[Theory] NOW: 2/12 Talks at TTIC: Ian Abraham, Carnegie Mellon University

[Mary Marre]

*When:*       Friday, February 12th at* 11:10 am CT*



*Where:*      Zoom Virtual Talk (*register in advance here
<https://uchicagogroup.zoom.us/webinar/register/WN_AtjkUtMlQxSS5wkgA19iuA>*)



*Who: *        Ian Abraham, Carnegie Mellon University



*Title:*         Runtime Active Learning for Reactive Robotics

*Abstract:* Robotic systems rely on human engineered guidance, programming
of tasks, and oracle information that enable them to operate in our world.
What happens to robotic systems when we are unable to perform as an oracle,
creating an absence of information about what is known and unknown to a
robot? Can we expect robotic systems to be intentional about how they seek
out information necessary to operate? And what are the necessary
requirements for them to explore and navigate the complexities that they
will face as we make them operate in increasingly unstructured environments?

In this talk, I argue that having the ability to actively acquire and seek
out informative data that improve robot learning is a stepping stone
towards autonomy detached from human engineers. Using existing methods and
tools from hybrid control theory, I first create the theoretical groundwork
for improving the learning capabilities of robots by combining various
modes of learning subject to robot dynamics and stability constraints. The
problem of active learning through experimental design, where anticipated
sensor data is optimized, motivates the use of methods from ergodicity and
ergodic exploration that enable robots to sample and sense from multiple
information sources while simultaneously ensuring that the robot does not
ignore unexplored parts of an environment. I illustrate the use of
ergodicity as a promising approach for learning models in complex and
spatially sparse environments using only rudimentary contact sensing. These
results are extended to more general active learning in dynamic
state-spaces where robot safety and the quality of informative measurements
are balanced using hybrid control theoretic analysis and known stabilizing
controllers. Last, I argue that we should not only care about active
learning, but also how we model and represent the dynamics of robotic
systems. The class of infinite linear embeddings is presented as a
candidate model that simplifies and improves the control and active
learning capabilities of robotic systems. Through simulated and
experimental application, I illustrate the potential of the presented
approaches for pushing the boundaries of robotic systems towards being more
capable, self-sufficient, and curious systems that intentionally seek out
the unknown and complex nature of interacting in our world.

*Bio:* Ian Abraham is a Postdoctoral scholar at the Robotics Institute at
Carnegie Mellon University. He received the B.S. degree in Mechanical and
Aerospace Engineering from Rutgers University and the M.S. and Ph.D degree
in Mechanical Engineering from Northwestern University at the Center for
Robotics and Biosystems. His Ph.D. work focuses on developing formal
methods for robot sensing and runtime active learning. During his Ph.D. he
interned at the NVIDIA Seattle Robotics Lab where he worked on robust
model-based control for large parameter uncertainty. He also participated
in the DARPA OFFSET FX-3 Urban Swarm Challenge and is the recipient of the
2019 King-Sun Fu IEEE Transactions on Robotics Best Paper award.



*Host:* Matthew Walter <mwalter at ttic.edu>




Mary C. Marre
Faculty Administrative Support
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 517*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*


On Fri, Feb 12, 2021 at 10:00 AM Mary Marre <mmarre at ttic.edu> wrote:

> *When:*       Friday, February 12th at* 11:10 am CT*
>
>
>
> *Where:*      Zoom Virtual Talk (*register in advance here
> <https://uchicagogroup.zoom.us/webinar/register/WN_AtjkUtMlQxSS5wkgA19iuA>*
> )
>
>
>
> *Who: *        Ian Abraham, Carnegie Mellon University
>
>
>
> *Title:*         Runtime Active Learning for Reactive Robotics
>
> *Abstract:* Robotic systems rely on human engineered guidance,
> programming of tasks, and oracle information that enable them to operate in
> our world. What happens to robotic systems when we are unable to perform as
> an oracle, creating an absence of information about what is known and
> unknown to a robot? Can we expect robotic systems to be intentional about
> how they seek out information necessary to operate? And what are the
> necessary requirements for them to explore and navigate the complexities
> that they will face as we make them operate in increasingly unstructured
> environments?
>
> In this talk, I argue that having the ability to actively acquire and seek
> out informative data that improve robot learning is a stepping stone
> towards autonomy detached from human engineers. Using existing methods and
> tools from hybrid control theory, I first create the theoretical groundwork
> for improving the learning capabilities of robots by combining various
> modes of learning subject to robot dynamics and stability constraints. The
> problem of active learning through experimental design, where anticipated
> sensor data is optimized, motivates the use of methods from ergodicity and
> ergodic exploration that enable robots to sample and sense from multiple
> information sources while simultaneously ensuring that the robot does not
> ignore unexplored parts of an environment. I illustrate the use of
> ergodicity as a promising approach for learning models in complex and
> spatially sparse environments using only rudimentary contact sensing. These
> results are extended to more general active learning in dynamic
> state-spaces where robot safety and the quality of informative measurements
> are balanced using hybrid control theoretic analysis and known stabilizing
> controllers. Last, I argue that we should not only care about active
> learning, but also how we model and represent the dynamics of robotic
> systems. The class of infinite linear embeddings is presented as a
> candidate model that simplifies and improves the control and active
> learning capabilities of robotic systems. Through simulated and
> experimental application, I illustrate the potential of the presented
> approaches for pushing the boundaries of robotic systems towards being more
> capable, self-sufficient, and curious systems that intentionally seek out
> the unknown and complex nature of interacting in our world.
>
> *Bio:* Ian Abraham is a Postdoctoral scholar at the Robotics Institute at
> Carnegie Mellon University. He received the B.S. degree in Mechanical and
> Aerospace Engineering from Rutgers University and the M.S. and Ph.D degree
> in Mechanical Engineering from Northwestern University at the Center for
> Robotics and Biosystems. His Ph.D. work focuses on developing formal
> methods for robot sensing and runtime active learning. During his Ph.D. he
> interned at the NVIDIA Seattle Robotics Lab where he worked on robust
> model-based control for large parameter uncertainty. He also participated
> in the DARPA OFFSET FX-3 Urban Swarm Challenge and is the recipient of the
> 2019 King-Sun Fu IEEE Transactions on Robotics Best Paper award.
>
>
>
> *Host:* Matthew Walter <mwalter at ttic.edu>
>
>
> Mary C. Marre
> Faculty Administrative Support
> *Toyota Technological Institute*
> *6045 S. Kenwood Avenue*
> *Room 517*
> *Chicago, IL  60637*
> *p:(773) 834-1757*
> *f: (773) 357-6970*
> *mmarre at ttic.edu <mmarre at ttic.edu>*
>
>
> On Thu, Feb 11, 2021 at 5:46 PM Mary Marre <mmarre at ttic.edu> wrote:
>
>> *When:*       Friday, February 12th at* 11:10 am CT*
>>
>>
>>
>> *Where:*      Zoom Virtual Talk (*register in advance here
>> <https://uchicagogroup.zoom.us/webinar/register/WN_AtjkUtMlQxSS5wkgA19iuA>*
>> )
>>
>>
>>
>> *Who: *        Ian Abraham, Carnegie Mellon University
>>
>>
>>
>> *Title:*         Runtime Active Learning for Reactive Robotics
>>
>> *Abstract:* Robotic systems rely on human engineered guidance,
>> programming of tasks, and oracle information that enable them to operate in
>> our world. What happens to robotic systems when we are unable to perform as
>> an oracle, creating an absence of information about what is known and
>> unknown to a robot? Can we expect robotic systems to be intentional about
>> how they seek out information necessary to operate? And what are the
>> necessary requirements for them to explore and navigate the complexities
>> that they will face as we make them operate in increasingly unstructured
>> environments?
>>
>> In this talk, I argue that having the ability to actively acquire and
>> seek out informative data that improve robot learning is a stepping stone
>> towards autonomy detached from human engineers. Using existing methods and
>> tools from hybrid control theory, I first create the theoretical groundwork
>> for improving the learning capabilities of robots by combining various
>> modes of learning subject to robot dynamics and stability constraints. The
>> problem of active learning through experimental design, where anticipated
>> sensor data is optimized, motivates the use of methods from ergodicity and
>> ergodic exploration that enable robots to sample and sense from multiple
>> information sources while simultaneously ensuring that the robot does not
>> ignore unexplored parts of an environment. I illustrate the use of
>> ergodicity as a promising approach for learning models in complex and
>> spatially sparse environments using only rudimentary contact sensing. These
>> results are extended to more general active learning in dynamic
>> state-spaces where robot safety and the quality of informative measurements
>> are balanced using hybrid control theoretic analysis and known stabilizing
>> controllers. Last, I argue that we should not only care about active
>> learning, but also how we model and represent the dynamics of robotic
>> systems. The class of infinite linear embeddings is presented as a
>> candidate model that simplifies and improves the control and active
>> learning capabilities of robotic systems. Through simulated and
>> experimental application, I illustrate the potential of the presented
>> approaches for pushing the boundaries of robotic systems towards being more
>> capable, self-sufficient, and curious systems that intentionally seek out
>> the unknown and complex nature of interacting in our world.
>>
>> *Bio:* Ian Abraham is a Postdoctoral scholar at the Robotics Institute
>> at Carnegie Mellon University. He received the B.S. degree in Mechanical
>> and Aerospace Engineering from Rutgers University and the M.S. and Ph.D
>> degree in Mechanical Engineering from Northwestern University at the Center
>> for Robotics and Biosystems. His Ph.D. work focuses on developing formal
>> methods for robot sensing and runtime active learning. During his Ph.D. he
>> interned at the NVIDIA Seattle Robotics Lab where he worked on robust
>> model-based control for large parameter uncertainty. He also participated
>> in the DARPA OFFSET FX-3 Urban Swarm Challenge and is the recipient of the
>> 2019 King-Sun Fu IEEE Transactions on Robotics Best Paper award.
>>
>>
>>
>> *Host:* Matthew Walter <mwalter at ttic.edu>
>>
>>
>>
>> Mary C. Marre
>> Faculty Administrative Support
>> *Toyota Technological Institute*
>> *6045 S. Kenwood Avenue*
>> *Room 517*
>> *Chicago, IL  60637*
>> *p:(773) 834-1757*
>> *f: (773) 357-6970*
>> *mmarre at ttic.edu <mmarre at ttic.edu>*
>>
>>
>> On Sun, Feb 7, 2021 at 11:14 AM Mary Marre <mmarre at ttic.edu> wrote:
>>
>>> *When:*       Friday, February 12th at* 11:10 am CT*
>>>
>>>
>>>
>>> *Where:*      Zoom Virtual Talk (*register in advance here
>>> <https://uchicagogroup.zoom.us/webinar/register/WN_AtjkUtMlQxSS5wkgA19iuA>*
>>> )
>>>
>>>
>>>
>>> *Who: *        Ian Abraham, Carnegie Mellon University
>>>
>>>
>>>
>>> *Title:*         Runtime Active Learning for Reactive Robotics
>>>
>>> *Abstract:* Robotic systems rely on human engineered guidance,
>>> programming of tasks, and oracle information that enable them to operate in
>>> our world. What happens to robotic systems when we are unable to perform as
>>> an oracle, creating an absence of information about what is known and
>>> unknown to a robot? Can we expect robotic systems to be intentional about
>>> how they seek out information necessary to operate? And what are the
>>> necessary requirements for them to explore and navigate the complexities
>>> that they will face as we make them operate in increasingly unstructured
>>> environments?
>>>
>>> In this talk, I argue that having the ability to actively acquire and
>>> seek out informative data that improve robot learning is a stepping stone
>>> towards autonomy detached from human engineers. Using existing methods and
>>> tools from hybrid control theory, I first create the theoretical groundwork
>>> for improving the learning capabilities of robots by combining various
>>> modes of learning subject to robot dynamics and stability constraints. The
>>> problem of active learning through experimental design, where anticipated
>>> sensor data is optimized, motivates the use of methods from ergodicity and
>>> ergodic exploration that enable robots to sample and sense from multiple
>>> information sources while simultaneously ensuring that the robot does not
>>> ignore unexplored parts of an environment. I illustrate the use of
>>> ergodicity as a promising approach for learning models in complex and
>>> spatially sparse environments using only rudimentary contact sensing. These
>>> results are extended to more general active learning in dynamic
>>> state-spaces where robot safety and the quality of informative measurements
>>> are balanced using hybrid control theoretic analysis and known stabilizing
>>> controllers. Last, I argue that we should not only care about active
>>> learning, but also how we model and represent the dynamics of robotic
>>> systems. The class of infinite linear embeddings is presented as a
>>> candidate model that simplifies and improves the control and active
>>> learning capabilities of robotic systems. Through simulated and
>>> experimental application, I illustrate the potential of the presented
>>> approaches for pushing the boundaries of robotic systems towards being more
>>> capable, self-sufficient, and curious systems that intentionally seek out
>>> the unknown and complex nature of interacting in our world.
>>>
>>> *Bio:* Ian Abraham is a Postdoctoral scholar at the Robotics Institute
>>> at Carnegie Mellon University. He received the B.S. degree in Mechanical
>>> and Aerospace Engineering from Rutgers University and the M.S. and Ph.D
>>> degree in Mechanical Engineering from Northwestern University at the Center
>>> for Robotics and Biosystems. His Ph.D. work focuses on developing formal
>>> methods for robot sensing and runtime active learning. During his Ph.D. he
>>> interned at the NVIDIA Seattle Robotics Lab where he worked on robust
>>> model-based control for large parameter uncertainty. He also participated
>>> in the DARPA OFFSET FX-3 Urban Swarm Challenge and is the recipient of the
>>> 2019 King-Sun Fu IEEE Transactions on Robotics Best Paper award.
>>>
>>> *Host:* Matthew Walter <mwalter at ttic.edu>
>>>
>>>
>>> Mary C. Marre
>>> Faculty Administrative Support
>>> *Toyota Technological Institute*
>>> *6045 S. Kenwood Avenue*
>>> *Room 517*
>>> *Chicago, IL  60637*
>>> *p:(773) 834-1757*
>>> *f: (773) 357-6970*
>>> *mmarre at ttic.edu <mmarre at ttic.edu>*
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/theory/attachments/20210212/278f7d43/attachment-0001.html>