Newsgroups: comp.robotics
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!howland.reston.ans.net!EU.net!sun4nl!phcoms4.seri.philips.nl!misf1!rvdouder
From: rvdouder@ms.philips.nl (Rob van der Ouderaa)
Subject: Re: Limbed 'Stick Insect' Robot
Message-ID: <1995May3.105107.28829@ms.philips.nl>
Sender: news@ms.philips.nl
Organization: Philips Medical Systems, Best
X-Newsreader: TIN [version 1.2 PL2]
References: <sdimaio.1.2FA10C48@eleceng.uct.ac.za>
Date: Wed, 3 May 1995 10:51:07 GMT
Lines: 45

sdimaio@eleceng.uct.ac.za wrote:
: Learning Algorithm for a Six (or Four) Legged Stick Insect Robot.

: Is there somebody out there who has successfuly created a 
: 'learning algorithm' for for a limbed robot?

: I am trying to formulate a final year thesis and am interested in this
: problem. The thing is that I need to be able to 'see' a definite way 
: of doing this before jumping in. Once I start, it is too late to turn
: back. If it fails... well, so do I.

: So far I have considered the following VERY GENERAL idea.

: Using the general idea of the Genetic Algorithm:

: - Each leg can have a standard 'action' or movement, but the  
:   timing of this movement may change.
: - The timing of each leg's movement relative to the other 
:   legs may change as well as the order in which the legs move.
:   
: Using the above characteristics as input variables to a function 
: whose output is a measure of the success of the walking algorithm (in 
: otherwords related to the movement of the centipede), one can apply 
: the genetic algorithm in order to converge on the most successful 
: walking strategy? How could I formulate this problem?

: How about a Neural Net?


: If there is anybody out there who has had any experience with problem
: I would greatly appreciate some ideas!


There is a project with a 6 legged robot called Genghis. I think it is
done at the MIT. Somewhere on the net there is a file with a MSDOS
simulation of the movement of this kind of robot. The file was called
something like GENGHIS.ZIP. You can use archie to find it.

I don't know how they control the movement of the robot but I thought I
would just give you a pointer. The simulation showed several ways that
the robot could move using it's legs.

Hope this helps,

	Rob
