F T,
which I love. Uh they're they're they're out of Japan
and they had just been bought by Google prior to
the trials. Now, wait a minute, I'm sure you're gonna
get to this, but I don't remember seeing that team
(09:48):
name when I was looking over the finals. Story about
that just a moment, Yes, yes, So Shaft really just
looked robo butt. In these trials, of thirty two possible points,
they earned twenty seven, which was an entire seven more
points than even their closest runner up, which was our
friends I H n C Robotics. Third place, at nine
(10:10):
points behind Shaft was a team called the Tartan Rescue
out of Carnegie Melon. Now I've heard about that one.
So what happened with the Google acquisition of this Japanese company? Uh? Well,
in June, Google announced that Shaft would be dropping out
of the finals in order to pursue work on a
(10:32):
commercial Google project. So, in other words, they said, you
know this, this research work is fantastic, it's great, but
we need to dedicate these people to building something that's
going to become a real world product and not a
proof of concept kind of tasting out things in a
scenario environment, and I suspect it also had something to
(10:53):
do with Google's prior decisions to stay out of military
related projects, which they have not infrequently said, you know,
like thanks guys, but no thanks, We're doing good not
funding military. We've seen we've seen it go the other
way where people from projects that were part of the
military ended up at Google later. But that's different. Yes, absolutely,
(11:14):
so they dropped out in June, and the finals took
place this past weekend as of when we're recording this
podcast on June five and six. Yeah, over in California
that they had a dartbed had a contest as well
for high school students to create a video explaining what
they how they felt robots fit into the future of culture, society,
(11:38):
how they were going to integrate, and how they would
be an important part of our lives. And the winners
got to go and witness the finals. So that's kind
of interesting. There was a big audience there there there
have been at the past couple I don't I'm not
sure if there was an audience for the virtual trials,
but definitely for the preliminary the physical trials. Yeah. Yeah,
(12:00):
what's interesting also is that if you watch the videos
from this, I mean that crowd is not shy, they're psyched.
It's so beautiful. It's like a sporting event. Yeah, it
really is. And spoiler alert, a lot of these robots
fell down a lot, and hearing the crowd react to
the robots falling is heartbreaking. Is they're they're also invested in.
They're not like laughing at the robots, which honestly I
(12:22):
tend to do when I'm just sitting on my computer.
Especially you see, like the animated gifts and like the
anime gifts make it look hilarious. Like there's a robot
that it takes a walk, it takes one step forward,
immediately collapses and and the gift makes it hilarious. But
you know when you kind of way, yeah, just like
it's kind of like like, I mean, it's it's that
(12:43):
ultimate in being taken down a peg right where you're right,
we're ready to compete, one step falls apart. Like it's
it's like almost like a national Lampoon vacation kind of moment.
You figured the Griswolds where the team behind Yeah, yeah,
but but the but the audience who was there? What
would own? Yeah, every every time a robot fell down,
(13:05):
like because everyone wanted everything to go Obviously, the the
performance here meant pushing the field of robotics forwards. So
it was more than just cheering on your favorite It
was about cheering on the the advancement of robotics as
a discipline. Yeah. And and these tasks were not easy
(13:27):
tasks that were set to the robots. They were they
were challenging. Yeah. They Now when we start saying them,
they're gonna sound easy because you're going to be thinking
about it in the terms of a human being completing
these tasks. And they're very simple. But keep in mind
that all these things are tasks that robots have to
be engineered to do. And only that, but the robot
has to be engineered to do all of these tasks root. Yeah,
(13:51):
and you think about most robots that you would see
in your day to day life, Like if you have
a rumba, that rumba is designed to clean the floor.
It's not designed to clean the floor plus do the dishes,
plus or watch your kids, or walk your dog or whatever.
It can't do those things and you wouldn't expect it to.
Yet these are robots that have to complete these eight
(14:12):
different tasks and have one robot to do it. Oh
yeah yeah, and with minimal supervision from the from the team,
right yeah, because again your communication is limited. And the
idea was build a robot that can autonomously do as
much of this as possible, not not expecting any robot
to be like the Terminator and not need any further
(14:33):
direction like go that way and save the people, like
there's nothing like that, but limited, you know, the more
limited than the human interaction, the better. So Task number
one drive. So so here's how this worked. They had
to drive a Polaris Ranger XP nine hundred, which is
essentially like a little not quite an altering vehicle, but
(14:56):
similar STV kind of thing. Yeah, not even in su
be more like a little almost like think of like
a souped up golf cart, so a little bit more
advanced than the golf cart. Um, and the teams had
five minutes to alter the vehicle without the use of tools,
so that their robots could operate it. So if the
robot was at a form factor that did not easily
(15:17):
fit into this, they could alter the vehicle, but they
couldn't use tools to do it. So this made me
think of like cutthroat kitchen. I'm picturing robotics teams just
like smashing their hands into panics to reform them or something.
You can alter this vehicle, but you can only use
gummy worms. Yeah, it is a little bit of a
of a fun twist on that. So the way the
(15:40):
challenge would start is the robots would already be behind
the wheel. They did not have to get into the vehicle.
They could be placed in the vehicle. The vehicle was
already running and in high gear. They said that they
chose high gear because it would move the most smoothly
out of the gears. But the teams were welcome to
design a robot that could shift gears if need be,
(16:01):
but they didn't have to. Uh. So then the robot
had to drive the vehicle from the starting position to
a different position through an obstacle course, not really an
obstacle course, but a driving course. Uh. If the robots
was to uh collide with one of the barriers and
move it, then it would not be eligible for a
(16:22):
point for the task. And we should say these eight
tasks represent eight points. So teams were judged on two criteria,
the number of points that they were able to accumulate
and the time it took the robot to complete the
series of tasks. So if you were able to do
all eight and you had the lowest time, you win.
Uh So if you if you ran ran to an obstacle,
(16:46):
you've got no point. You could also choose for your
robot to walk from the starting point to the next task,
but then you also time penalty share. Yeah, time penalty,
and you don't get a point because the robots not
operating the vehicle. Um. So really the robot bare minimum
had to be able to operate the accelerator and the
steering wheel, although you would hope it could also operate
(17:06):
the brake um because otherwise the robot would just be
easing off the accelerator so that it would essentially coast
into the final position. And then it had to uh
to get completely past a cut off point that said,
you know here, once the vehicle has passed this point,
you have completed the task without going so far as
(17:28):
to smash into the next obstacle. All right, So that's
that was the first task. Second task was called egress,
which I love. It's that's terrific vocabulary. Yeah, it's it's
it's a little nicer than get out of the DIRN car.
So that that's the southern version, right. So egress is
(17:50):
essentially what I just said. The robot had to get
itself out of the vehicle. It could exit out of
either side, so it was not required to exit from
one side versus the other. It had to be able
to stand up or otherwise get into its you know,
mobile format, because you didn't have to have a bipedal robot.
Most of the designs were, but you didn't have to
(18:11):
design one that way. It's just that when you're talking
about operating stuff that humans operate bipedal, since that's how
we are, that tends to beat. Yeah. So then it
had to move itself to the next uh task position.
So the whole judging, the whole thing is get is
allowing itself to get out and walk over to the
(18:32):
next point, and that would be another You would get
a point for the egress part of the task if
you were able to do that, or rather if your
robot were able to do that. After that was door. Okay,
so here's one of those things that sounds incredibly simple.
The robot had to open a door and walk through
the doorway. That's it that's the challenge. So so drive
(18:57):
a car is on equal playing field with open and
walk through it through a door. Yeah. However, spoiler alert,
this was one of the challenges that was the most
difficult for the robots. Uh So. It was a push door,
and it was a it was a central, a lever
handled door. It kind of like the one we have
(19:17):
to the studio where you have to either push down
or lift up on the levers and then push the
door to simultaneously push. There was no threshold, so the
robot didn't have to step over anything. That's very kind. Still,
a lot of robots fell over in this point. And
part of the reason why the robots were having so
much trouble because that the doors with when you include
(19:37):
the door jam, was only thirty three and a half inches,
and some of these robots were wider than that. So
they had to turn themselves, they had to pivot in,
or they had to fold their limbs inward and then
just kind of, you know, wiggle on through. I've got
all these great images of robots attempting to get through
a doorway. So yeah, that's the thing is that, in fact,
(19:59):
it prompted one guy in the audience to say, when
the robot uprising inevitably happens, just close your door and
you'll probably be safe. If you have a door and
you're at the top of a set of stairs, you're
doubly safe. Um. So, next was Valve, which is where
a robot had to buy a game on Steam. Okay, yeah,
(20:20):
preferably something like uh, well there's the new Fallout, so
pre ordering Fallout. Oh yeah yeah, actually no, no, it's
it was actually physical Valve. It was not not the
not the game company Valve. It was a terrible joke
that I wrote in our notes. Actually in there, I
was willing to go with it. You know, I think
that I think that navigating complex internet software is really
(20:43):
part of your future of robots. Yeah, you know where
your robot could just sit at home and buy all
those games for you. Yeah, exactly. I mean you're not
gonna play him anyway, You're just buying them because they're
on sale. Right, Really hate putting my credit card information
into Yeah, this is going to be valuable. You just
do it once with the robot and then you're done,
maybe less with maybe less with disaster situations. So in
this case, in this case, they were operating an actual valve.
(21:07):
It was a circular handle. They had to turn in
a counter clockwise direction at least three sixty degrees that
would be considering opening the valve. So this was another
one that was a little tough. Was you know, design
a robot that can grip a circular valve and then
turn it in this way. Also, the teams only knew
that the valves size would be somewhere between four and
(21:29):
sixteen inches in diameter, so they couldn't a gap. Yeah,
they couldn't just plan in advance, like all right, well,
we'll just design the robots so that it's hands are
you know, are precisely molded to this particular valve, right, Like,
let's just assume it's twelve inches. It'll be fine. That's
not necessarily the case. The next step was probably my
favorite out of all of them. Oh this is this
(21:51):
is the kool aid Man challenge. Yeah. Wall. This was wall,
but it wasn't just breaking through a wall, which is
the way it was always described to me when I
heard about, well it's really cutting through a wall, and
and that's and I and I said breaking through in
the in the previous trial challenge. But but it's really
cutting through a wall using a drill actually, and the
wall was dry walls half inch thick drywall, and there
(22:15):
was a shape that you your robot had to cut
the shape out of the wall, so I think it
was like a giant circle, at least the one I
saw it was a giant circle. Where the idea was
that the robot had to go pick up a drill.
The drill was an off position, so I had to
squeeze the trigger to have it go on. Turn it on, yeah,
and then insert the drill uh inside the line of
(22:38):
the shape, trace around that line on the inside, and
remove the rubble. The idea being that there might be
a time where a robot would have to have to
respond to an emergency get access to controls, but the
entrance to the control room is otherwise blocked, so this
would be a way of getting access to stuff without
being without having to go through the doorway. So it's
(23:00):
kind of a cool thing, and the pictures and video
from this section are really interesting. One robot got a
little over excited, dropped the drill, then immediately turned around
and walked to the next task, gave up. There were
two drills by the way, so if you dropped one,
you could pick up the other one you being the
robot um. But yeah, that was my favorite. Uh. The
(23:20):
next one was just called surprise, And this was one
of those things DARPA like DARPA told all the teams,
there will be a challenge that your robot will have
to do, but we're not going to tell you what
it is. Yeah, oh yeah, and if we didn't mention
it earlier, the teams had all of these tasks and
specifications for all of these tasks, like rough specifications way
(23:44):
in advance, like at least a year in advantage. They
had to know what the robot was going to have
to be capable of doing in order to engineer the robot.
Really of course, of course, so surprise what what wound
up being our surprise? Well, at least on one day,
and I have to look into it to find out
if it was the same for both, because they said
that they had the option of changing out the surprise
task each day, so that way, no to make it
(24:07):
more fair, So the teams that go on day two
don't have an advantage over the ones that day one.
But at least on one day it was to pick
up a plug and plug it into an electric outlet,
which apparently one robot tried like for fourteen minutes and
then just gave up one of those things. Again. I
just have you ever seen the video of the one
little robot arm that picks up like I think it's
(24:29):
like it's supposed to be a little piece of food
and it's supposed to put it at the mannequin's mouth,
but just starts slamming the fist into the mannez. That's
what I imagine. Um, But you know, so some of
the robots had trouble with this task of plugging, And honestly,
I've had issues where I have been unable to plug,
like my computer into an outlet. I can't blame the robots. Yeah, no,
(24:50):
not at all. Uh. Then we've got we've got two
more left. Yeah, so we've we've got we've got rubble. Yes,
this was just traversing debris or a terrain yield. Uh.
Some people said it either involved walking on top of
the debris or walking through the debris, but the again,
whichever works for you, I suppose. Yeah, and considering the
(25:12):
scenarios um uh inspiration the Fokushima event, than having rubble
to get through is is a realistic possibility. So some
of these robots just would walk over them, like the
Boston Dynamics approach is to maintain balance while walking over
uncertain terrain. Some of them were designed so that they
could go onto four limbs and they had like treads
(25:35):
on four limbs, I could roll over it. In fact,
the Carnegie Melon one had that, and the the Um treadmills.
The treads rather were essentially on the shins and foe
arms of this robot, so not on the hands and
feet on so it would get down on its shins
and its foe arms and roll across. Uh. So that
(25:57):
was kind of cool, fascinating. The last one was a
set of stairs that had a rail on the left side,
no rail on the right. And as we all know
from Dr Who not that the doaleks are robots, but
they have robotic elements yea, and robots hate stairs. Robots
hate stairs. You might have heard about Asimo, which is
capable of walking up and downstairs, but only if you
(26:17):
program exactly how many stairs are in the sequence and
how high they are. It is not good at just
you can't just put it in front of any staircase
and be like, go suck at yeah and be like
I'll see you on the top, and it doesn't happen
that way. So stairs are tricky, and so you know,
you hear a lot of crowds, Like if you watch
the last few minutes of any of these runs where
(26:39):
the robots are climbing the stairs, you hear the crowd
going nuts as the robot very slowly and carefully traverses
the stairs, like like there's a four minute video of
the of the winning team just slowly going four minute
might be a little much, but it's you seeing it
slowly going up each step and everyone the crowd is
(26:59):
just going the entire uh. And then, of course, because
this is a darker challenge, there was a twist. Yeah,
it's like a m night shya milan. Yeah. Yeah, you know,
you gotta gotta keep those robots on their proverbial or
literal toes as the case. Maybe. So this year, each
of the eight challenges had been practiced with a time
limit of thirty minutes per challenge, but on the day
(27:22):
of the competition, the robots were given an hour to
complete all eight. That's incredible together, so ha, that like
gets my blood pressure going, like just thinking about what
those robotics teams on site must have been feeling. Yeah.
I mean you're talking about you thought you were going
to have four hours and you have one hour. That's
(27:44):
a big change. So let's talk about the folks who
actually won. Uh So, Like I said that, the first
place team came from Korea, South Korea specifically, and that
was Team k A I S T KIST KIST and
they had DRC Hubo as their robot. The winning time
(28:04):
was Remember they had one hour to complete all these
four minutes twenty eight seconds. They did complete all eight tasks,
so uh more than fifteen minutes to spare. And like
I said, I was watching the this one and the
crowd went totally bonkers. To keep in mind, now, they
were not on the first day's events, so at the
(28:26):
end of the first day there was a different team
that was in the lead. So this was one of
those things where you yeah, um. And there's also a
preparation video for this particular robot that was hilarious. Did
you see this. It's like watching one of the movies
from the Rocky franchise because the robot is like in
(28:46):
a training montage. Uh So, it's showing the team working
on various things for the robot to do, like the
kind of precise movements that needs to do to do
things like turn valves to plug a hose into evalve,
that kind of stuff. And it also showed the robot
doing push ups to show the robot doing a fighting pose. Yeah,
(29:08):
something like like okay, this is this is cute. Also
doing stuff like climbing a ladder. It climbed the ladder backwards,
like like the part of the torso you would identify
as the front. Keep in mind that robots don't necessarily
have a front and back. That could be it's only
it's only what you have told it is. But to them,
it looks like it's backing up, and it's the way
(29:30):
it's arms bend. It looks like it's backwards because elbows
are going up and it's lifting itself up backwards up
a ladder. I suppose, I suppose that the balance spots.
That makes way more sense. We should all be going
up ladders backwards. Really, when you think about it now,
and this ladder, I should say it's more like a
step ladder than not a ladder ladder, not a rung ladder,
more like a step ladder. So I had a little
rail and everything so the robot could hold on to
(29:53):
the sides, so they won first. Second place was our
friends team I h M C out of Pensacola. They
were using an altered Boston Dynamics Atlas bought called Running Man,
and upon completing the course, it struck this amazing victory
post at the top of the stairs and then promptly
fell right over like it also completed all eight tasks
(30:17):
and did so in fifty minutes twenty six seconds, so
it was just about six minutes slower than the other
one in a competition. Third place was Carnegie Melon's Team
Tart and Rescue and it's robot Chimp. So Chimp is
a robot that has really long arms. Oh that one okay.
It's also the one that had the treads on its
(30:39):
forearms and shins. Now, Chimp was able to complete all
eight tasks in fifty five minutes fifteen seconds. Chimp was
the leading contender at the end of day one. Oh
Chimp also had fallen over and was the only robot
to fall over and right itself. All the other robots
had to have the team's come over and look at
(31:02):
the robot and try and get it back up on
its feet. Yeah, it's it's it's legopendages its ponds. Yeah.
So but but you know, chimp was in a shoot
all help and got up on itself. And I watched
the video of this, and this is freaky to watch
the video because it's essentially twisting its legs and in
(31:25):
ways that organic lifeform legs do not twist, and using
those treads to help get it back up to a
four m base. Yeah. Yeah, that that one, um, that
that would kind of creeped me out. It's it's a
little bit creepy looking. I think it's reminded me of
like maybe like the Wheelies from Return to Oz or something. Yeah,
(31:45):
I don't know, I can see that. So those three
teams were also the only three to complete all eight tests.
None of the other teams managed to do that. Four
teams completed seven, one team completed six tasks, two teams
complete five tasks, two teams completed four tasks, four teams
completed three tasks to complete to one completed one, and
(32:08):
then four teams didn't complete any successfully. Yeah. But they
it shows how like even even as we said these challenges,
like you clemoss stairs and for most people, that's not
a big issue. You open a door for most people,
not a big issue. Um, the walling across debris could
be a little tricky depending upon the type of debris.
But a lot of these are tasks that most people
(32:31):
would find incredibly simple. But it demonstrates how these simple
tasks are really tough engineering challenges. Like when you figured
that only three teams were able to successfully complete all
eight and only just under an hour, right, And these
are yeah, sensibly some of the some of the best
teams in the world in robotics right now, incredibly talented,
(32:53):
intelligent and and motivated people. So the cool thing is
that it shows us while these tasks have been really difficult,
they are ones that can actually be achieved through robotics,
and um, it's it's pretty pretty cool stuff. The whole
point was to show that robotics could be pushed to
(33:17):
completely specific tests, and it was taking them out of
the lab. That's one of the big one of the
big criticisms about technology in general is that we often
see it in the R and D phase in the laboratory,
where it's a very controlled environment. And in that controlled environment,
of course, you would expect success rates to be much
higher because you have control over everything. But but most
(33:41):
robots are not going to be in laboratories if they're
ever going to be useful. Yeah. Yeah, For if we're
talking about a first respond to robot, it has to
be able to adapt and respond to real world situations.
And this is what that challenge was really about, was
how well could you design a robot that could, uh,
you know, you know ahead of time what it's going
(34:02):
to have to do. And to some extent, that's going
to be the truth for at least the foreseeable future
where robots. Robots are going to be designed to do
specific things and anything outside of that they may or
may not be able to do based upon the stuff
you've programmed into that robot. But you know, it all
depends upon the design and programming of the robot with
(34:25):
foresight thinking these are the things this robot is typically
going to encounter, so here's how the robot needs to
be able to respond, and anything that goes outside of
that is really hard to design in a robot right now. Yeah, Yeah,
it's that old old problem of how do you tell
how do you tell a machine to specifically do something
when you don't know what that thing is. Yeah, and
(34:48):
so the special you know, surprise was an example that
it was. It was a variation on something that the
robots were already doing. So it wasn't like it was
completely lf field. It wasn't like, uh, you know, um,
you've you've got a cook an omelet, wasn't something like that,
you know it right, yeah, go buy games on Steam
(35:11):
and wasn't anything like that. So uh, you know, but
even then you see that some robots had real problems
with that. But this does mean that we are pushing
the field of robotics further as a result, And I
think that by identifying specific challenges, it gives engineers the
focus they need to be able to design the systems
(35:33):
that are capable of meeting those challenges. Right, So instead
of just having a wide open field where you say,
there's going to be an emergency situation, design a robot
to respond to it without any more direction. That's hard
to do. So the other point being that that this
also helped advance the the idea of how much communication
(35:55):
to humans need to be able to have with the
robot for it to be able to complete these challenges.
So a lot of these challenges the DARPA was saying, well,
we expect that some of the robots at least will
be able to respond to a fairly simple command and
carry it out, like go from this point to that point,
don't fall over. Not all the robots were able to
(36:17):
do that, obviously, but the dark but also the organization
I was saying, we don't expect any robot to be
able to go from start to finish with no human
contact contact at all, because to complete a single task
is one thing, but to chain them together in a
logical formation is another. So you wouldn't tell the robot, Hey,
(36:38):
I need you to drive to this point, Get out
of the vehicle, walk up to this door, open it.
Go through the door. You're going to see a wall
with a shape on it. Cut the shape out of
the wall, remove the rubble. Then you're gonna have something
we don't know what it is. Then you're going to
have to like you can't. You can't just tell a
robot that. It's much easier for right to to be
(36:59):
able to put that that input in every once in
a while to go like, oh, here's the part where
they're stairs, exactly, this is the stairs part. And because
otherwise it's trying to climb the vehicle and it's just
not you know, yeah, it's identified you know, a little
Timmy as the first step, and that's not going to
go well, uh, at least not for Timmy. So the
(37:20):
comparison I wanted to draw here was one of the
earlier DARPA challenges, the Grand Challenge. So with the Grand Challenge,
we saw all these teams trying to build automas cars
that could go through various obstacle courses. The first one
was a desert course. Then they had a simulated urban
environment complete with you know, traffic and other things estrians
(37:42):
that the cars had to contend with. And we saw
simulated pedestrians. Yes, yes, we saw no let me no
little Timmy, We're gonna need another Timmy. I think that
was from Dinosaurs, the television series. Um. But the uh,
the takeaway I wanted to give was that while that
(38:04):
wasn't intended to make automous cars reality the very next day,
what it did was it pushed the discipline forward, literally
pushed it forward, and we saw lots of people from
the various winning teams go into other organizations, including Google,
to develop the autonomous cars that are on the brink
of coming out on the consumer market. Now, within the
(38:27):
next few years, we we will probably see the first
steps and they'll probably be limited it first. But it's
because of that grand challenge that we're as far along
as we are. Sure. Yeah, and all these challenges, of
course also just raise public interest and excitement in the fields.
And you know, watching video of families that were out
(38:49):
watching the finals was so amazing, like getting to watch
the looks on kids faces and and yeah, and hearing
the cheers and the and the gasps from the crowd
whenever the robots were doing anything. It really does seem
to serve as an inspiration for people to get into
the stem part of education, right, to look into science, technology, engineering,
(39:09):
and uh, mathematics. Yeah. I was like, what's that fourth one?
That that one that we're all terrible at in this
podcast room English lit. Major If you listen to that
other episode where I suddenly reverted to my primal self.
Uh yeah. This this is one of those things where
you really you've got to admire it from multiple reasons. Um,
(39:30):
the idea of we're trying to solve a real world problem.
What happens in the face of an event like Fukushima.
How can we design something that could potentially save countless lives? Also,
how can we inspire the next generation of engineers to
want to get into that, to do that, and to
go beyond when we've even envisioned so far. So it's
(39:52):
really been great to look into this. I'm glad we
were able to do an episode about it and talk
about all the fun and foibles of robots, um whether
they're upright or falling over. Uh. It was really a
lot of fun to talk about this. So if you
guys have suggestions for future topics that we can cover
(40:12):
here on Forward Thinking, please write in and let us
know we love hearing from you. You guys have been
sending us some great ones, and we'll be doing another
listener mail episode before too long because we we accumulate
those and some of them, some of them are short
enough where we can do a couple in a single episode.
And then some of you guys have really complicated questions
that you want answered, and we're going to tackle those
(40:34):
as well. But if you have one you haven't voiced
it yet, send us an email FW thinking at how
stuff works dot com, or drop us a line on Twitter,
Google Plus, or Facebook. At Twitter and Google Plus, we
are f w thinking at Facebook. Just search fw thinking
in the search bar, will pop right up and leave
us a message. We'll talk to you again really soon.
(41:00):
For more on this topic and the future of technology,
visit forward thinking dot com, brought to you by Toyota.
Let's Go Places,
Fw:Thinking News
Hosts And Creators
Jonathan Strickland
Joe McCormick
Lauren Vogelbaum
Popular Podcasts
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.