[android-discuss] Re: Simulating a compass with bearing and accellerometer? - Peli - com.googlegroups.android-discuss

14 messages in com.googlegroups.android-discuss[android-discuss] Re: Simulating a co...

From	Sent On	Attachments
efon...@gmail.com	15 May 2008 20:22
efon...@gmail.com	17 May 2008 04:48
James	18 May 2008 05:47
efon...@gmail.com	19 May 2008 04:32
Alex Pisarev	19 May 2008 05:51
efon...@gmail.com	19 May 2008 10:13
Peli	20 May 2008 05:21
James	20 May 2008 05:47
James	20 May 2008 06:20
efon...@gmail.com	20 May 2008 13:27
efon...@gmail.com	20 May 2008 15:50
efon...@gmail.com	21 May 2008 04:42
efon...@gmail.com	21 May 2008 06:11
efon...@gmail.com	21 May 2008 15:09

Subject:	[android-discuss] Re: Simulating a compass with bearing and accellerometer?
From:	Peli (peli...@googlemail.com)
Date:	05/20/2008 05:21:03 AM
List:	com.googlegroups.android-discuss

I think it should be rather bearing and orientation sensors that you have to combine to simulate a compass:

The accelerometer can not indicate any direction except "down" (direction of gravity). Imagine the following thought experiment: Put a 3-axis accelerometer somewhere, and rotate it slowly around the z-axis (that points upwards). The accelerometer values will not indicate any change, since the direction for "down" relative to the device will always stay the same.

This will also not happen if you include 2 or more accelerometers into one device: Accelerometers can not keep direction.

What you need is an orientation sensor - made of gyroscopes. This could keep orientation information, but is larger and more expensive than accelerometer sensor.

Here is one example that combines accelerometer, orientation, and compass sensors: http://www.microstrain.com/3dm-gx2.aspx but this would be too expensive for phones. Probably, the compass sensor alone would be cheaper (but I'm just guessing here..), so that it could make more sense to simulate an orientation sensor using compass and accelerometer :-)

You can try it out in the OpenIntents Sensor simulator: "Yaw"ing will not change accelerometer values, but orientation and compass values. http://www.openintents.org/en/node/6

On May 19, 7:13 pm, "efon...@gmail.com" <efon...@gmail.com> wrote:

Thanks.

On May 19, 7:52 am, Alex Pisarev <alex...@gmail.com> wrote:

As far as I remember, OpenIntent's Sensor Simulator managed to emulate compass somehow, however, didn't look details.

Regards, Alex

On May 19, 3:33 pm, "efon...@gmail.com" <efon...@gmail.com> wrote:

Hi James,

This is helpful. Thank you. It is good to know someone is thinking about this. After reading the IEEE abstract, a little clarification on use is always helpful.

Our applications are eyes free, because the phone should not be competing for those revenue generating resources. So use will happen with the phone stored on the person’s body, say, in a shirt pocket, held to the person's ear, or on their belt. The compass needs to work no matter how the phone is oriented. The IEEE abstract seemed unclear on how it would work in a handheld device. It would be great for cars, though. With the two accelerometers, there is a calibration routine where a person wears the phone normally and walks in one direction. 10 ft should be enough to get enough calibration to be useful. It can recalibrate in the background and alarm or adjust ...

Sensor Fusion is a new word for me. Differential sensors of finite resolution have been around for a long time. Consider the roach, or any bug with antennae. They sample air at distant points allowing the organism to select a direction. The longer the antennae, the smaller the gradient the organism can detect with sensors of a fixed resolution. Long antennae help folks figure out what is going on. Sampling acceleration at distant points is going to give you better information on angular velocity and acceleration (how fast you are spinning) than using a single sensor in the same way.

Thank you for pointing out that radio field interaction can provide information... since a human body can influence that, it is probably good to not rely on that method. Two sensors a fixed distance apart should require little attention and provide good results across many devices once it is engineered.

My job is to show why it is worthwhile to spend that dollar for pedestrians who don’t read maps. Android has the tools to do that, even in today’s SDK.

On May 18, 7:48 am, James <jame...@yahoo.com> wrote:

A MEMS chip can be a collection of sensors (temperature, accelerometer (x,y,z), atmospheric pressure, Hall effect sensor (compass), ...) all built into the same chip. In mass production the chip could be relatively inexpensive.

http://ieeexplore.ieee.org/Xplore/login.jsp?url=/iel5/20/35967/017046....

However, an Android shortcut would be to use the GPS sensor and your relative direction of travel to produce a compass bearing over 100 feet of uniform travel. For each model of cell phone the antenna sensitivity changes as you rotate the cell phone about a point. This could potentially be tied in with relative position movement to estimate a compass bearing about a point.

But as I said, the lookup table would be different for each model of cell phone.

This type of engineering where you take two sensors with low resolution to combine their results to provide greater resolution is called "Sensor Fusion".

Basically a cell phone antenna signal does NOT have the same signal profile when you rotate left versus rotate right. This can be capitalized upon to determine the relative bearing of which the compass is facing. Coupled with the cell phone GPS the relative bearing can be referenced to the true bearing. A lookup table can provide a correction factor and thereby produce Magnetic Bearing; vis vi Compass.

James Dunn Table of Contents - Similar Insights related to technology applicationshttp://blog.360.yahoo.com/jamesbdunn?p=207

On May 15, 9:23 pm, "efon...@gmail.com" <efon...@gmail.com> wrote:

I have no idea how much a cell phone with a compass costs. I don't see why it would be expensive if there were enough of a market to micro machine it like all those tiny mirrors. > Who knows these answers?

Thanks, Ed- Hide quoted text -

- Show quoted text -- Hide quoted text -

- Show quoted text -