My robot life.. not shot on red (although i wish it was)... DARPA Robotics Trials...

[william howell]

here's a little glimpse into the robotic research world... I know some of you worked on Transformers... but.. i am lucky enough to work on forefront of robotic research... these things are MUCH slower than our imaginations... and very unsexy compared to Optimus Prime... but... in the end.. they're still kind of cool..

my world is very similar to the film industry.. however... instead of a huge production crew... it's literally just me and a coworker...

Robotic research relies HEAVILY on video... from high speed observation to troubleshoot problems, to onboard vision systems... it's really intense..

shot on sony fs700, fs100, canon 5d mark iii, black magic cinema camera, black magic pocket cinema camera...

if anyone wants to come hang out with me this year and help us document our robotic research for the DRC finals, please... come to Pensacola Florida... bring all your red's...

We have so much to film, not just humanoid Robots...

enjoy...

 

[Les Dittert]

When I watched some of the event, I was thinking that stop motion would be the only way to make it interesting for the public ... man do the current robots spend a lot of time 'thinking' about the next tiny movement during the contests !

 

[william howell]

actually...


Robots don't do any of the "thinking". It's best to look at them like extremely complex bulldozers.

There is a human at a computer terminal controlling every movement that the robot makes... which is why these things take so long... the computer that the operator is using does the "movement" calculations. There are some onboard processes, but for the most part... it's the human at the terminal that is doing the majority of the "decision making".

Figuring out what actions should be "automatic" and what should be human controlled is the challenge...

think of your experience as a human being... there are two types of actions that your body performs... one type is "automatic"... these things include balance, breathing, controlling blood flow to muscles, etc... then there are the "deliberate" actions... these include walking to the refrigerator to get some mustard, opening the car door to get in, pointing a camera at something...

the automatic actions happen "under the hood"... you don't think about them.. these are the functions that the computer needs to handle... balance, joint angle positions, hydraulic fluid control..

the deliberate actions require the human brain to make the "decisions" that lead to actions.. these are the functions that the human operator needs to control... Grabbing a drill, cutting a hole in cement, turning a valve...

we look at robotics, and all computer systems for that matter, as devices that "augment" the human being... There is an operator that is actually making the decisions as to what the robot should do... the robot is just a remote controlled extension of the operator.


One has to understand that these are highly specialized devices built for the sole purpose of allowing human beings to enter environments that are too dangerous for humans to traverse...

Three of the reactors at Fukushima are not accessible by humans... As unsexy as it sounds... humanoid robots are designed to go into human centered environments and do really unsexy jobs... like pulling the radioactive material out of the ground at Fukushima.

wheeled and tracked robots are useless there. We need the ability to remove debris from in front of doors, open doors, traverse collapsed stairwells, walk over rubble... turn valves... operate machinery... all remotely.

sorry for all of the ...'s but I've had ALOT of coffee this morning..

 

[Neil Dearman]

how much do those robots cost? I want one!!!! Have his head to be a RED DRAGON

 

[william howell]

Robots are EXPENSIVE...

and the manpower behind transport and operation makes them about as useful as a politician.

which is why we're working on them... If RED were to "donate" a dragon to our cause, i'd do my damnedest to replace it's head with it...

 

[Les Dittert]

William,
I thought the robots had to be autonomous ( like in the Darpa robot driving on/off road contest ).
Having a human in the loop is 'cheating' IMO. Heck, then you could have a motion capture stage and just drive the robot with that to some degree, with the robot only maintaining basic balance tasks , for example.
I would permit tethering to power sources and compute power for it, but only allow humans to intervene to halt the robot from making a costly mistake, thereby disqualifying it from the contest.
Didn't the teams have fairly detailed descriptions of each setup, so the software would know what image/scene processing it would be challenged with ?

I like watching Bigdog and similar, even though they are more dumb.

 

[william howell]

The Darpa Grand Challenge was the quest for the autonomous vehicle..

The Darpa Robotics Challenge is the quest for a usable tool for human operators to do dangerous jobs at safe distances. Autonomy is not a requirement... In fact.. those teams that did rely on autonomy weren't able to adjust for changes in their environment, like wind blowing doors shut... robust adaptability is one of the main goals, because... the real world is anything but a controlled environment.

Think about it like your camera... Do you use any automatic settings? do you want autofocus on the whole time?... The camera allows us, the operators, to accomplish whatever it is we want to do... but... it's not going to replace the human's ability to "frame a good shot" or figure out when to rack focus... or who to point the camera at.

The human brain is far too complex a piece of machinery to just discard in the loop.

Driving is basically how the "civilized world" walks now... the infrastructure is in place to make it autonomous... maps... gps data.. brightly colored signs... painted roadways... sensors all over cars... etc.... It's a really simply problem to deal with.. basically... you deal with three variables... how fast do I need to go?... which direction should i point the car?... when should i stop?...

humanoid robots have 80+ degrees of freedom... the computational requirements to function in the "walking world" are substantially more problematic than driving.

 

[william howell]

this is the video we released to the public a few weeks before the trials. Basically, the whole Darpa Robotics Challenge is a field wide effort to advance humanoid robots, so it was much less of a competition than it was a field wide gathering, so each team released a "this is how we think we're going to do in the trials" video.

Unlike the industry that "plans out" shots... we have to be reactionary... usually there isn't time much time to set up shots or lighting equipment, so doing it on the fly is a requirement. It's a tough job when the action is happening... because there are no breaks... it's go go go.... then someone breaks the code... and everything stops....

 

[ericyoung]

Fascinating.

 

[william howell]

Les... i just read your post on the Back to the Future thread...

you have no idea how much some of the projects you've worked on has inspired our entire robotics staff... that's awesome..

i wish that I had enough money to pay for any interested member of the red forum to visit our facilities... and bring their cameras of course...

our geeks LOVE the movies everyone here has worked on...

 

[william howell]

Fascinating.

i think so too.

i've got 14 years of footage that i'm going to start building a documentary with.. we recently invested in a bunch of thunderbolt raids, and are in the process of porting all of our media into one place... needless to say... i'm excited..