Make drone fly autonomously from point A to B

Question

I want to develop a autonomous drone which is able to fly from point A to point B in an indoor location. The drone starts at point A and with a button push it should fly to point B. The points have a distance of around 2m, the drone should land on point B with an accuracy of +- 20cm. There are no obstacles between the points.

I thought about using GPS, but GPS is way too inaccurate for an indoor localization. I also looked into UWB localization methods based on a DWM1000 IC. UWB does work with a good accuracy but it requires 3 anchor points which use the DWM1000 + one DWM1000 for the drone, so I would need 4 of those chips, they are quiet expensive.

I also looked into LIDAR and SLAM and I think this would work to make the drone fly autonomous from point A to B.

Do you know any other maybe simpler solution for the described task?

Answer 1

3d lidar scanners are in the thousands of dollars for low end. You probably want an rgbd camera like the real sense.

If your environment features are static and prerecorded a regular camera can be used but you won't get absolute distance.

Answer 2

Electronic gryo IC's with drift correction knowing the initial and final condition is zero velocity and short hop time are the way to go IMHO.

How you do that is up to you. e.g.

The position change is the integral of velocity change and the second integral of acceleration changes.

A calibrated gyro measures velocity in degrees per second (dps). So as long as the bandwidth of the acceleration is within the selected chip BW for some BW defined at -3dB and -40dB (?), the latter will affect accuracy because acceleration is the 2nd derivative of position error which increases the magnitude for error if there is a sampling error. Thus I might expect a tradeoff for BW and method of oversampling, decimation and resolution of the required dps measurement with SNR.

• where -3dB BW = 0.35/Tr (10 to 90%) and -40dB is 20 dB per order of magnitude on the slew rate if more than a 1st order response.

From my past experience, I used a carrier signal derived from an OCXO PLL with 1e-11 stability and mixed with dual antenna Rx to generate phase shift for range and azimuth in a 2D model tracking of Black Brandt rockets over a 500km height for recovery. Yours is a much smaller range ;) so harder and might require a much higher carrier phase-locked to an OCXO mixer. Although I found if the surrounding has no interference to VHF signals, a return loss bridge makes an excellent people intrusion detector from many many wavelengths away with increasing levels of cycling return signals with motion-sensing using a cheap RF diode and dipole antenna to a directional coupler measuring mV or return loss transmitted. You might think of a phased-array RX network around the room and detect the phase shift of the transmitted carrier around the room but likely would get unexpected reflections from objects here and there.

Answer 3

You could try something homecooked, like a very low power RF beacon (on a public frequency), or maybe UV/ultrasound, and simply maximise received signal strength while staying on a fixed horizontal plane. This would probably get you with 20cm and might be fairly simple as a first order solution. Of course, you might be looking to eventually scale your solution to a less controlled environment, it depends on what you eventually wish to achieve.

EDIT : actually you miss an important piece of information in your question : is the location of the target known to the drone in advance or not? or does the drone have to find it in a radius of about 2m? And does the problem you are trying to solve allow you to place a "marker" on the target? how is the target position actually defined?