I have a quadcopter equipped with PX4FMU board. You may download its datasheet from HERE. I wonder whether it is possible to program the quadcopter to autonomously follow a path like circular motion without any human interference. Are the built-in sensors enough for this task? I also wonder how accurate the built-in GPS is? I read that it gives coordinates with a radius of 5m as error.
$\begingroup$ The answer is going to be "yes, with some degree of accuracy that may or may not fit your needs". What are those needs? $\endgroup$– IanJan 21, 2014 at 16:36
There is everything on that board that you'd need for basic autonomous flight:
A gyro (MPU6000) + an accelerometer (L3GD20) + a compass/magnetometer (HMC5883L) to measure your roll, pitch and yaw angles and a barometric pressure sensor (MS5611) to measure the altitude.
There is no GPS sensor though - Only the connector to use easily add one. You're right though, GPS will only give you 5/10 meter resolution (but that's not due to the sensor being in accurate, but rather the technique/method itself - Don't forget that it's using satellite in space (!) to do that measurement, so it's pretty good really!)
What that means is that GPS is only good for large scale flight path. Achieving autonomous motion like this is a lot more difficult of course (in the video, they are "cheating", as they have cameras up in the ceiling watching down and tracking exactly where the quad is located).
The board comes will lots of other goodies such as a buzzer, GPIO ports... but if you don't care about these, you're probably better off getting each sensors separately.
$\begingroup$ There actually are GPS sensons that are much more accurate. I heard of it from this Kickstarter project, which claims an accuracy of 4 centimeters. $\endgroup$– danijarJan 20, 2014 at 20:52
$\begingroup$ Well, a trajectory is actually a set of points in space which are to be followed. Since a point is represented by x,y and z, we need to know the position of the quadcopter to allow tracking these points and apply adjustments if it drifts (feedback control). Knowing roll, pitch and yaw angles don't give information about the position, do they? $\endgroup$ Jan 21, 2014 at 4:50
$\begingroup$ That's right... which is why the guys in the video have mentioned in my answer have cameras on the ceiling watching the quads from above (I think they have IR dots on them, so the infra red cam can track them). That's one way to do it I suppose, but I'm sure you'll agree it's difficult to replicate and, of course, only works indoors. $\endgroup$– dm76Jan 21, 2014 at 9:05
$\begingroup$ The only other way I know of (if anyone knows of a clever idea, please shout!), is to use "visual servoing" and use a actual camera to navigate. Similarly, there are so-called "optical flow" sensors that can help: unmannedtechshop.co.uk/px4flow-cam-optical-flow-sensor.html For indoors, you could also look at using Ultra Sound proximity sensors to detect the distance from the walls $\endgroup$– dm76Jan 21, 2014 at 9:11
I think the answer depends upon the degree of autonomy you require. If, by "fully autonomous route", you mean just the ability to "autonomously follow a path like circular motion without any human interference"... then yes, that kind of function has been built into quadcopters with even less processing power and with almost identical sensor suite. But if your definition of autonomous route following widens, it is hard to say.
Btw... most GPS typically have an error radius of +/- 5m.
Has anyone tried hooking a soloshot camera base to a quadcopter? That could give you accurate mapping of an area without requiring a GPS. Just a half a minute of scanning with a camera and the base can track anyone with a tag. Find some way to take that data and map flight paths would be amazing.
$\begingroup$ "Find some way..." - Do you have any suggestions as to how to achieve that? $\endgroup$ Dec 27, 2015 at 22:48