Inline graphics on this page: 28K

 

Message from a builder

                       
From an email conversation between Dizzy's designer
and builder Han van Gelderen.
On 19 Dec 2001, Han van Gelderen wrote:

> Building him was no problem at all. And it took a little less
> time than I expected. So I can tell you that another happily
> bleeping little robot of yours has made it into this world :-)

Han,

Nice to hear!

> The building instructions are fine, though it would be nice to
> see a picture of a horseshoe crab.
Horseshoe crab   
More about this fascinating animal in KIJK May 2001, page 46-50. On the Web: www.horseshoecrab.org
> I followed the instructions except for the first charge of the
> battery. After glueing the motors in place, I built the dinner
> table. When that was finished, I thought I might as well charge
> the battery right away. So I connected the 330E resistor and the
> battery using a few bits of wire, en left it on for 5 hours. This
> has the advantage that Dizzy is ready to go when the building is
> done; no need to try your patience waiting for the battery to
> charge.

There has to be a procedure for charging a really flat battery,
when Dizzy has been stored too long. In that case, it has already
been soldered in place, and the method described in the building
instructions is safest. Providing two different solutions might be
confusing.

> In edz_bouw.htm you advise to lightly touch the folding lines
> with a knife, in order to make a sharply folded hood. I always
> use the point of a darning needle to enable a sharp fold. That
> point has a radius of about 0.4 mm, so you won't push it through
> the paper by accident, at least not as easily as a knife. You
> might use the point of a small screwdriver instead. Anyway, I'd
> never use a knife.

Good idea; I changed the instructions.

> A very minor detail is that the instructions never tell you when
> to push the 7400 IC into its socket. When I finished the charger
> PCB, I mounted it in its box and did all the wires. Then I
> compared the result with photo on my screen. Hey, there is a
> difference, the IC is missing from its socket!! Well, just to be
> sure (also because the controller is mentioned quite often) I
> checked the instructions, and didn't find that particular action
> in the text.
> No disaster of course, but it's something I didn't notice until I
> compared my food box with the photo.

Corrected. Maybe the instructions are perfect now :)

> Adjusting the food box was also quite easy. I hardly had to alter
> the setting of the variable resistor. And I didn't use a file on
> the LD271. I thought I'd first try it as is, and it worked fine.
> Dizzy has no trouble finding his food from any corner of his
> terrarium, sized 1 x 1 meter.

It will usually work well enough like that, but interference by
reflected IR light becomes more likely.

> I had to put a sheet of hardboard on the floor. The circular
> motions drag the silver wire castors more or less sideways across
> the floor, causing them to catch in a carpet.

A smooth surface also reduces dust trouble. My own Dizzy moves
about on a sheet of rubber. Very quiet and I rarely need to clean
his tires and drive shafts.

> Now that I've seen Dizzy in action, some questions come to mind.
> I don't know the thinking behind the program running inside the
> controller, or how reactions depend on information from the
> sensors.

Dizzy assumes that motion around him is less likely to escape
detection if he is in a dark corner, looking in the direction where
most light is coming from. That is the situation he tries to
achieve, as far as the position of his terrarium, the wider
environent and his limited senses allow.

> When I tease him a little by aiming a flashlight on the part of
> the terrarium watched by the LDR, he'll start to move (after
> beeping and bleating for a while). What is the purpose of his
> movements? Does he want to reach the light, or avoid it?

The first four steps are in the direction he was watching, as a
sign of `curiosity'. Then Dizzy will stop and look around,
searching for a dark corner, or a dark coloured object. In my
office, he often locks on to my denim pants. When he's hit a few
obstacles on the way, he will turn to the light, or towards a
bright colour. Then he waits for new activity in his field of view.

> I noticed that his motors run a little too long to keep him
> moving along a straight line. After changing wheels about four
> times (four steps, if you will), he ends up about 90 degrees off
> his original direction. So each step is larger than 180 degrees,
> more like 200 degrees. Is that as it should be??

Dizzy aims for a `step' of about 180 degrees. The timing is
corrected for the battery charge level, but also depends on the
motors' current consumption and the internal resistance of the
battery, which in turn depends on its age. A newborn Dizzy will
often tend to curve clockwise, while a middle-aged one should
usually travel in more or less straight lines.
As you can see, Dizzy will in any case have no trouble finding his
dinner table. When his IR receiver and the beacon are functional,
he reaches his food quite fast. And just like a real animal, he
doesn't allow a broken sensor to entirely defeat him. Dizzy will
try to feel his way to the dinner table, when he sees no IR light.
The prototypes can do that fairly easily.
A real surprise was the way in which two Dizzies managed to share
`food', as there was only room for one on the dinner table. After
fiddling about for a while, the second robot got one of his feelers
against number one's free feeler, while just touching the edge of
the table with a castor. The feelers of a Dizzy are connected, so
this allowed him to steal half the charge current...

> When Dizzy is switched on, he rotates around his top axis for a
> while. What is he doing with the information gathered??

He is looking at the levels of light and darkness around him, and
he will stop with his eyes aimed at a brighter area or colour. Then
he waits for about 20 seconds, allowing the person who switched him
on to put his `shell' back in place. And then he begins to monitor
his environment for signs of motion.

> When Dizzy begins to move, after a few changes in light level,
> he'll stop now and then to turn briefly around his top axis.
> What is he doing?

After every four steps, he looks for a `dark' direction to aim his
next four steps at. The algorithm sets a timer for 360 degrees of
turn. Every time when the light value sampled by the `eye' is
darker than the darkest so far, the timer is reset for another 360
degrees. Eventually, the timer clocks down to zero, with the eye
aimed in about the direction from which the winning light value was
recorded. Then another four steps are performed, with the eye
checking his own motion (to make sure he isn't stuck).

> By the way, perhaps a fun idea to increase Dizzy's options, if
> there is room:
> Give him a little ear, for him to react on differences in sound
> level. For instance, when you clap your hands, he might bleep
> back and briefly move to indicate that he was startled.

I'll put all suggestions on my list for the next version.

> When a sound persists, you might make him move towards the source
> when the level is between certain limits (50-70dB), and away from
> it when the sound is louder than the higher limit. I've no idea
> whether something like that can be done, if you can determine
> the rising or falling of the sound level within about 20
> centimeters of motion. Much will of course depend on the distance
> between Dizzy and the source of the sound.

Not much can be done with sound level alone, but the Sider Sound
System offers possibilities with specific sounds:

         http://www.xs4all.nl/~sbolt/e-spider_sound.html
           
> Anyway, I think the little robot is great fun, en he works quite
> well.

Thank you :)

Note that Han is rather overqualified, as builder of something
as simple as Dizzy. Have a look at his own pages, especially at
the crystal radio on:
http://www.xs4all.nl/~gelderen/

previous (assembling its food) / next (questions and answers)