DPRG List  

[DPRG] Hello again DPRG. (Re-Introduction) and long message.

Subject: [DPRG] Hello again DPRG. (Re-Introduction) and long message.
From: Brad garton bgart at iadfw.net
Date: Mon Apr 16 11:15:45 CDT 2012

I was in the DPRG when I lived in Dallas in 1998. I sort of competed (not
very successfully) in the 1998 RoboRama. I moved to Austin and due to the
nature of my job, I have spent very little time in robotics since then. I
have always wanted to complete the design for Tom Servo that I started back
in 1998.  The sensors and controllers available have changed over the years
and I am currently trying to get back up to speed again.   That's what I'm
working on now. 


My original interest in robotics came from visiting mit.edu cmu, and
Stanford research sites. and I became very interested in Rodney Brook's
ideas on robot programming.  When I came to DPRG I discovered people like
David Anderson had already implemented much of what I wanted to do, and
these robots performed very well at least in a structured environment. I
have continued to watch people progress in building personal mobile robots. 


Now I have begun implementing again, and as I am older,  I want to consider
the big picture (who knows how long I have to work on robots).  One thing
that has been observed is that the state of the Art in personal robotics has
not progressed much from where it was in 1998. Is this true? I think there
is at least some truth to the comment.  SR04 was and is, one of the most
capable robots I have observed built by an individual.  Few of us have the
skills to match that design, and we would do well to emulate David's work,
in my opinion.  However, what then? 


Why is robotics so hard and why is it difficult to move the state of the art
forward? If you look at research going on, in industry and the university,
there are some new sensor designs available now  (Kinect, SICK etc) and some
new techniques (SLAM) and these have been shown to be useful in the DARPA
grand challenges for example. Not to mention the Quadrotor designs from
UPenn.   All of these are very interesting developments to be sure.  All new
since 1998. How can an individual hope to learn and implement robots that
are more capable with the new techniques and technologies available now?


I think robotics is hard for the neophyte and I have tried to think of why
that is. Here are some things I came up with based on my own experience. 


1.       It spans multiple disciplines. 

a.        Programming 

b.       Mechanical Design 

c.        Embedded Systems Design 

2.       Why is robot Programming Hard

a.        It is not a typical Program, where system events are predictable,
and functions proceed in an orderly sequential manner.  Events are by nature
random, and multiple things need to be done apparently simultaneously. 

b.       It spans multiple programming Domains. 

                                                               i.      Low
level machine level programming 

                                                              ii.      High
level behavioral or AI programming. 

c.        We do not understand for the most part how the analogs in Nature

3.       Why is the embedded system design problem harder than a lot of
other embedded systems design

a.        Robots must carry everything with them.  Power, Computing
Resources, Mechanical resources

4.       Why is the mechanical aspect of a robot design difficult? 

a.        Dynamics

b.       Actuators are generally power hungry and not so efficient

c.        Environments are complex, not simple

d.       Need to use lightweight materials due to the problem of carrying
everything with us. 

I came to robotics from a background in Electrical and Computer Engineering.
The things I did in my work time were to design embedded systems and
eventually I designed workstations for a 3 letter Acronym Company in Austin.
I feel like my skills circa 1998 were up to the Embedded System Design
aspects of robotics.  They have atrophied considerably since then. 

Rodney Brooks's ideas allowed us to implement a very simple "AI", which was
perfect for the environment of embedded system design in 1998, but perhaps
we have hit the limits of that architecture alone. It would seem we need to
develop higher level behaviors at the least.  I still think it can be part
of a more capable robot, though I'm not so sure the behaviors we currently
use are complete. 


I was an ok programmer, but I was not prepared for the challenge I
encountered in implementing my robots. I had zero experience in Mechanical
design, past an engineering mechanics course in College.  The result was a
capable embedded system, on poorly functioning mechanics that kept me from
getting to the point of writing interesting Software.  I am correcting these
problems now, and may get to the level of being able to complete RoboRama,
given sufficient time, but I would like to hear some ideas on how to
approach this problem in general. 


In particular I want to consider teaming, or crowd sourcing robotics
development.  It appears that we can progress more rapidly if we have more
canned solutions for the basic problems of robotics, and rather than
tackling the whole enchilada at once, combine skills in teams to produce
functioning robots, with the key domains above represented.  What do you
think?  I can see advantages and disadvantages but want to hear what the
group thinks. 





-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.dprg.org/pipermail/dprglist/attachments/20120416/dc94eddb/attachment.html 

More information about the DPRG mailing list