CompassSensor issues

[Kaye]

Hello all,

I've recently started working on a project focusing on swarmed exploration. Right now I am in the very early stages, but am having some trouble with the CompassSensor (I'm trying to avoid the CompassPilot since it has been deprecated, I am using the 0.9.0 build).

The issue I'm having is that the sensor doesn't seem to be behaving at all as it should. This is my calibration method, which seems to work correctly:

Code: Select all

private void calibrate()
   {
      compass.startCalibration();
      pilot.setRotateSpeed(18);
      long start = System.currentTimeMillis();
      pilot.rotateRight();
      while((System.currentTimeMillis() - start) < 40*1000);
      pilot.stop();
      compass.stopCalibration();
   }


The issues that I'm having come after this step. If I call

Code: Select all

compass.resetCartesianZero();
float degree = compass.getDegreesCartesian();


degree will be some nonzero value. If I turn the robot by 90 degrees using a DifferentialPilot, the Cartesian degree value will change by some seemingly random interval, usually around 140 degrees.

Is this the expected behavior and I'm somehow misunderstanding, or am I doing something incorrectly? I could post the full code here, but there is a rather large amount of irrelevant code that isn't really relevant.

Also, the documentation mentions using a CompassSensor along with a PoseProvider, and I've been a little bit confused on how one would accomplish that. Can anyone shed some light for me?

Thanks for any help,
Jordan

[Aswin]

Are you familiar with the compass sensor? I ask this because it is easily disturbed. so the problem might as well be hardware as software.

[Kaye]

Thanks for the reply!

I am not at all familiar with the sensor, or with NXT in general. This is actually my first time working with robots of this nature.

Is anything specific I should know or some guidelines that could help me to avoid this problem?

[gloomyandy]

Readings from a compass sensor can be very heavily influenced by motors and other metal objects, and the sensor needs to be level to operate correctly. I would recommend that you create a small test program and use an NXT with only the compass sensor attached to see how it behaves in your environment. The readings you are reporting do seem a little odd, so I would try and simplify things as much as possible and make sure you fully understand what is going on. If you are not familiar with the NXT then you may wan to spend some time browsing this forum looking at common problems that people run into. In particular if you are displaying data on the NXT screen and using that to judge the output of the sensor you may want to read this thread:
viewtopic.php?f=6&t=2937
which describes a problem that a lot of users have when displaying data on the LCD...

Andy

[Kaye]

Thanks for the input Andy.

After some frustrating testing, it appears that the issue is in fact interference from other sensors/the NXT itself. Can you suggest any kind of configuration that may solve this issue? It is a bit difficult because my project requires 2 motors, an EOPD, an ultrasonic, and an infrared sensor in addition to the compass. Should I just find some way to attach the compass at a distance from all of the other metal components?

Jordan

[Aswin]

Jordan,

The compass is a difficult sensor to use. It should be placed at least 15 cm from the brick and motors.
Even then you should be aware of iron in the environment that might disturb the compass. Speakers, furniture and concrete, or rather the iron within the concrete, can disturb compass readings.

Some compasses do allow for calibrating. Calibration can help to overcome some disturbances. But don't expect miracles from it.

[kirkpthompson]

Hi.

Have you tried quantum magnetic shielding?
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=1062382&tag=1

Best,
-K

P.S.

[Aswin]

Well, I did. But it also shielded the earth magnetic field. Also, it is not available in dark bluish gray.

[Kaye]

Aswin,

Thanks for the 15 cm benchmark - I will try to find a way to build the robot keeping this in mind tomorrow and report back with the results.

Jordan