27 months later…

It’s been 27 months since I first mentioned a quadcopter project on this blog.  And finally, I think I can say this initial phase of this project is finished and as good as it can be.

There’s a few bits and pieces I should do with HoG just to tidy up:

  • I want to get a longer video – probably at the local park – to prove how well behaved she is.
  • I’ll try her with flight plans that involve horizontal movement but I’m not expecting any surprises there – with motion processing, a hover is just a special case of movement – if hover works, then so will movement
  • I may add the ability for her to fly in circles – that’s a little different as it involves a fixed angle along one axis and a fixed velocity along the other – but again, other than calculating angles and speeds to give the size of circle required, there’s little else to do.

I’m struggling to get excited about using the magnetometer / barometer to refine orientation and altitude – I don’t think she needs it for the purposes she’s designed.  Likewise, I’m not bothered about adding GPS.  But I do have everything I need should I change my mind.

I will probably add the laser tracking as this will give me the ability to control her:

  • 2 lasers on opposite arms pointing at the ground but slightly inwards used to attain and maintain a fixed height
  • a third laser in my hand to guide her around at that fixed height.

I hope to be able to show her at the next Cotswold and Cambridge Jams but I’ll need a lot more flights between now and then to be confident she won’t misbehave and fly into anyone.

As for now, I’m feeling a little deflated, demotivated and just a bit tired.  Time for a break.

13 thoughts on “27 months later…

  1. Well I think you done an amazing job. Had you asked me two years ago if it was possible to perform an autonomous flight using (cheap readily available) inertial sensors alone, I’d have said “no”!

    • That’s the advantage of starting from complete ignorance and sheer bloody mindedness courtesy of my parents’ Yorkshire genes! I didn’t know it couldn’t be done with just a Raspberry Pi, Linux, Python and some sensors so I set out to prove it could.

    • Ultrasound / barometer? Too conventional – it has to be hard to be engaging and fun! Not sure yet about the double lasers for height, may give in to the altimeter yet, but the idea of taking her for a walk around a room on a laser leash / lead seems cool to me! What type of lasers? Just the standard laser pointers (those 1mW ones from China) – It’ll probably only work indoors with even ambient light, but still it’d be cool to take a quadcopter ofr a walk!

  2. Hey, congrats on achieving this! I’ve been rather quiet on your blog for a while (due to house move + new baby arrival) ; I remember you struggling for a long while with stability and calibration… What would you say has been the thing(s) that sorted it all out?

    • Biggest thing? Umm, so many increments – the roller-coaster started it’s final descent 6 months or more ago with understanding of the rotation matrices. From there, code performance to capture all the data, both in the python code and the underlying GPIO library. Most significant was the poorly described (in the data sheet) zero-g calibration which is a lot less sensitive to temperature than 1g I was using, and possible to do in 30 minutes with winter temperature differential indoors compared to out. Probably a few more other bits and pieces I can’t remember as the roller-coaster was going pretty fast towards the end. That why it’s such a blunt finish when the ride ends so rapidly.

      • P.S. Congrats on the new house, and especially the baby. Alton Towers or a Quadcopter can’t compete. Have fun along the way – I’ve a son of nearly 7 and a daughter of nearly 4 and they are both the best things that have happened to me but in a very life-changing way – roll with it and enjoy the ride!

      • Actually Dave, scratch all that – the most significant step was the addition of the Butterworth low pass filter to allow me to separate gravity from net acceleration, so regardless of temperature, and other sensor inaccuracies, a hover was always zero net acceleration. The rest in comparison were all refinements.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.