>
> Raymond Skarratt u wrote:
> >
> > We can probably get <10cm accuracy with ultrasonic beacons, depending on
> > humidity and temperature variances. I think that the idea of having
>
> Surely we should exhaust all possibilities of onboard sensing before we
> ever contemplate beacons or any other artificial modification to the
> environment.
Well like everything else it's a tradeoff. Complete onboard sensing may
require a significant investment whereas an external source may greatly
simplify some problems you present below. We should examine the sensing
problem from one end of the spectrum (complete onboard) to the other
(completely mapped and beaconed) so that we can understand the tradeoffs.
>
> We need to build a base first, then build the sensing system bit by bit.
> Sovling the simplest problems first and building on those foundations to
> solve more of the problem.
Agreed. I'm really uncomfortable when it comes to the mechanics of building
such a beast. Sensing and control electronics and programming are trivial
compared to the three basic items any project any of us undertakes requires:
infrastructure, mobility, and cutting. I really wish we could focus here
and folks could throw out some cookbook recipies for us to try out.
>
> Beacons sound nice when you first consider them, but all they really do
> is move the problem from the robot to another location. Worse, they are
> extra entities that bring their own problems. An active beacon is a
> major undertaking: power needs to be run to the beacon, the beacon needs
> to be installed to certain specifications, a method must be implemented
> to ensure that failure of the beacon is reported in a timely fashion
> (don't want Mowbot rolling off to regions unknown). Failure can take
> many forms, simple power failure, frequency failure, even failure of the
> transdure itself.
Dave, while beacons do bring new issues to the party, they are not that
difficult to solve. For example I'll take a quick swipe at two of the three
you presented above: power and failure.
Power: Rechargable battery contained within the beacon.
Failure. Mowbot stops and signals alarm when a beacon is lost.
While I agree that beacons are probably a last resort, their issues are not
insurmountable.
>
> I believe the stages should be:
>
> 1) Onboard sensors
> Until all possibilities of solving a particular problem are
> exhausted.
Again I'm not so sure. Onboard electronics can solve a lot of problems, but
at what cost for load, power, and intelligence? Let's develop a tradeoff
curve which designates what it takes to solve particular problems onboard. Then
each of us can choose a point on the curve we,re comfortable with and implement
it.
> 2) Passive beacons
> fences, wires, posts etc, that the existing onboard sensors can use.
Agreed. All existing boundaries should be utilized as they are relatively
easy to sense.
> 3) Active beacons
> when all else fails.
Once again, active beacons can play a part but we need to develop an
understanding of the costs versus the benefits.
>
> Any talk of positional information (like XY co-ords, multi-beacon
> referencing) creates another problem; They imply that the robot carries
> a detailed co-ordinate map of the environment. That means that before
> the robot ever moves, a detailed survey of the property must be made.
Not exactly. From the beginning I've been thinking of active beacons as
extensions of the passive beacons, their primary job is to artificially
define "do not cross" boundaries for the Mowbot. Positional info is a bonus
and need not be mapped a priori. The Mowbot can internalize positional
info to indicate areas it has already covered. So instead of depending
on the "seek for uncut grass then flood fill" algorithm solely positional
information recorded during the session indicating what has already been
cut can be eliminated from the search. It can also help the Mowbot to define
a finished condition, when all the areas without the boundaries (natural
and artificial with the active beacons) has been mapped as cut, it's done
for now.
> These are in my opinion just more entities that do little to add value
> to the project, in fact they are good reasons for not pursuing it. If,
> to get a robot lawnmower going on my property I had to:
> 1) build the robot
Well that's a given.
> 2) perform a detailed survey of my property
Depends. If the benefit is several years worth of not having the cut grass,
I may undertake this. Personally I think it's unnecessary...
> 3) install beacons (at considerable expense and modification to my
> property, ie digging cable trenches, installing concrete bases for the
> beacons)
> 4) ensure the security of said beacons from weather damage, vandalism
> etc.
> I would consider that an unreasonable venture.
For 3 and 4 I think we have different perceptions: initially I was thinking
infrared posts that were battery powered and mobile. Simply stick it in
the ground where you wanted an artificial boundry and be done. You've given
me a bit more to think about.
I too am not willing to go beyond sticking it in the ground and turning it on.
The sensor can even be smart enough to turn itself off until it hears the
Mowbot talking to it.
And just a cute thought: Any reason the active beacons could not be mobile
and roll themselves to the right spot at grass cutting time? Then when the
Mobots (I'm thinking more than one now ;-) are done, then whole team rolls
right on back into the shed? Idle speculation mind you.
Active beacons what do not create any a priori situations and don't require
massive reworking (or mapping) of the property could be useful. Let's not
dismiss them summarily.
BAJ