a Kalman filter to fuse gyro and compass readings

[nxtreme]

Even better, do we know whether anyone else has made a compass sensor that doesn't have this limitation?

The Honeywell HMC6343 compass looks fairly easy to use, and it is tilt-compensated. You might also be able to use one of the other (cheaper) magnetometers that SparkFun sells, they just don't seem quite as easy to use. I have a HMC6352 that I got from SparkFun and it seems to work fairly well. The specs are quite similar to the HiTechnic compass sensor. In fact, does anyone know exactly what sensor the HT compass does use? It'd be interesting to know...

And no, I don't work for SF, just a happy customer .

[physics-matt]

I'm sure - no: I am convinced that your method does not achieve the required accuracy to me (estimated 0,1-0,2% navigation accuracy for a maze mapper and solver, size about 10m x 7m for all rooms on the floor, driving around about 1 hour at a go).

Well, it was never designed to. If you want to achieve this then two things are certain: first, you are going to need an additional position sensor and second, you stand more of a chance using a compass over odometry to get the bearing, though I think the best strategy would use both. For instance, I seem to recall someone on the nxtasy forums who built a robot that combined the compass, ultrasonic and odometry readings in a particle filter and got pretty good results.

Anyway, I'm glad we managed to vindicate the HT compass sensor.

@nxtreme: Thanks for the suggestions, I'll take a look.

Matt

[rghansen]

What I would like to know is exactly why the compass has such a strong dependence on its tilt (i.e. how far it is off being horizontal). Having a better understanding of the underlying physics of it may help to find better ways to compensate. Even better, do we know whether anyone else has made a compass sensor that doesn't have this limitation?

Matt

I assume it's because of inclination. The magnetic field in our part of the world point fairly steeply into the earth. This can cause large errors if your compass isn't level.

[rghansen]

Sounds like you are talking about SLAM (simultaneous location and mapping). It's well worth reading up on (wikipedia is a good place to start) and if you can get it working on the NXT then you'll get a lot of kudos!

Matt

Thanks for the SLAM tip.

[HaWe]

IIRC and IYRC the guy with the PF had a finished room model which was compared to US readings.
As my robot is moving in a first of all unknown environment, this virtual model has to be built up by moving through the rooms.
No virtual room model - no corresponding US echo model - no PF which got data to refer to and to rely on.

It's not that easy as you think it would be.

Your compass approach unfortunately doesn't seem to solve the navigation problems. But maybe aswin's KF will be able to.

BTW, "physics"-matt: have you finished studying physics? Have you already got your master degree or even a PhD? I'm not sure if I'm talking to a natural scientist or a student...

[physics-matt]

It's not that easy as you think it would be.

I never said it was easy!

I'm just trying to point you to work that I know about that may help you along the way to solving your problem, even though it doesn't solve your whole problem. If you would rather I don't try to help in this way in future please say so.

BTW, "physics"-matt: have you finished studying physics? Have you already got your master degree or even a PhD? I'm not sure if I'm talking to a natural scientist or a student...

I don't think a true physicist ever finishes studying physics, and as for my qualifications... does it make any difference? I'd rather persuade you by the strength of my argument than by the fact that I hold any particular qualification.

Matt

[HaWe]

Please calm down and quit arguing on these forums or we'll have to issue warnings, lock topics, and/or other dire actions. For the record, I have an undergraduate degree in Mechanical Engineering (never used beyond college) and work as a software engineer.

The original contents of this post have been deleted.

John Hansen

[linusa]

I don't think a true physicist ever finishes studying physics,

Top! That's the right attitude!

and as for my qualifications... does it make any difference? I'd rather persuade you by the strength of my argument than by the fact that I hold any particular qualification.

And right again. This is btw. one the fantastic character traits of R. Feynman. He was taught to "not believe uniforms" or degrees or whatever of that kind. When having discussions about physics, he was able to talk to his opponent freely, without being influenced about his position. He could focus purely on the actual "core physical question". Being able to talk to authorities in his field and not believing them everything they said just because they were famous or had a good reputation was one of his great strenghts -- see how it lead to his success...

I recommend to watch this google tech talk: http://video.google.com/videoplay?docid ... 1522818645 . The "open source" topic fits, I mean we ARE a sourceforge community now

[HaWe]

I am calm.
<content removed>[JCH: I am glad that you are calm. Please be less aggressive and annoying when you calmly post next time. Thanks!]

[aswin0]

I have my thoughts on the discussion compass versus odometry.

To my experience odometry very soon leads to bad results, getting worse over time. Compass based localisation immediately gives bad results, Results that, being translated to a position, also getting worse over time. My Kalman filter gives bad and good results mixed in time, the bads being small for some time, but still they add up once translated to a location. So, all methods are useless over time unless there is a recovery mechanism, a reference system, that can be used to calibrate the location.