lets assume this non-flight scenario:
There is a street with supermarkets on the right and left.
I am walking down the street and into each supermarket, then down all the aisles and on to the next supermarket.
Obviously there is no GPS reception inside.
But I have a GPS lock when walking to the next supermarket.
Any chance to map the aisles inside the supermarkets using dead reckoning?
What hardware to use?
Ardu Pilot Mega with Oilpan?
I am a beginner with all things UAV and looking for advice what to look into.
To my knowledge commercial solutions start at 2000+ $US and are out of the question. Tell me when I am wrong.
If it helps I can roll a wheel to measure the exact distance traveled.
At first, I thought you were building an AUV since I couldn't picture anyone flying in and out of buildings.
I'm fairly ignorant but it seems like you could get the job done with a 3d compass, a 3 axis gyro, a GPS, and a pedometer.
If you already have the arduIMU, it would make sense to just use that and get a 3d compass if you don't already have one.
If you don't have either, I'd consider using an arduino, magnetometer, and a sparkfun 3 axis gyro.
If you think you may eventually get involved in autonomous robots, the oil pan + DIY Drones magnetometer would be the "cry once" solution.
List of arduino shields that will allow you to log data:
Indoor guidance can be reasonably accurate with a magnetometer and a surveyors wheel (trundle wheel). If you need to be more discrete, you could substitute a pedometer for the trundle wheel but with less accuracy unless you are very well practiced at walking with a constant pace. If you are able to be *less* discrete, you could do without the magnetometer and use a variation of the "south pointing chariot" to keep direction.
ArduIMU is going to drift way too fast for this application.
As you imply, expecting GPS indoors is not a good idea. But, one never knows. I'm an old timer and remember when we had single channel hopping receivers that took 20-30 minutes to get a lock under a blue sky. Yesterday I had a MediaTek maintain (maintain, not obtain) lock inside of a concrete block building with a steel roof. Blew my mind. Probably depends on obscure things such as how much water and iron are in the concrete, lucky window positioning, maybe if/how the roof is grounded. Unfortunately I was not in a position to log data and see if there was a position jump from multipath or other issues.
To be honest I am looking for something like an affordable alternative to the 2000 $US Honeywell DRM 4000.
It integrates GPS and dead reckoning and works outdoors as well as inside buildings, with walking and without pedometer or wheel.
The hardware seems to be comparable to an Ardu Pilot Mega with Oilpan and extra compass board?
http://www.magneticsensors.com/products.html (On bottom of page)
I'm sure that you are. But accelerometers that can do that are why it costs $K.
How often does the Klaman filter deliver it's output on ArduIMU or APM with Oilpan?
What do I get? 40 Hz?
I like to use it to record roll and tilt data using a LogoMatic SD card logger.
There is *no* Kalman filter in the ArduXXXX system.
It uses DCM.
The difficulty is the fusion of sensors. It's easy to make a navigation system with GPS telling you where you are, but the moment you start judging your position based upon other sensors, you start accumulating error. Allow me to lend you an example.
Suppose you're mapping a supermarket with a GPS that works in the parking lot, accelerometers, a pedometer, a compass and gyros. Suppose the device that contains these sensors is all combined in one case, thus barring misalignment of sensors.
When you walk into the store, you lose GPS signal. Suppose the metal racks in the supermarket create sufficient error in the compass to allow you to have a 1 degree heading error. For every 100 yards you walk, you can acquire about 5 feet of error to the left or right. The pedometer will indicate velocity, which will constrain accelerometer runaway while you're in the store. Once you get out of the store, the lack of velocity and accelerations will cause the position error to not disappear so quickly when the GPS signal reappears.
The mechanism to fuse all of these signals together which is most commonly used is the kalman filter. I've done some work with this filter, however, it's a beast to compute. You'll have to be very good at optimizing to get it to run on a mobile device.