Alt-BEAM Archive

Message #13918



To: alt-beam@yahoogroups.com
From: rob mantid@morphine.neuron.net
Date: Wed, 12 Dec 2001 11:28:59 -0800 (PST)
Subject: Re: [alt-beam] Truth or Dare



> To be heretical about it, the phrase "obstacle avoidance" is literally
> slathered all over BEAM literature. I've got a huge problem with
> that

This is a neat objection you have. But is the objection to the method of
[saving the machine from getting stuck on obstacles], or to the term
itself? What new term would you suggest?

I can't imagine that you don't see the tactile sensors as useful. You are
right that they dont avoid contact with the obstacles, but merely smacking
into the obstacles doesn't change the behavior of the motors. So rather
than obstacle avoidance, we have a "deal with a crash" system. You would
like to see a design to truly avoid the crash?

And if even using a stream of IR photons is contact... Hmmm.

I suppose you could just passively recieve photons already bouncing off
the object (light?) and store those in a kind of mental map. That is I
think the simple way to describe what humans to do "avoid obstacles".

Its kindof fun to imagine an obstacle so fragile or dangerous that brush
by a wire sensor would destroy it, trigger a bomb, or something. Doesn't
seem as useful as the "crash sensors" are, but lets keep dreaming.

What else could be (passively) done to avoid an object? What clues are
there other than light? Is it just a matter of waiting for visual sensors
cheap and efficient enough?

I have seen some interesting programs that analyze input from cameras, but
can't imagine a beam approach to the same.



Home