August 4, 2025 • 18 mins
Advances in fundamental technologies enable robots to collaborate with humans, as well as with other robots. David Saldaña, assistant professor in the department of computer science and engineering at Lehigh University, discusses his work developing resilient and adaptive collaborative aerial robots.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:03):
This is the Discovery Files podcastfrom the U.S.
National Science Foundation.
Whether fully autonomousor in close collaboration with humans.
Robots are becoming more prevalentthroughout people's lives.
From the factory floorto the operating room
to space exploration,
the field of roboticsis deeply interdisciplinary, combining
advances in engineering with innovationsin computer science.
(00:24):
As technologies advance,so do the possibilities for what
we might use them for.
We're joined today by David Saldaña,assistant professor in the Department
of Computer Scienceand Engineering at Lehigh University.
His research focuseson developing resilient and adaptive
multi-robot systems capable of operatingin unpredictable environments.
Doctor Saldaña,thank you so much for joining me today.
(00:45):
Oh, thank you so much for the invitation.
I'm glad to be here and be ableto talk about our research work.
So as we move into the research
work, you're dealing with multi rotoraerial vehicles.
Is there a difference between those
and what people generallyjust call colloquially “drones”.
Yeah it's basically the same concept.
But in the multi rotor systemsthey could have four,
(01:08):
it could be a quadrotor,it could be eight, it could be six.
But it's the same idea.
It's just a vehicle that is in the air.
What are some of the ways these kind ofvehicles are currently useful?
There has been a boom in their usesince the last 5 to 10 years.
We've seen them as toys in Christmas.
We see a lot of photographyin real estate.
(01:30):
You can see that now,people like the high view of their houses.
In military purposes.
Drone delivery as well.
So it has been becoming more popular.
And what we are trying to dois like to move something
that 20 years ago seemed like irrealisticor just like a sci movie.
We tried to make it real.
(01:52):
So how might these aerial robotsgrab something?
Traditionally, what researchers did in
the last ten yearswas when they needed to move an object.
They put a mechanism on the droneto allow that to grab and move the object.
Right.
So we talk about a grippers, robot arms.
(02:13):
But these mechanisms are very heavyand they are difficult to operate.
The dronedoesn't have to move its body anymore.
It has to also move an external bodythat is articulated and brings
more problems to the system.
You mentioned the grippers.
I know you're developing hitchesas a different approach.
Can you talk a little bit
(02:33):
about how hitches might be beneficialin, say, the construction industry?
Robot arms are heavy and they need power.
They are expensive.
Now we are proposing somethingthat is completely different.
It's light.
It doesn't need a battery, very low cost.
Now the challenge is how can we do that?
Right?
Humans have been using ropes for centuries
(02:56):
since the beginning of the humanity,but they are not that popular in robotics.
So we want to bring thatarchaic technology and provide it
that these new capabilityto the aerial robots.
Now that we have ropes,we have a drone that we call the catenary
robot is basically two dronesconnected with a cable.
(03:18):
They curve of the cabledescribed, is called the catenary
curve based on these catenary robot.
We can have many of them andthey can start to interlace the cables.
They can start to move around themand move around the object.
So we can tie the objects.
And we can we can move the objects.
(03:40):
In construction, we need to move
materialsand tools from one place to another.
That's what the builders do all day.
Moving objects around.
So we want to give them a new tooland a helper who will come and move the,
the objects to difficult placeslike in the sky scrapers.
(04:02):
When we have like very high locations,
the drones will be able to pick the toolsfrom the ground and take it to them.
When you using dronesto try to move things into the skyscraper
as your example had there, what are someof the challenges you run into?
So the main challenge comes from the ropesor the cables that we try to manipulate,
(04:22):
because they can have infiniteconfigurations with infinite shapes.
So we need to give the capability of therobots to manipulate those cables, right?
To slide them from each other.
One of the difficult problemsthat we face is trying to tie the knot.
Once we make the knot, tying it.
(04:43):
It's complicated task for the drones,especially because that depends
on the material, how the rope wasfabricated, how elastic it is.
So we have all these properties,
and the way that we doit is by interacting with the cable.
We humans, we are really good at thatbecause we have experimented
(05:03):
with many ropesand we understand how they work.
But for robots,they need to come and see, like,
okay,if I pull, this is what's going to happen.
And it has to be coordinatedbecause we are trying to do this
task with cooperative robots.
So if I pull, I shouldn't pull too hardor I will pull my colleague there.
(05:24):
So I'm thinking about them workingas in a collaboration
or in the swarm state,if there's a whole bunch of them.
How much are they aware of each other?
How much is it like one program, I guess.
What are the sort of the strategiesfor them interacting with each other?
Oh, especially in this concept
of making hitchesand making knots in the air.
The fundamental robot that we haveis two drones carrying a cable,
(05:49):
and then they can start to interlacewith each other right?
But in this setting, the distanceor the obstacles are not as important
as knowing if one of the cables is aboveor below the other.
That's going to determine that if I pull,I'm going to have a cable in my hand
or I'm going to have an object.
(06:11):
For that, we need to understandthe topology of the system.
And that's where we planbased on topology,
not based on a metric system,because a centimeter will not be much
in the metric system, but for the cablesit will be completely different, right?
The main idea here is that we have a team
(06:31):
of eight catenary robots,and their objective
is to interlace their cablesto be able to transport an object.
Right.
For that, they need to calculate a planabout how to move.
How to move one cable aroundthe other is like meeting.
We need a sequence of cablegoes up, cable goes down.
(06:52):
And they should be able to figure it outautonomously.
Right?
That's done based on the team,the teammates that they have
and also based on the object,because the object has some specific shape
that they need to go aroundand they start to wrap
to create a knotthat will help to move the object.
Thinking about them interactingwith each other as a team
(07:13):
there, would one unitmaybe have a specialty?
Thinking about knitting, there's a certaindexterity that's involved with that.
Would certainrobots need to do certain parts of it,
or would you ideally haveeverybody could do whatever?
So far we define two types of drones.
Two drones are like strong.
They are the ones who are going to pulland carry the object.
(07:35):
And some of your drones are just helpers,right?
We can think when we tie our shoes,we pull the two ends with strain right?
But while we make the knot,we don't need that strain.
We just need to hold it on one positionwhile the lace passes around.
We have the same thing.
Some drones will only be able tohold the cable while the others go around.
(07:58):
Initially you mentioned how,how heavy, like an arm or a grabber is,
in the cases of these helpersthat are doing some of the finer detail,
what kind of attachment or tooldo they need to do that job?
That's part of what we are trying to do.
Something that is very lightweight,maybe like a 3D
printed little piece, like a hook.
(08:19):
The object’s only to to hold the cable
for a little bit of time,so maybe they just pass the hook,
they hold itand then they can just release it.
For the control of them,do you have to operate them
independently or is it more of beingyou program it to do something.
So our objective is that you tell me,
(08:41):
I need to move this chairfrom this room to this other room,
or I need to move this tool to the topof an skyscraper, right?
And the robot should be able to figure outwhere to pass the cables
to grab the object, to make a knotand be able to move it.
Right?So for that, they need to coordinate.
They need to understandwhat the others are doing.
(09:03):
We have different approaches for that.
And one of them iswe define our own type of hitch.
That was the special part of thatwe have been doing the last three years
because, what my PhD student did was like,he went through
all the marine knots and checkingwhich one we can do with drones, right?
(09:23):
And after some research,he figured out that none of them
was like easy to make and easy to untie.
So we started to thinklike how we want to control the friction.
At some pointwhen we pull. We wanted to, we wanted to
release the object right?
But so depending on the actionsthat we do,
(09:45):
we we still want to hold it or release it.
And then we design our own hitch
and that hitch is based on a polygonthat we can adapt to any object.
And the robots know the polygon
and they will know that they belongto one side of the polygon.
So they can do it in parallel.
They can do their jobwithout thinking about the others.
(10:07):
It’s like, I need to place my cable here,then I need to wait for my neighbor,
and then we can make a twistto start to interlace the cables.
Have you had any caseswhere one of them runs into an issue
where maybe, maybe it's damaged,maybe it shuts down,
maybe the battery dies or something,
but something happensto where it can't do its part.
(10:28):
What does the first one do at that point?
I think right nowthere are more failures than success.
Yeah, it's like all the things happen.
Like a funny one is that you have eightrobots trying to lift an object, right?
And one of the drones stopped working.
And then now you have seven drones tryingto lift the object and a dead robot,
(10:51):
and they still can do it.
So it's quite interesting howthe system can adapt so they stop working.
The cables slide when they shouldn't,
or the cablesthey get stuck when they should slide.
And that's where our research comes in.
How we can make the cablesdo what we want, right
when we want to slide we will do it.
(11:13):
We humans, we have some actions.
Like we shake objects, we pull harder.
But dronesstill don't have those abilities yet.
Thinking about that
that way, like you're
picking up the rope and you want it here,but it's here right now.
Like a human might just go, oh,nudge it over a little bit.
What's the drone gonna do?
We are using reinforcementlearning for that.
(11:35):
So at that point when it's like
we have the model of the vehicle,we have the model of the cable.
But there are so many variables insidelike the friction,
how the cables are interlacedthat what we tell the algorithm is like.
Now you have to learn.
Now you have to do some interactionand see what is the best action
(11:55):
that will move thethat crossing point where we want.
What are the challengeswith that training?
Are you using reinforcementlearning primarily? Yes.
So on one side, what we use for,to interlace and determine
how the robot should move around,we use topological planning.
And for the partwhere we need to slide the cables
(12:19):
we need to tie, we need to applydifferent actions with them.
That's when we use reinforcement learning.
Specifically,we use model based reinforcement learning,
and we try to combine the models and the
all the dynamics and the mathematicsthat we know with what we don't know.
That's the challenge, right?
Getting that real time reaction.
(12:40):
Yes. Especially aerial robotscome with challenges different
from the reinforcement learningapplied to Atari, where you can run it
many of times and eventuallyit will solve the video game.
In aerial robots, the batteries don't lastthat long, so they need to learn fast.
They cannot make big mistakesbecause then they will crash.
(13:03):
So these are the main challenges here.
Like how to do it fast, how to do itaccurately and safe.
The stakescan be a little more catastrophic.
Yes, especially when we are trying
to move objects in the air and everythingfalls and break.
What would you say is the biggest benefit,maybe
of having a collaborative robot system?
There are multiple advantages.
(13:25):
One of them is the versatility,how we can associate
number with thehow big or how heavy the payload.
In some cases, maybe tworobots are enough to move an object.
Maybe I want to move a piano, thenmaybe I need a hundred robots to do that.
Right?
Or we want to pass the cablesinto different parts to make it safe.
(13:47):
So that versatility that I can addand remove
drones on demandis the main advantage of the swarm, right?
And the cooperative work, once they definethe objective, they can do it as a team.
Can you talk about some of the challenges
interactingwhen it's multiple units in a team?
One of the main challengesis the logistic.
(14:10):
More robots mean more components,more batteries to change
and especially more points of failure.
So we are trying to be resilient to that.
And as I mentioned with the example,like if one robot fails,
the system should still workor a few robots fails,
the system should still be ableto make the task that we asked them to do.
(14:32):
So one of the kind of ideas that people onmy end of things got excited
about was how this might be used indisaster response or emergency response.
Can you talk a little bit about whereyou might like to get the technology
to in that regard.
Especially in disaster response, right?
Having multiple robots,we have this versatile path right?
(14:55):
We don't need to come like,oh, we are going to move these type
of objects.
No, It's like whatever that we havein front of us, we will adapt to them.
Right.
So we have a versatile solution
and we will be able to move objectsand manipulate things.
that autonomous systemswere not able to do before.
For example, we can move plastic covers
(15:17):
to protect houseswhen there is like heavy rain and stuff.
Robots will be able to movethat they bend.
They are not easy to to manipulate.
The wind comes from one sideand now the robots need to adapt to that.
So these type of objectsare very challenging to move.
But when there is a disaster,
we will be able to move those objectsthat in the past we were not able to.
(15:41):
You recently got a career award.
Can you talk a little bit about whatgetting recognition from NSF means to you?
Oh thank you.
It means a lot like we have been workingin this type of problems.
Some of the problems are a little bitof science fiction.
Like some people don't believe ohyou can do that, cables are too complex.
(16:04):
But then we keep workingand then we get some advances,
and some people start to thinkthat it is possible, right?
So having the support from NSF to,to push these ideas forward,
it has been great for me and for my labin my career award, we work in
that similar problem is not exactlysame problem with the with the cables.
(16:26):
In that case what we did with objectsthat are soft when we try to move objects
that they don't stay rigid like a boxsometimes
like, piece of wood can bend.
People try to move their mattresswhen they move from one house to another.
It always bends.
We need to give the robots the capability
to move objects that are non rigid.
(16:49):
And we got the supportto move these ideas forward.
I have one last question for youand that's thinking about the future.
What are kind of your next stepsas you develop these robots.
So this is very,a very recent research area.
In the last 3 to 5 years
what we want is to have this workworking in the lab.
(17:12):
Right. We want to start movingobjects in the lab.
We want to be consistent.
As I told you,now we have more failures than success.
There are so many variables
that we need to understandhow to make a reliable controllers.
And safe controllers, right.
So we are working on that,
developing the theoryto support these autonomous vehicles.
(17:35):
Our plan for the next ten yearsis to have these
in the real world,not being in the sci movies anymore.
We want them to be part of the real life.
We want to move an object to a houseand is like, oh,
I just call my team of drones.
They do itand they go back to their station.
Special thanks to David Saldaña.
(17:56):
For The Discovery Files, I’m Nate Pottker.Watch video versions
of these conversations on our @NSFscienceYouTube channel.
Please subscribe wherever you get podcastsand if you like our program,
share with a friendand consider leaving a review.
Discover how the U.S.
National Science Foundationis advancing research at NSF.gov.
This is the Discovery Files podcastfrom the U.S.
National Science Foundation.
Whether fully autonomousor in close collaboration with humans.
Robots are becoming more prevalentthroughout people's lives.
From the factory floorto the operating room
to space exploration,
the field of roboticsis deeply interdisciplinary, combining
advances in engineering with innovationsin computer science.
(00:24):
As technologies advance,so do the possibilities for what
we might use them for.
We're joined today by David Saldaña,assistant professor in the Department
of Computer Scienceand Engineering at Lehigh University.
His research focuseson developing resilient and adaptive
multi-robot systems capable of operatingin unpredictable environments.
Doctor Saldaña,thank you so much for joining me today.
(00:45):
Oh, thank you so much for the invitation.
I'm glad to be here and be ableto talk about our research work.
So as we move into the research
work, you're dealing with multi rotoraerial vehicles.
Is there a difference between those
and what people generallyjust call colloquially “drones”.
Yeah it's basically the same concept.
But in the multi rotor systemsthey could have four,
(01:08):
it could be a quadrotor,it could be eight, it could be six.
But it's the same idea.
It's just a vehicle that is in the air.
What are some of the ways these kind ofvehicles are currently useful?
There has been a boom in their usesince the last 5 to 10 years.
We've seen them as toys in Christmas.
We see a lot of photographyin real estate.
(01:30):
You can see that now,people like the high view of their houses.
In military purposes.
Drone delivery as well.
So it has been becoming more popular.
And what we are trying to dois like to move something
that 20 years ago seemed like irrealisticor just like a sci movie.
We tried to make it real.
(01:52):
So how might these aerial robotsgrab something?
Traditionally, what researchers did in
the last ten yearswas when they needed to move an object.
They put a mechanism on the droneto allow that to grab and move the object.
Right.
So we talk about a grippers, robot arms.
(02:13):
But these mechanisms are very heavyand they are difficult to operate.
The dronedoesn't have to move its body anymore.
It has to also move an external bodythat is articulated and brings
more problems to the system.
You mentioned the grippers.
I know you're developing hitchesas a different approach.
Can you talk a little bit
(02:33):
about how hitches might be beneficialin, say, the construction industry?
Robot arms are heavy and they need power.
They are expensive.
Now we are proposing somethingthat is completely different.
It's light.
It doesn't need a battery, very low cost.
Now the challenge is how can we do that?
Right?
Humans have been using ropes for centuries
(02:56):
since the beginning of the humanity,but they are not that popular in robotics.
So we want to bring thatarchaic technology and provide it
that these new capabilityto the aerial robots.
Now that we have ropes,we have a drone that we call the catenary
robot is basically two dronesconnected with a cable.
(03:18):
They curve of the cabledescribed, is called the catenary
curve based on these catenary robot.
We can have many of them andthey can start to interlace the cables.
They can start to move around themand move around the object.
So we can tie the objects.
And we can we can move the objects.
(03:40):
In construction, we need to move
materialsand tools from one place to another.
That's what the builders do all day.
Moving objects around.
So we want to give them a new tooland a helper who will come and move the,
the objects to difficult placeslike in the sky scrapers.
(04:02):
When we have like very high locations,
the drones will be able to pick the toolsfrom the ground and take it to them.
When you using dronesto try to move things into the skyscraper
as your example had there, what are someof the challenges you run into?
So the main challenge comes from the ropesor the cables that we try to manipulate,
(04:22):
because they can have infiniteconfigurations with infinite shapes.
So we need to give the capability of therobots to manipulate those cables, right?
To slide them from each other.
One of the difficult problemsthat we face is trying to tie the knot.
Once we make the knot, tying it.
(04:43):
It's complicated task for the drones,especially because that depends
on the material, how the rope wasfabricated, how elastic it is.
So we have all these properties,
and the way that we doit is by interacting with the cable.
We humans, we are really good at thatbecause we have experimented
(05:03):
with many ropesand we understand how they work.
But for robots,they need to come and see, like,
okay,if I pull, this is what's going to happen.
And it has to be coordinatedbecause we are trying to do this
task with cooperative robots.
So if I pull, I shouldn't pull too hardor I will pull my colleague there.
(05:24):
So I'm thinking about them workingas in a collaboration
or in the swarm state,if there's a whole bunch of them.
How much are they aware of each other?
How much is it like one program, I guess.
What are the sort of the strategiesfor them interacting with each other?
Oh, especially in this concept
of making hitchesand making knots in the air.
The fundamental robot that we haveis two drones carrying a cable,
(05:49):
and then they can start to interlacewith each other right?
But in this setting, the distanceor the obstacles are not as important
as knowing if one of the cables is aboveor below the other.
That's going to determine that if I pull,I'm going to have a cable in my hand
or I'm going to have an object.
(06:11):
For that, we need to understandthe topology of the system.
And that's where we planbased on topology,
not based on a metric system,because a centimeter will not be much
in the metric system, but for the cablesit will be completely different, right?
The main idea here is that we have a team
(06:31):
of eight catenary robots,and their objective
is to interlace their cablesto be able to transport an object.
Right.
For that, they need to calculate a planabout how to move.
How to move one cable aroundthe other is like meeting.
We need a sequence of cablegoes up, cable goes down.
(06:52):
And they should be able to figure it outautonomously.
Right?
That's done based on the team,the teammates that they have
and also based on the object,because the object has some specific shape
that they need to go aroundand they start to wrap
to create a knotthat will help to move the object.
Thinking about them interactingwith each other as a team
(07:13):
there, would one unitmaybe have a specialty?
Thinking about knitting, there's a certaindexterity that's involved with that.
Would certainrobots need to do certain parts of it,
or would you ideally haveeverybody could do whatever?
So far we define two types of drones.
Two drones are like strong.
They are the ones who are going to pulland carry the object.
(07:35):
And some of your drones are just helpers,right?
We can think when we tie our shoes,we pull the two ends with strain right?
But while we make the knot,we don't need that strain.
We just need to hold it on one positionwhile the lace passes around.
We have the same thing.
Some drones will only be able tohold the cable while the others go around.
(07:58):
Initially you mentioned how,how heavy, like an arm or a grabber is,
in the cases of these helpersthat are doing some of the finer detail,
what kind of attachment or tooldo they need to do that job?
That's part of what we are trying to do.
Something that is very lightweight,maybe like a 3D
printed little piece, like a hook.
(08:19):
The object’s only to to hold the cable
for a little bit of time,so maybe they just pass the hook,
they hold itand then they can just release it.
For the control of them,do you have to operate them
independently or is it more of beingyou program it to do something.
So our objective is that you tell me,
(08:41):
I need to move this chairfrom this room to this other room,
or I need to move this tool to the topof an skyscraper, right?
And the robot should be able to figure outwhere to pass the cables
to grab the object, to make a knotand be able to move it.
Right?So for that, they need to coordinate.
They need to understandwhat the others are doing.
(09:03):
We have different approaches for that.
And one of them iswe define our own type of hitch.
That was the special part of thatwe have been doing the last three years
because, what my PhD student did was like,he went through
all the marine knots and checkingwhich one we can do with drones, right?
(09:23):
And after some research,he figured out that none of them
was like easy to make and easy to untie.
So we started to thinklike how we want to control the friction.
At some pointwhen we pull. We wanted to, we wanted to
release the object right?
But so depending on the actionsthat we do,
(09:45):
we we still want to hold it or release it.
And then we design our own hitch
and that hitch is based on a polygonthat we can adapt to any object.
And the robots know the polygon
and they will know that they belongto one side of the polygon.
So they can do it in parallel.
They can do their jobwithout thinking about the others.
(10:07):
It’s like, I need to place my cable here,then I need to wait for my neighbor,
and then we can make a twistto start to interlace the cables.
Have you had any caseswhere one of them runs into an issue
where maybe, maybe it's damaged,maybe it shuts down,
maybe the battery dies or something,
but something happensto where it can't do its part.
(10:28):
What does the first one do at that point?
I think right nowthere are more failures than success.
Yeah, it's like all the things happen.
Like a funny one is that you have eightrobots trying to lift an object, right?
And one of the drones stopped working.
And then now you have seven drones tryingto lift the object and a dead robot,
(10:51):
and they still can do it.
So it's quite interesting howthe system can adapt so they stop working.
The cables slide when they shouldn't,
or the cablesthey get stuck when they should slide.
And that's where our research comes in.
How we can make the cablesdo what we want, right
when we want to slide we will do it.
(11:13):
We humans, we have some actions.
Like we shake objects, we pull harder.
But dronesstill don't have those abilities yet.
Thinking about that
that way, like you're
picking up the rope and you want it here,but it's here right now.
Like a human might just go, oh,nudge it over a little bit.
What's the drone gonna do?
We are using reinforcementlearning for that.
(11:35):
So at that point when it's like
we have the model of the vehicle,we have the model of the cable.
But there are so many variables insidelike the friction,
how the cables are interlacedthat what we tell the algorithm is like.
Now you have to learn.
Now you have to do some interactionand see what is the best action
(11:55):
that will move thethat crossing point where we want.
What are the challengeswith that training?
Are you using reinforcementlearning primarily? Yes.
So on one side, what we use for,to interlace and determine
how the robot should move around,we use topological planning.
And for the partwhere we need to slide the cables
(12:19):
we need to tie, we need to applydifferent actions with them.
That's when we use reinforcement learning.
Specifically,we use model based reinforcement learning,
and we try to combine the models and the
all the dynamics and the mathematicsthat we know with what we don't know.
That's the challenge, right?
Getting that real time reaction.
(12:40):
Yes. Especially aerial robotscome with challenges different
from the reinforcement learningapplied to Atari, where you can run it
many of times and eventuallyit will solve the video game.
In aerial robots, the batteries don't lastthat long, so they need to learn fast.
They cannot make big mistakesbecause then they will crash.
(13:03):
So these are the main challenges here.
Like how to do it fast, how to do itaccurately and safe.
The stakescan be a little more catastrophic.
Yes, especially when we are trying
to move objects in the air and everythingfalls and break.
What would you say is the biggest benefit,maybe
of having a collaborative robot system?
There are multiple advantages.
(13:25):
One of them is the versatility,how we can associate
number with thehow big or how heavy the payload.
In some cases, maybe tworobots are enough to move an object.
Maybe I want to move a piano, thenmaybe I need a hundred robots to do that.
Right?
Or we want to pass the cablesinto different parts to make it safe.
(13:47):
So that versatility that I can addand remove
drones on demandis the main advantage of the swarm, right?
And the cooperative work, once they definethe objective, they can do it as a team.
Can you talk about some of the challenges
interactingwhen it's multiple units in a team?
One of the main challengesis the logistic.
(14:10):
More robots mean more components,more batteries to change
and especially more points of failure.
So we are trying to be resilient to that.
And as I mentioned with the example,like if one robot fails,
the system should still workor a few robots fails,
the system should still be ableto make the task that we asked them to do.
(14:32):
So one of the kind of ideas that people onmy end of things got excited
about was how this might be used indisaster response or emergency response.
Can you talk a little bit about whereyou might like to get the technology
to in that regard.
Especially in disaster response, right?
Having multiple robots,we have this versatile path right?
(14:55):
We don't need to come like,oh, we are going to move these type
of objects.
No, It's like whatever that we havein front of us, we will adapt to them.
Right.
So we have a versatile solution
and we will be able to move objectsand manipulate things.
that autonomous systemswere not able to do before.
For example, we can move plastic covers
(15:17):
to protect houseswhen there is like heavy rain and stuff.
Robots will be able to movethat they bend.
They are not easy to to manipulate.
The wind comes from one sideand now the robots need to adapt to that.
So these type of objectsare very challenging to move.
But when there is a disaster,
we will be able to move those objectsthat in the past we were not able to.
(15:41):
You recently got a career award.
Can you talk a little bit about whatgetting recognition from NSF means to you?
Oh thank you.
It means a lot like we have been workingin this type of problems.
Some of the problems are a little bitof science fiction.
Like some people don't believe ohyou can do that, cables are too complex.
(16:04):
But then we keep workingand then we get some advances,
and some people start to thinkthat it is possible, right?
So having the support from NSF to,to push these ideas forward,
it has been great for me and for my labin my career award, we work in
that similar problem is not exactlysame problem with the with the cables.
(16:26):
In that case what we did with objectsthat are soft when we try to move objects
that they don't stay rigid like a boxsometimes
like, piece of wood can bend.
People try to move their mattresswhen they move from one house to another.
It always bends.
We need to give the robots the capability
to move objects that are non rigid.
(16:49):
And we got the supportto move these ideas forward.
I have one last question for youand that's thinking about the future.
What are kind of your next stepsas you develop these robots.
So this is very,a very recent research area.
In the last 3 to 5 years
what we want is to have this workworking in the lab.
(17:12):
Right. We want to start movingobjects in the lab.
We want to be consistent.
As I told you,now we have more failures than success.
There are so many variables
that we need to understandhow to make a reliable controllers.
And safe controllers, right.
So we are working on that,
developing the theoryto support these autonomous vehicles.
(17:35):
Our plan for the next ten yearsis to have these
in the real world,not being in the sci movies anymore.
We want them to be part of the real life.
We want to move an object to a houseand is like, oh,
I just call my team of drones.
They do itand they go back to their station.
Special thanks to David Saldaña.
(17:56):
For The Discovery Files, I’m Nate Pottker.Watch video versions
of these conversations on our @NSFscienceYouTube channel.
Please subscribe wherever you get podcastsand if you like our program,
share with a friendand consider leaving a review.
Discover how the U.S.
National Science Foundationis advancing research at NSF.gov.
Popular Podcasts
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.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.