This video shows me flying a Syma X5C quadcopter, with the factory closed-source flight controller replaced by a Quanum Pico 32-bit brushed FC. I followed the directions in the "Visible Drone" chapter of this excellent new book, but, lacking the eye/hand coordination required to do surface-mount soldering, I used a Quanum Pico instead of the Teensy-based FC shown by the author.
A simple 3D-printed mount allowed me to attach the new controller with some 3M VHB tape. 1.25mm picoblade connectors, available cheap on ebay, connect the FC to the brushed motors on the Syma. I flashed my own firmware on the board, using PIDs from a 3DFly/Pico vehicle I'd built.
In addition to having a Spektrum DSM receiver onboard, the Quanum Pico has built-in Molex connectors for attaching I^2C and UART devices. So I'm hoping to use the modded Syma drone to experiment with proximity sensing, optical flow, and the like.
Comments
Mouse optical-flow sensors are useless for MAVs. Believe me, I've tried. The APM community gave up on them pretty quicikly, too.
@Simon, you got my brain working overtime... What about getting back to the roots ?
https://github.com/ArduPilot/ardupilot/tree/5ddbcc296dd6dd9ac9ed631...
Can get the mouse optical flow for 10$ " bangood , add the https://www.pololu.com/product/2490 and it makes a lightweight ( in terms of price ,mass and processing) Optical Flow test bed ?
Great !!!
Got a couple of 120 - 140 & 180 quads ready for the task, all flying on F3 & cleanflight.
I will get into BreezySLAM, one I'm finished with the Garmin Lidar_lite v3 & PX4FLOW integration on my BBBMINI.
This Garmin Lidar is quite an impressive beast, but I have to rewrite the driver...there's no free lunch ;-)
@Patrick: yes, obstacle avoidance is second on my to-do list, right after altitude-hold. With its floating-point co-processor, the F3 is better suited to such tasks than the F1, and with over twice the clock speed than F1/F3, the F4 is overall the best choice at this point. (I believe an F7 flight controller is in the works.) For both tasks, I want to use small, lightweight sensors like the MB124XX line from MaxBotix and VL53L0X proximity sensor. None of these processors would however be fast enough to do anything useful with lidar, such as SLAM. As you can see here, many single-board computers, with Gigahertz-speed processors, aren't even up to the task.
Once I've got something working with a few of the inexpensive proximity sensors, I'm hoping to work with Geof Barrows on adding one of his optical-flow sensors to the mix.
@Simon yeah right. Are you planning on implementing oboject avoidance within the F3 or F4 ?
I guess there is enough memory & processing power available to do basic lidar stuff ?
I was about to tell you about Geoffrey L. Barrows , and I just saw that he made a speach at your place ... So one day we might benefit from a collaboration between You and Geof ?!?
@Patrick I bought two of the CJMCU controller/frame units but was unable to flash my own firmware on them. These days I'm more interested in developing for the F3 and F4 boards anyway.
@Simon, did you tried the CJMCU controler ?
I have one here , but I never been very successful making it perform with cleanflight, this controller was slow and sloppy, felt like there was too much lag in the process loop. Might give it a try with hackflight..
@Patrick: Don't you know I'm Loco? ;^)
Cool suggestions, but I'm trying to avoid beacon-based systems, focusing instead on how flying animals do it (sonar, optic flow, etc.) E.g., https://www.pololu.com/product/2490
Cool, another hackflght project !!
Next logical step would be a localisation system like :
Position Tracking using HTC Vive's Lighthouse : https://github.com/ashtuchkin/vive-diy-position-sensor
or
Crazyflie TDoA Loco Positioning Sytsem : https://www.bitcraze.io/blog/
These would make ''the perfect lab'' ;-) .Keep on the good work Simon