MIT Helicopter
Here is a new video from MIT's CSAIL lab demonstrating their small autonomous robotic helicopter. The
video shows a compact yet very able platform, integrating many sensors for flight dynamics and
environment mapping. The MIT team won the 2009 International Aerial Robotics Competition. Watch
the whole video to see the mapping illustrations. The technical paper that fully describes their work is:
Autonomous Navigation and Exploration of a Quadrotor Helicopter in GPS-denied Indoor Environments.
This paper has a diagram showing the multi-level hierarchical control system that is structured as I've
come to expect from competition-winning autonomous vehicles. The lowest sub-symbolic level maintains
the helicopter's pitch and roll within a 1 ms feedback control loop using accelerometers and gyroscopes.
The next higher sub-symbolic level operates a 100 ms feedback control loop to maintain pose and
obstacle avoidance with respect to the helicopter's surroundings using a laser rangefinder and stereo
cameras. The final and topmost level is symbolic, performing trajectory planning with three second
temporal granularity.
It appears that the symbolic processing can be performed by the helicopter's on-board lightweight
processing power. In contrast, certain heavy computational tasks, which I suppose includes
sub-symbolic realtime perception of stereo images, are performed by a wireless networked laptop
computer cluster. I think that this research points to how domestic and workplace robots will be
organized. Mobile robots which have a local operating environment need not carry around heavy
unshared computer processors. Rather, when wireless bandwidth permits, a networked computational
resource can occupy a fixed location, and offload heavy sub-symbolic computational tasks from the
mobile robot.
I bring this research to your attention because it demonstrates how robust symbolic processing can be
grounded to the real world by a hierarchy of increasingly less abstract sub-symbolic levels, and because
I believe that this architecture is generally applicable to AGI.
Cheers.
-Steve
Stephen L. Reed
Artificial Intelligence Researcher
http://texai.org/blog
http://texai.org
3008 Oak Crest Ave.
Austin, Texas, USA 78704
512.791.7860
It is indeed generally applicable to all locomotive robots. I would tend to class this and other robotic
application as AI rather than AGI.
The holy grail of robotics is to get a robot which can :-
1) Navigate in a general way.
2) Has manual dexterity as good or better than humans.
Why not AGI? Well "The man who knows how works for the man who knows why". Dexterous locomotives
can perform any task that humans can perform - true. However a high level task still needs to be broken
down.
It should of course be pointed out that the performing of generalized mechanical tasks would provide an
enormous increase in robotic capabilities. One talks about a "singularity". Singularities can take many
forms. If we were to get a robot that could construct other similar robots - A Von Neumann machine, that
would be a breakthrough although perhaps not a Singularity. It should be pointed out too that a VN
machine only needs dexterity and locomotion. It does not require superhuman intellect.
- Ian Parker
Good to see people paying attention to robotics AGI.
I don't think you quite nail the AI/AGI difference though for robotics. And I would argue that robotics must
provide the defining paradigm for AGI (well actually there ain't going to be any other kind, other
perhaps than some useful simulations).
I think there's a definitive broad robotic AGI paradigm.
It is this - a true AGI robot must be a wayseeker(/goalseeker) ..
as distinct from AI robots wh. are still wayfollowers(/destinationfollowers).
IOW it must be able to seek and find its way to an ill-defined goal through a strange,unstructured
environment
as distinct from AI robots which follow the way laid down by their programs and external guides (eg
GPS).through basically familiar, structured environments to well-defined destinations.
(There can be some confusion here in that AI robots may *appear*.to navigate somewhat strange
environments - rooms they and their program haven't seen before - but if they're rooms of a certain type
then they're basically familiar).
You might say, AI robots navigate familiar cities, true AGI robots will be able to navigate unfamiliar
jungles.
True AGI robots will be true explorers - as opposed to guided tourists - (and there is no other way to do
that than what I've called adventurous trial-and-error - being able to freely explore new and strange
paths).
This is a good test for anyone who thinks algorithmic AGI can work - what's the algorithm for exploring
an unfamiliar jungle? You must say where you start, continue, carry on and end - *before* you've seen
the jungle - no problem-o?
The reason one can say this is a definitive paradigm is that it is the de facto scientific paradigm of all
living creatures - they are all described as wayseekers/goalseekers , and that's what animals do both
physically and metaphorically in all their activities. Find their way as opposed to follow their way
narrow-AI-style.
Mike Tinter
Andy Rubin who works for Google and who invented Android OS has his backyard "protected" by robotic
drones.
Next
Index
Here is a new video from MIT's CSAIL lab demonstrating their small autonomous robotic helicopter. The
video shows a compact yet very able platform, integrating many sensors for flight dynamics and
environment mapping. The MIT team won the 2009 International Aerial Robotics Competition. Watch
the whole video to see the mapping illustrations. The technical paper that fully describes their work is:
Autonomous Navigation and Exploration of a Quadrotor Helicopter in GPS-denied Indoor Environments.
This paper has a diagram showing the multi-level hierarchical control system that is structured as I've
come to expect from competition-winning autonomous vehicles. The lowest sub-symbolic level maintains
the helicopter's pitch and roll within a 1 ms feedback control loop using accelerometers and gyroscopes.
The next higher sub-symbolic level operates a 100 ms feedback control loop to maintain pose and
obstacle avoidance with respect to the helicopter's surroundings using a laser rangefinder and stereo
cameras. The final and topmost level is symbolic, performing trajectory planning with three second
temporal granularity.
It appears that the symbolic processing can be performed by the helicopter's on-board lightweight
processing power. In contrast, certain heavy computational tasks, which I suppose includes
sub-symbolic realtime perception of stereo images, are performed by a wireless networked laptop
computer cluster. I think that this research points to how domestic and workplace robots will be
organized. Mobile robots which have a local operating environment need not carry around heavy
unshared computer processors. Rather, when wireless bandwidth permits, a networked computational
resource can occupy a fixed location, and offload heavy sub-symbolic computational tasks from the
mobile robot.
I bring this research to your attention because it demonstrates how robust symbolic processing can be
grounded to the real world by a hierarchy of increasingly less abstract sub-symbolic levels, and because
I believe that this architecture is generally applicable to AGI.
Cheers.
-Steve
Stephen L. Reed
Artificial Intelligence Researcher
http://texai.org/blog
http://texai.org
3008 Oak Crest Ave.
Austin, Texas, USA 78704
512.791.7860
It is indeed generally applicable to all locomotive robots. I would tend to class this and other robotic
application as AI rather than AGI.
The holy grail of robotics is to get a robot which can :-
1) Navigate in a general way.
2) Has manual dexterity as good or better than humans.
Why not AGI? Well "The man who knows how works for the man who knows why". Dexterous locomotives
can perform any task that humans can perform - true. However a high level task still needs to be broken
down.
It should of course be pointed out that the performing of generalized mechanical tasks would provide an
enormous increase in robotic capabilities. One talks about a "singularity". Singularities can take many
forms. If we were to get a robot that could construct other similar robots - A Von Neumann machine, that
would be a breakthrough although perhaps not a Singularity. It should be pointed out too that a VN
machine only needs dexterity and locomotion. It does not require superhuman intellect.
- Ian Parker
Good to see people paying attention to robotics AGI.
I don't think you quite nail the AI/AGI difference though for robotics. And I would argue that robotics must
provide the defining paradigm for AGI (well actually there ain't going to be any other kind, other
perhaps than some useful simulations).
I think there's a definitive broad robotic AGI paradigm.
It is this - a true AGI robot must be a wayseeker(/goalseeker) ..
as distinct from AI robots wh. are still wayfollowers(/destinationfollowers).
IOW it must be able to seek and find its way to an ill-defined goal through a strange,unstructured
environment
as distinct from AI robots which follow the way laid down by their programs and external guides (eg
GPS).through basically familiar, structured environments to well-defined destinations.
(There can be some confusion here in that AI robots may *appear*.to navigate somewhat strange
environments - rooms they and their program haven't seen before - but if they're rooms of a certain type
then they're basically familiar).
You might say, AI robots navigate familiar cities, true AGI robots will be able to navigate unfamiliar
jungles.
True AGI robots will be true explorers - as opposed to guided tourists - (and there is no other way to do
that than what I've called adventurous trial-and-error - being able to freely explore new and strange
paths).
This is a good test for anyone who thinks algorithmic AGI can work - what's the algorithm for exploring
an unfamiliar jungle? You must say where you start, continue, carry on and end - *before* you've seen
the jungle - no problem-o?
The reason one can say this is a definitive paradigm is that it is the de facto scientific paradigm of all
living creatures - they are all described as wayseekers/goalseekers , and that's what animals do both
physically and metaphorically in all their activities. Find their way as opposed to follow their way
narrow-AI-style.
Mike Tinter
Andy Rubin who works for Google and who invented Android OS has his backyard "protected" by robotic
drones.
Next
Index
 | |  | |  | |  |