09:03am - GSoC and Olivier
Quite a bit of activity on Gitter
Sebastian is working on smart-RTL to find a safe path home
Sebastian’s update:
Two options
Breadcrumb
fly path in reverse
After cleaning up
Occupancy grid
Define gridpoints as safe or unknown
Breadcrumb is most likely to work on an embedded system
Back-of-envelope calculations seems to show occupancy grid using two many datapoints
Prototype written in Python
Clone and play! https://github.com/squilter/safe_rtl_viz
Visualiser in Python https://www.youtube.com/watch?v=sQf46iinOi8
Started to reimplement in C to embed
Need to make sure it will actually work on the board
Q: When do calculations happen?
Q: Which thread?
100 crumbs, add once per second
Pruning hard done when buffer full or at RTL
Should probably be done in io thread
Vehicle wants to RTL and RTL isn’t ready? Complicated callback mechanisms would be bad
Kind of like terrain which does calculations in io callback
Pierre is working on Rover reference model
Funding request in for hardware http://discuss.ardupilot.org/t/funding-for-gsoc-student-hardware/18394
Maker-Faire rover on its way
Donation receipt to be sent to Nick
Huge PR made
Rover motors library https://github.com/ArduPilot/ardupilot/pull/6461
Miguel who is working on sensor-head
Tridge has met a bit
Good progress
SITL implementation in next week or two
Sync up two process of the SITL simulator (plane)
TCP sockets
“Master” controls sensor hub
Will probably change to something like “producer”/”consumer”
Tridge says initial implementation should be a stand-alone-firmware but later the sensor producer should have a flight stack too
Low-latency is key!
Sriram is working on big, loud helis
Working on UAVCAN EFI https://github.com/ArduPilot/ardupilot/pull/6438
Siddharth
Moving along gradually
Parses logs in javascript
Suggestions made on data structures to make it memory and computationally efficient
Hamsa
3D interface for SITL
No code yet
Update in next day or two
Anastasia
Is waiting on hardware
Test boards on the way
Working on RPi and PixRacer in meantime
09:33am - Contribution of the month
Jaime for MakerFaire
Shared with Nick
09:34am - Philip and intel
Intel are dropping Edison, Joule and Galileo http://www.cnx-software.com/2017/06/19/intel-issues-end-of-life-notices-for-galileo-galileo-2-edison-and-joule-boards-modules/
Dropping the entire market
Just after a lot of people put a huge amount of effort in MakerFaire
Many people hurting on this
Philip’s done a huge amount of work on carrier boards for these devices recently
Philip sent a please explain email
Intel Aero?!
Nobody we know can sensibly comment on this ATM
TX2 carrier boards?
Jorgen
09:40am - Tom and intel
Firewall issues for people in intel who would like to join
Could open Mumble on a different port
Experimentation later this week
09:43am - Randy and Copter
-rc8
Some outstanding issues
Dsmx receiver-not-working issue
Turnigy 9xr pro doesn’t turn on
But Randy now has hardware
UAVCAN baro issue needs to go in
Complaint in about waypoint yaw behaviour
Erle-brain still crashing
LiteWare Lidar on serial connection has longer delay than when using i2c
Autotune leaning http://discuss.ardupilot.org/t/autotune-fails-on-y6-3-5rc7/18457/2
Switching in pitch and roll during tune?!
Pitching to get to target and twitching for roll at same time?
Leonard has a PR in for it
Randy is generally happy
Not sure if there’s anything critical enough to hold up the release
Maybe soft-launch next week
Home location being updated in-flight?
Could happen if they take off without lock
Apparently they took off in a GPS mode
Maybe the GCS is moving the home point?
Perhaps the EKF is resetting home position somehow?
More info to come
Solo-on-master guys have been flying -rc8 raw from the build server, loading parameters and all is wonderful.
The non-green-cube guy is living on the edge but so far so good
Release 3.5 with a note that you should use something like the green cube?
Philip recuses himself on this subject
Randy says it will go out with a note
Controller health messages on each axis would be nice
09:57am - Plane
Report of crash detector falsely triggering
Tom to look at
Crash detector is not compatible with deep stall
09:59am - back to Randy and Copter
Rob says a lot of stuff is broken for heli in Copter 3.4
Servo test function broken
Tridge used this yesterday on 3.5beta OK
Direct drive tails don’t work
Both sorts
Acro controller issues
Tridge points out that 3.5 has flown a lot on various stages of development
Tridge noted the auto-upgrade for params from 3.4 to 3.5 didn’t work correctly especially for servo reversals
Tridge is happy to look at a list
Helping to set up a heli ATM
Soft release for 3.5?
Maybe
10:07am - tridge
Small copters have special requirements
tridge re-added 1-axis accelcal
6-axis Accelcal is too hard
EKF3’s ability to learn offsets allowed this to happen
Not as good as doing a full accel cal
May also be useful on impractical-to-manipulate vehicles
No reboot required
No change to 6-axis
Power lines too close to compass
Per-motor compass compensation added
Works really, really well https://github.com/ArduPilot/ardupilot/pull/6448
UX is too, good
Runs motors one-at-a-time
Running one motor at half throttle won’t lift the vehicle
Put vehicle flat, say “go” and it does the job without further interaction
Huge improvement on what we had
Bonus feature: quadplane-like run each motor in sequence feature added
Tridge is thinking of doing a simple compass cal
Point your vehicle north
To within 5 degrees
Runway angles are often known
180 degrees doesn’t work….
EKF failsafe will kick in
Needs GPS fix
Results from compass cal are often very poor
Made up for by the EKF
Learning on EKF limit is ~50milligauss
There’s a script that does compass offset estimation based on flight data
Different strategies
Sphere fitting
Gyro
GPS
Ability to turn off compasses
compass_type_Mask
Disable classes of compass
Not by bus etc
Don’t load lis3mdl compass driver
Early here GPSs
Would be nice to have checkboxes for each compass on the system
Option to set offsets on other compasses based on a “known good” compass?
Could be done on the GCS
We have large max-offsets for compasses now-adays
For Disco’s hatch magnets
Works just fine
Magnetised USB cables!
Beware!
Sensors can saturate…
There was a bug in BCM driver recently
If you see 20,000+ it’s probably this bug which has now been fixed
PixFalcon has a compass issue
LED creates oscillation in the z-axis
150milligauss!
1Hz
FET too close to compass
Already has the buzzer problem
NTF_LED_BRIGHT
PixRacer also shows this
10:07am - MdB and landing abort fly-away
stuck in abort mode
Needs a specific mission setup
Only makes sense for deep-stall
PR is in master now
10:43am - NuttX and PX4 have moved forward on various things
We should look at updating NuttX https://github.com/ArduPilot/ardupilot/issues/6462 https://github.com/ArduPilot/ardupilot/issues/6463
Ports for new boards
Any actual bugfixes, ‘though or new stuff for existing boards?
Specifics needed
For this to pop to the top of the list
10:47am - mavlink.io
Becoming more tightly couple to px4
Dronecode shortcuts
Issue will be raised to add ArduPilot mavlink links to mavlink.io
10:53am - hardware with Philip
RTK GPS with an IMU in it with controllable LEDs
CAN FLOW
CAN ESCs
Open motor control project
CC-by-NC 3.0
Fee for commercial use to help keep things moving
Apparently ArduPilot and Hex are receiving an award for work in Shenzhen
Jani will be talking there as Philip can’t make it
11:00am - report highest temp MdB
New mavlink message for highest-temperature
Or just a health sensor flag?
Mavlink2 message to convey temperature information
11:04am - takeoff flightmode
Tom is thinking about this
Throw vehicle and let it get up to height and loiter
Needs an offset and a direction to loiter at
Take direction vehicle was pointed in when armed