DEV call 19 Dec 2016 3pm PST = 19 Dec 2016 2300 UTC






  • AP_Sync
  • wiki next steps (move to AWS server, etc)
1 Like

10:02am: Randy - Copter update
3.4.3 is the official version
Heli support issue to do with motor runup; poshold needed a patch to ensure runup complete
Projects area ha fixes in it
Beta release should be outside soon; 3.4.3rc1
Starting to think about 3.5
Beta-testing early-January
M8P RTK support would be nice says PR
Object Avoidance
Working on it for althold mode
No absolute position usually means no knowledge of velocity
Using distance-to-lean0angle conversion
Pilot can override by stick
Vector combination is cool; simply adding the vectors can lead to issues
Instead vectors are broken into vertical and horizontal components
Final vector is formed by adding the maximum vectors for up/down/left/right

10:20am: Randy AP_Sync
TX1, Edison, RPi
Might try to do MQTT telemetry as part of AP_Sync
MQTT dataflash-logging
MQTT telemetry links

10:26am: Grant and Rover
Release is on the way today!
Grant is working on release notes now
Loiter functionality demo may be produced later in the day
Next version will have some avoidance work; in particular, stopping!
Discussion on the sensor-offset code that Paul wrote
Having the sensor offsets relative to the centre of the rear axle might be a good idea
Maybe a little forward will give various advantages (Paul)
Particularly at high speeds
Determining you have a sensor offset issue would be done by looking at things like velocity innovations on your off-centre GPS unit

10:36am: tridge and EKF
EKF3 is in thanks to Paul
No-accel-cal capabilities
Have been flown on Grant’s Firstar
Will allow Copters to take off without accel cal
Tested by setting accel calibrations to zero
Did pick up a few issues
Limits on accel learning that can be removed
HW limits mean sizew of EKF is constrained
Range beacon work will be going into EKF3
No new features into EKF2
Enabling EKF3 is one parameter, zero-calibration is a second
Spinning fast may confuse the learning
May lead to toilet-bowling when spinning stopped
Paul will be looking at corner cases in next month
Replay + ekf3 needs testing
Ride-along logs will be useful, particularly on copter
Maybe saving the calculated offsets based on EKF3’s confidence levels
Some sort of learning supervisor required
Also needed for magnetometer learning
No-mag accel calibrations are next
Might be able to use magnetic field intensity tables to aid calibration
Learned offsets are available in XKF/EKF reports
Latitudes above 60 degrees might be an issue
Might be able to add surveyed-North assumption on startup
Possibly use RTK GPS?
FMUv2 and FMUv3
PX4-v3 is now built as well as px4-v2
Px4-v3 includes CANBUS
Rover is small enough that v3 will be OK everywhere
Copter also just fits
GCS will need updating to know about v3
(Plane) Disco secondary throttle issue is still outstanding
Has been flagged for 3.8

11:20 hardware with Philip
New HW GPS have turned up with HW fixes
PixHawk2.1 next batch in January
Orientation will be zero on future 2.1s
Power module Zener diode are sometimes installed backwards; they will go pop, but they are not required anyway

11:24 web stuff with Randy
Looking at moving Discuss to AWS servers
AWS server for discuss ~US$2,500/year
16G ram, 4 core Xeon, 500G SSD, 1 public IP
Luis suggests Azure might be a cheaper option
30-40% cheaper claims Microsoft
Buzz fixed some of Randy’s problems
Might need to pay on someone’s credit card then getting SPI to reimburse
Question as to what we should do next to improve the website
Randy thinks is a little neglected
Randy thinks that Sphinx might be a better idea than current system
Randy bought up a few oddities about the’s layout
Buzz thinks the tech is fine but we really need more content
Maybe a blog post might get some response?
Previous example of same didn’t
Randy will raise issues against the Wiki issues list
Olivier has stepped up to make some adjustments to the front page
Concrete5 runs front page (edited through browser)
Sphinx runs the “wiki” (edited through git)
Question as to whether should just be ardupilot or whether it should be more general content
“Relevant and of interest” is the rule of thumb
E.g. d-shot, while not currently implemented on ArduPilot is certainly of general interest to ArduPilot users

11:37am Randy: Need to spend more of the money
Any updates from funding committee?
Dev call will get an “updates from funding committee” section
System administration funding proposal?
Content funding proposal?
Someone to look out for random cool and interesting things, blog-post them
E.g. the ornithopter youtube videos
Try to become single-source of bits about ardupilot
Replace facebook, vimeo, random other blog-posting sites, diydrones, rcgroups etc
Paid blog posts?