Newsgroups: comp.robotics
Path: brunix!uunet!mdisea!mothost!lmpsbbs!ECK001@email.mot.com
From: ECK001@email.mot.com (Chris_Kerlin)
Subject: Re: vision guided robots
Organization: Motorola_Adv_Mfg_Tech
Date: Thu, 20 May 1993 19:22:06 GMT
Message-ID: <1993May20.192206.21526@lmpsbbs.comm.mot.com>
References: <1993May19.131333.10264@EE.Surrey.Ac.UK>
Sender: news@lmpsbbs.comm.mot.com (Net News)
Nntp-Posting-Host: 145.2.119.41
Lines: 21

In article <1993May19.131333.10264@EE.Surrey.Ac.UK>, j.pretlove@mech.surrey.ac.uk (John Pretlove) writes:
> 
> I have a CCD camera pair on the end of a robot arm which (quiet successfully)
> guides the robot arm to intercept an object in the workcell. To determine the
> relationship between the world and the robot coordinate system I carry out
> camera-to-world calibration and toolpoint-to-eye calibration.
> Not only are these calibrations tedious and complex but they are also error-prone.
> 
> My question to the newsgroup is: what are the alternatives ?
> 
> ...

Robots can sometimes be programmed to perform "tedious and complex" tasks :)
I must assume that U are performing the calibrations with human intervention.
Have U tried locating an object so that it can be picked up by the robot
without calibration, then have the robot pick, place, look, pick, rotate
180 deg and look again. We have found this sufficient for many fixed-focus
systems. Additional scenarios are possible if a third camera is located such
that it can see the toolpoint...

------------------------------------------------------------------------------CK
