Where to start with GPS and accelerometer.

cbmeeks

I have a side project that just came up. I'm an amateur at best so please forgive any dumb questions.

While I've been using the propeller for some audio/video stuff, that has all been purely for my gaming fun. But I have another project (or potential project) that could turn into something more.

Anyway, I want to start slow but the general idea would be a drone that would relay its GPS coordinates and accelerometer coordinates. The drone would eventually do other things but for now, I just want to communicate with it. Imagine a little drone flying around a room and recording items on a shelf.

So I figured the best thing to do is simply get a GPS and/or accelerometer module, put it on a breadboard and walk it around for now. I can worry about the actual "flying machine" later on.

But, doing a quick search on Parallax and SparkFun indicate there are several GPS/accelerometer modules and they look pretty pricey. $50+ for something that is just a proof on concept (on my own dime) is pretty hefty. :-)

However, I'd be willing to go by the suggestions here and if a $$ kit is what I need then so be it.

OK, so, being a total GPS/accelerometer newbie, here is what I would like help finding if possible:

GPS and accelerometer "chip/module/breakout board"

Resolution good enough to know it's on the "left" side of a large room or the "middle" of a large room, etc.

Accelerometer resolution to know that it's at the bottom of a tall shelf or the top.

None of this is probably making sense. But, like I said, imagine a small drone who has a job of going from shelf to shelf and scanning objects.

Of course, propeller drivers for the hardware would be great too but I am also familiar with Atmel products.

Thanks

Publison

cbmeeks wrote: »

But, doing a quick search on Parallax and SparkFun indicate there are several GPS/accelerometer modules and they look pretty pricey. $50+ for something that is just a proof on concept (on my own dime) is pretty hefty. :-)


Thanks

Accuracy comes with a price. Cheap GPS's will not get a signal inside a house.

The new Parallax GPS:

https://www.parallax.com/product/28509

has pretty good accuracy, (<2 meters), but that is outside under ideal conditions. Inside, (if you can get a signal), is less than ideal, it has an accuracy of >2.5 meters, (7' or so).

That will not map an interior room to any accuracy.

User Name

The least expensive GPS I know of that is still worth buying is this one:

http://hobbyking.com/hobbyking/store/__31135__NEO_6M_GPS_Module.html

It is better with relative displacements than it is with absolute position. So as you moved around the house you would notice changes in Lat and Lon.

I don't think you'd find accelerometer data as useful as you imagine...such as in determining whether you're on the top shelf or bottom shelf. A double integration of accelerometer measurements to obtain position would be wildly subject to drift, for example. OTOH, they're pretty effective in determining orientation wrt gravity. So for yaw roll and pitch...

Publison

BTW, Altitude from a GPS can vary by 15 meters,nature of the beast.

cbmeeks

Thanks for the suggestions.

What would you guys recommend in the $50 - $100 range?

Also, that's why I was *assuming* that a accelerometer would makeup for the inaccuracies of the GPS indoors. I was thinking maybe the drone could register itself with 4 known spots.

For example, think of a shelf having four barcodes in the TL, TR, BL, BR corners. The drone would scan those codes and "mark" it self there. Then it would be a matter of divide/conquer for locating other shelves, etc. within those bounds.

Of course, finding the four markers is another problem but you get the idea.

Thanks.

Publison

cbmeeks wrote: »

Thanks for the suggestions.

What would you guys recommend in the $50 - $100 range?

Also, that's why I was *assuming* that a accelerometer would makeup for the inaccuracies of the GPS indoors. I was thinking maybe the drone could register itself with 4 known spots.

The Pam modules are one of the best out right now. See Parallax item I quoted.

The accelerometer can not compensate for the inaccuracies of the GPS, the data from the two do not compliment each other.

prof_braino

Even if your room is 30 meters on a side, the GPS will only tell you that you are someplace close to there, in relationship to the enitre world. So for flying around a room, maybe GPS is like using a telescope when you want a microscope. Instead of GPS, maybe you want an interial measurement system. This might be the accelerometer and gyroscope combination. At least this is the path I took.

The cheapest experiement might be done with a 9DOF or 10 DOF IMU module from ebay. There are not the greatest, but they are the cheapest. Once you get one of those figured out inside the room, then look into the fancy GPS offerings, for outside in the park.

JasonDorie

An accelerometer reads instantaneous accelerations due to external forces, like gravity, being moved or bumped, and so on. You *might* be thinking of an altimeter / barometric pressure sensor, which measures air pressure and can be used to calculate altitude, often to a resolution of 10cm.

cbmeeks

Ah, yes, those are good ideas. I was really thinking it would be a mixture of these technologies to get accuracy.

Oh, and the room size is a little bigger than 30 meters per side. It's 1.5 million square feet. LOL

GordonMcComb

Inertial systems can have unbounded error. That is, the error grows and grows until it's corrected by some system that can provide a waypoint. Any system you develop should have one or more methods of determining physical waypoints, similar to lighthouses on the coast of the sea. You should use a robust interial sensor that includes a compass, so that you can re-orient yourself to a known direction in the room. The 9DOF IMUs have such a compass (3 axes each for accelerometer, gyroscope, and compass). Many flight boards for quadcopters already have such a sensor.

There are methods, now fairly well discussed in other threads here and on the Internet, of using various types of beacons for finding a position within a defined space. I believe Pololu has such a system, though I don't think it's $50-100. You need ideally three, beacons to locate yourself within a grid. Since you're flying, your work is made more difficult because you no longer can rely on simple plane geometry to triangulate your position.

Another method, more complicated but more flexible, is a "navigational star" projection onto the ceiling. A camera on your drone pointing up captures the projection of dots in a particular pattern. The angle and position of the dots can be correlated into position. I imagine you'd need the drone to fly level for each reading; the IMU can be used for that.

If you have a CMUCam or other video camera/image processor, you could perhaps "teach" your drone to recognize objects in your room. Doorways and windows are a good place to start, and work well if your vision system can detect outlines (some of these are just simple color blog analyzers). For ease in programming, you might need to fly a specific height. Here's where a couple of Pings -- one mounted under, the other on top -- that report distance to floor and ceiling, assuming 8-12 foot ceilings. These are well beyond the $100 range.

cbmeeks

GordonMcComb wrote: »

Inertial systems can have unbounded error. That is, the error grows and grows until it's corrected by some system that can provide a waypoint. Any system you develop should have one or more methods of determining physical waypoints, similar to lighthouses on the coast of the sea. You should use a robust interial sensor that includes a compass, so that you can re-orient yourself to a known direction in the room. The 9DOF IMUs have such a compass (3 axes each for accelerometer, gyroscope, and compass). Many flight boards for quadcopters already have such a sensor.

There are methods, now fairly well discussed in other threads here and on the Internet, of using various types of beacons for finding a position within a defined space. I believe Pololu has such a system, though I don't think it's $50-100. You need ideally three, beacons to locate yourself within a grid. Since you're flying, your work is made more difficult because you no longer can rely on simple plane geometry to triangulate your position.

Another method, more complicated but more flexible, is a "navigational star" projection onto the ceiling. A camera on your drone pointing up captures the projection of dots in a particular pattern. The angle and position of the dots can be correlated into position. I imagine you'd need the drone to fly level for each reading; the IMU can be used for that.

If you have a CMUCam or other video camera/image processor, you could perhaps "teach" your drone to recognize objects in your room. Doorways and windows are a good place to start, and work well if your vision system can detect outlines (some of these are just simple color blog analyzers). For ease in programming, you might need to fly a specific height. Here's where a couple of Pings -- one mounted under, the other on top -- that report distance to floor and ceiling, assuming 8-12 foot ceilings. These are well beyond the $100 range.

That's what I'm afraid of. The "well beyond $100 range" seems to pop up quite a bit in this field. Not sure the pings would work seeing that the floor to ceiling is probably 40-50' or more. But I can't believe the ideas that are swarming (no pun) in. I didn't think about the navigational star method but I like that.

One thing I was also aiming for was to be able to fly in the dark. Some of these buildings aren't 24/7 so they power down at night. Perfect time to collect inventory while the lights are out.

Anyway, lots to think about.

Thanks

GordonMcComb

This Wiki page covers many of the most common methods related to wireless for interior location finding:

http://en.wikipedia.org/wiki/Indoor_positioning_system

In addition there are the vision analysis and other methods I briefly noted.

For the Ping, you wouldn't need one for the ceiling if you always get within 6-8' or so of the floor in order to take a picture snapshot. You just need to be more or less at the same plane in order for the camera to reliably recognize the major objects in the room.

cbmeeks

GordonMcComb wrote: »

This Wiki page covers many of the most common methods for interior location finding:

http://en.wikipedia.org/wiki/Indoor_positioning_system

I wouldn't say any of them are cheap.

For the Ping, you wouldn't need one for the ceiling if you always get within 6-8' or so of the floor in order to take a picture snapshot. You just need to be more or less at the same plane in order for the camera to reliably recognize the major objects in the room.

That was a very good reference. Thanks!

I particularly enjoyed the magnetic positioning. Seems like that would be an awesome problem to work on. :-)

JasonDorie

A number of universities use "closed systems" that are basically motion capture systems to figure out this stuff. If your intention is to only ever use this in known spaces, you could do something similar. A WiiMote has an infra-red camera in it designed for point tracking. It will output the x & y coordinates of up to 5 points.

A couple of those and an infra-red emitting diode or two on your quad and you could use an external system to locate it within the space, and transmit the rough location back via XBee or something. It's crazy enough it might work, and it wouldn't cost thousands to implement.