############################################# Friday 7-7-2017
Put frame together.
############################################# Saturday 8-7-2017
Put motors, ESCs together. Connected Pi, Navio2, GPS antenna.
############################################# Sunday 9-7-2017
Bought galvanised steel M3 screws from Apple Tool & Gas. Lengths: 6mm, 8m, 10mm, 12mm. Attached flight controller, case, GPS antenna, 433 MHz transceiver to frame.
############################################# Wednesday 12-7-2017
Connected ACCST transceiver for handheld control, power module.
############################################# Friday 14-7-2017
Used some math to work out cartesian line to add 6 flight modes. Some of the numbers:
(982), 1166, (1238), 1295, 1425, (1494), 1555, 1685, 1814 18, 30.5, 43, 56, 69, 81 -64, -39, -14, 8, 38, 62
y/x = 25/100 = 25%
Switch A -> I5: y = 26x + 37 Switch B -> I6: y = 75x + 0
Channel 5 = I5 + I6
############################################# Saturday 15-7-2017 Flight attempt
Channel 2 (Elevation) inverted.
Added extra power module for redundancy.
Growing wobble of death. Investigated the timing of the ESCs. There is a slight delay, about 10ms, between two motors when running the motor test {16% throttle, x seconds}. I thought this must be an issue, but this is just due to timing constraints (I assume) when it comes to the real-time kernel in the raspberry pi, before it sends the commands to each motor via the 32-bit external co-processor / microcontroller.
############################################# Sunday 16-7-2017 First flight
I lowered the gain of the roll/pitch PID values by about 25%. Made the craft slightly less jerky, but enough that oscillatory transients were not present anymore. Smooth flying.
############################################# Monday 17-7-2017 Official commencement of Project E448
Met with Dr van Daalen today at 1pm for an ad-hoc meeting. We spoke about the approach, methodology etc with regards to the thesis. Two methods are a systematic approach, which is good, but slow. It can also lead to fixing the wrong problem. The other approach is focusing on the goal, and refining the process thereafter. The latter approach also leads to the discovery of unforeseen problems. One caveat is that subsections are incomplete, as opposed to the systematic approach.
Flew a test on front lawn in Alt-Hold and Stabilise. Flew a mission at De Hoop.
Mission:
- takeoff to 10m
- move west to waypoint
- land
At the transistion to landing, I discovered that the drone started toilet-boweling. I've seen this before, and handled the situation relatively well. Usually it increases the chance of a crash ten-fold (in my opinion). The greatest problem is that one can only go up and down.
After investigation, it seems that a fix is to attach an external compass, or (which I neglected after a compass re-calibration) a compass-motor-interference recalibration. This determines the magnitude of interference from the motors (induced by current) which affect the compasses.
The current compass-motor-interference calibration is set for previous compass offsets.
############################################# Tuesday 18-7-2017 First day of classes.
I still need to sort out wifi-reconnections, and to investigate why the 433MHz linkup is not receiving enough power. Range is antennuated to about 10m at most, and usually 2-3m.
I re-did the compass-motor test. This time I did a linear throttle increase, which resulted in a smooth non-linear result, instead of spiky when twitching / lowering the throttle. Interference peaks at about 30%.
When attempting a quick mission in the front lawn, the gps home was offset, which meant that the drone flew into the street. Luckily, I switched to Alt-Hold, and slowly brought the drone down at the other end of the road. Doing that in split second moments is hair-raising enough. In hindsight, perhaps putting it into Brake mode would pause it, but I'll need a quicker switch configuration to enable that.
############################################# Thursday 20-7-2017 First weekly skripsie meeting
Met with Corné, and finalised the aims for the project.
The aims are as follows:
- To investigate parameters within large-scale areas beneficial to the agricultural sector.
- To design and build an aerial observation platform in order to collect data and estimate these paramters
- Verification by physical measurement is an optional extension.
############################################# Friday 21-7-2017 Wifi-reconnect added
Tested the range of both 433 MHz HolyBro tranceivers. They have a receive sensitivity of about -130 dBm. The one sends a signal (from within a room) to roughly 700 meters (not LoS). LoS should be about 1.5km. The other one cannot get past 10m. This is using the same power supply. I verified this using an RTL2832 SDR. I spoke to FlyRobot
Spent many hours fiddling with wicd {cli, curses}, the interface file used by ifup / ifdown, wpa_supplicant.conf, iwconfig, ifconfig and trying to get wifi-reconnection to work. In the end, a rather custom interfaces file, and a custom bash script did the trick. It checks if it is connected to an ESSID every few seconds from iwconfig, and if not, it forces a reconnection with ifup --force wlan0.
############################################# Saturday 22-7-2017
Used up 1142mAh of 1300 mAh battery. Flew in front garden. All's well.
############################################# Sunday 23-7-2017
Connected camera. Extended flat 15-pin ribbon cable (200mm + 150mm). This took many hours actually, since it was only successful on the third variation. In fact, regarding the first variation, I didn't realise that the pins needed to be flipped.
############################################# Wednesday 26-7-2017
Got dronekit to work. Realised it only shows 8 channels.
############################################# Saturday 29-7-2017
Flew two sorties on the dehoop upper field. Was late in the evening so blurred shots, and I realised that I had accidently adjusted the lens focus.
############################################# Sunday 30-7-2017
Delicately refocused lens on camera.
Piped mavlink commands through mavproxy in such a way that ardupilot communicates with the GCS hostname, and not a manual IP address. It also communicates locally on two different UDP ports. One for ardupilot, and another for dronekit. Got rid of the HEARTBEAT mode 0 doesn't exist exception within dronekit from the GCS.
Discovered that CAM_FEEDBACK messages are related to the camera trigger by listening to a large stream of all the messages. Received them in dronekit.
############################################# Monday 31-7-2017
Windy day. Need to take photos using camera trigger. Successfully wrote code to trigger camera from mavlink messages. Photos were taken at the end of the day (6pm), and were lower quality due to ISO etc.
############################################# Wednesday 2-8-2017
Photos taken yet again at the end of the day in the front yard. Linking to gps data not always accurate.
############################################# Thursday 3-8-2017
Met with Corné, I set myself a goal of having IR photos from the air for next week's meeting.
Camera triggered upon mavlink message.
############################################# Friday 4-8-2017
Photos taken on De Hoop field. Most of the photos were in the dark, yet the first few were not bad. Sunlight really makes a difference.
############################################# Saturday 5-8-2017
Test flight in Cape Town.
############################################# Sunday 6-8-2017
In process of writing UDP linkup code.
############################################# Monday 7-8-2017
Unjumbled file names. Wait for connection from ardupilot / mavproxy. Mavlink restarts on IP changes. Screen daemon startup processes. Released uneccessary resources, show missed shots.
Implemented bash script allowing infrared photos to be taken simultaneously (via wifi -- temporary). Bash script is nodal, meaning that more cameras can easily be added, but for now it doesn't seem necessary.
############################################# Tuesday 9-8-2017
Time from ardupilot trigger via mavlink message. Notes: seems useful, but epoch time zeroed if internet not available upon boot. Solution: real-time-clock, synchronised via ethernet.
############################################# Wednesday 9-8-2017
Networking via ethernet. Faster trigger time. Notes: the CAMERA_FEEDBACK messages actually come through twice, with a 50ms difference. Instead of using time, a % 2 operation ensures no shots are missed. The problem is that since the camera captures are blocking, the message queue is blocked as well. If the mission requires too many shots, the shots will be out of sync. Solution, run the capture command in a thread, event based with a callback, and also implement a control system to pause at the requested position if the callback is taking too long. THe other solution can possibly be more overlap between shots.
Was going to take dual shots, but by the time I was ready to, there was cloud cover, which seemed to mess with the gps in the sense that it was badly offset by a few thousand kilometers. Also, in the spur of the moment, I had left the power supply for the INFRAPI back home.
Finally received the 433 MHz linkup again. Works sucessfully so far. Haven't yet range tested it. Camera platform not final, just testing, nevertheless, it is misaligned in orientation.
############################################# Thursday 10-8-2017
Met with Corné, nearly had IR photos at the ready. Weather was poor at the time. This time I planned to have IR shots at the ready by next meeting.
############################################# Friday 11-8-2017
Bad weather for IR shots (too cloudy).
############################################# Saturday 12-8-2017
Took test photos out of a window, during moments of slight sunrays between the clouds. Aligned dual photos by means of a SIFT, and landmark correspondences algorithm. Note: sunlight greatly improves quality of photos. However, horizontal photos are not necessarily ideal for scientific analysis, especially due to shadows.
############################################# Sunday 13-8-2017
Took dual infra-red photos, triggered from handheld transceiver. Unfortunately, the subsequent triggers are not synchronised, therefore they went out of phase. Processed useable NDVI photos.
############################################# Monday 14-8-2017
Spent the whole day fiddling with UDP, TCP solutions to essentially create a double-barrow shotgun which waits for both cameras to be ready before taking a simultaneous shot. The shots may be 20ms out, but it is sufficiently small to be satisfactorily aligned.
lapse
############################################# Thursday 7-9-2017
Discovered aerobotics. Will be handy.
############################################# Thursday 15-9-2017
Finally replaced faulty 433 MHz transceiver. Technician was poor and had sent to me three faulty ones previosly, so I went there to sort it out.
############################################# Monday 18-9-2017
Flew another few sorties with 4 new batteries. Have a dual battery setup where the one battery powers the on-board electronics, and can remain on while the flight battery is replaced. Saves some setup time.
The focus between the cameras differ and need calibration.
############################################# Tuesday 19-9-2017
Met with Chris Antonisson of Geoflight, and plan to fly together on friday. He has a calibration plate, besides a Parrot Sequioa, Matrice drone, Phantom etc.
############################################# Wednesday 20-9-2017
Calibration of focus for stereo camera setup using image feedback and a 2m ruler.
############################################# Thursday 21-9-2017
Built a stable platform for the stereo cameras to minimize any movement.
############################################# Friday 22-9-2017
Attempt to align cameras using video feedback. Fiddled with focus.
############################################# Saturday 23-9-2017
OpenCV. Attempt to install OpenCV for eclipse.
############################################# Sunday 24-9-2017
Eclipse route failed. Flight test. Due to short time, IR images were not captured. Attempt at stereorectification.
############################################# Monday 25-9-2017
Write report.
############################################# Tuesday 26-9-2017
Preparation for NDVI calibration.
############################################# Wednesday 27-9-2017
NDVI calibration plate.
############################################# Thursday 28-9-2017
Hand in chapter to Corne + meeting.