Dev Call Sept 3, 2018 2300 UTC

CraigElder · September 3, 2018, 10:59pm

Issues & Pull Requests

github.com/ArduPilot/ardupilot

AP_Declination: Fix the generation script for python 3

ArduPilot:master ← WickedShell:wickedshell/pyigrf-update

opened 11:23PM - 02 Sep 18 UTC

WickedShell

+64 -76

This replaces #8467 fixing up a couple of the kwargs that changed in the python …IGRF package, as well as adapting to the package name change. I squashed the commits from @scivision and dropped the updates to tables.cpp. My last commit is the only new commits that were needed and it can be squashed, I've simply left it alone for ease in reviewing. The summary of why this matters: The tool does not function with python 3 at the moment, and the package is only distributed for python 3. At the moment the only person with a working python2 install appears to be @tridge and this change allows others to actually regenerate the tables correctly.

github.com/ArduPilot/ardupilot

waf: fixes and changes

ArduPilot:master ← OXINARF:pr/waf-changes

opened 06:24PM - 02 Sep 18 UTC

OXINARF

+155 -50

Multiple things here: - fix one method name so it doesn't show up as a waf co…mmand - our waf script had a shebang hardcoded path, changed to use the Python provided by the environment - remove a ChibiOS "hack" that didn't allow selection of ChibiOS base as a board; we already have a method for this, which is to mark a board class as abstract - added an option to disable the new GCC dependencies; it fails too much for me (and others) - long-time coming patch to not fail when pkg-config is not present; it's not a required tool so it will only check for it when needed and failure to find it will only lead to an error if its presence is mandatory - biggest feature: allow to change board to build at build time instead of needing to re-configure the project: - waf, as other tools, is made to run configure once and have everything configured, including all variants (that we use for different boards) and then on build time select what you want to have built. We don't want to force everyone to configure every board, so implementing the choice of board build-time choice was more difficult due to auto-configure support: as configure is going to be run at different times for each board, we need to save the files, hash and options involved in the configure for each of the variants so we check the correct ones when looking for the need to configure again. I've done several tests, but it would be nice if other people did as well.

github.com/ArduPilot/ardupilot

Plane: Cruise target speed handling

ArduPilot:master ← WickedShell:wickedshell/cruise-failsafe

opened 12:08AM - 02 Sep 18 UTC

WickedShell

+40 -13

This presents a couple of small changes to planes handling of target airspeed se…lection in cruise (or fly by wire b). First, it no longer slows down to the minimum airspeed on RC loss when in cruise is a poor pattern, because it keeps the vehicle closest to stalling out, and moves it outside of the envelope it was primarily tuned against. The second option that this adds is it allows a user to specify that centered throttle stick should indicate trim airspeed, rather then being in the middle of the minimum/maximum airspeed. This allows for easier handling of center sprung throttle aircraft that have an asymmetric range above/below trim airspeed. Finally it adds an accessor in RC_Channel that allows you to request what the maximum range on a channel was configured to. This was added as plane uses a range of 100 right now on throttle, and this assumption is hard coded. I'd like to slowly work on removing this assumption, as this introduces problematic quantization of the input signal and is known to cause the target airspeed (and pitch) to experience oscillation. So this simply keeps me from having to revisit the assumption in once spot in the future.

github.com/ArduPilot/ardupilot

AP_HAL: Util: make vsnprintf and snprintf always null-terminate

ArduPilot:master ← peterbarker:snprintf-null-terminate

opened 01:24AM - 27 Aug 18 UTC

peterbarker

+54 -15

The C++ standard indicates these functions always return a null-terminated stri…ng. We should rename these functions if we're not going to conform to the standards. From https://en.cppreference.com/w/cpp/io/c/vfprintf : "Writes the results to a character string buffer. At most buf_size-1 characters are written. The resulting character string will be terminated with a null character" We are still not standards-compliant in the case a length of 0 is passed in, returning 0 where we should return the space that would be required to store the formatted string.

github.com/ArduPilot/ardupilot

Plane: Small cleanup to quadplane const (and clarity)

ArduPilot:master ← WickedShell:wickedshell/qplane-constiness

opened 06:09AM - 24 Aug 18 UTC

WickedShell

+27 -24

This is just a small set of sanity patches to the quadplane code that I came acr…oss while trying to address other things. In no particular order it does the following things: 1. Adds `const` to a number of functions, primarily intended to help track what can side effect. 2. Adds a couple of other objects we require to the free memory check. 3. Extracts a couple of redundant calls to millis() 4. Simplifies the logic, and marginally improves the speed of `QuadPlane::should_relax()`

github.com/ArduPilot/ardupilot

ChibiOS not respecting BRD_SAFETYENABLE intermittently

opened 11:13AM - 23 Aug 18 UTC

closed 05:24AM - 04 Sep 18 UTC

Pedals2Paddles

BUG Safety ChibiOS

### Bug report **Issue details** Pre-arm and arming check of safety switch int…ermittently fails even with parameter `BRD_SAFETYENABLE` disabled on initial power up and never clears until reboot. This never happened in nuttx, and I've experienced this on multiple occasions in ChibiOS (master and 3.6-RCx). Rebooting clears the problem until it happens again, seemingly at random. I can see that the switch handling is a little different in ChibiOS than Nuttx. Something seems to be sneaking by once in a while on initial power up. I don't have disarmed logging enabled, but I don't think there's anything in the DF log that would related to this? @rmackay9 for FYI. I experience this in copter, but I don't think it's a copter issue since the copter code just gets its information from board libraries. **Version** Copter 3.6 and 3.7 master **Platform** [ ] All [ ] AntennaTracker [X] Copter [ ] Plane [ ] Rover [ ] Submarine **Hardware type** Cube

github.com/ArduPilot/ardupilot

Plane: Do_change_speed => cannot return to default

opened 06:15PM - 17 Aug 18 UTC

jace26

When using the Do_Change_Speed command during an auto flight there is no way to …revert control of the speed/throttle to automatic control. For a AS change this isn't to terrible, as you can just create a Do_Change_Speed command to return to the default AS for the aircraft, but with throttle changes this is a huge issue. If I command a throttle change for whatever reason, I am from that point on responsible for throttle output through all phases of flight, until the FC resets. It appears that reloading WPs, changing modes out of auto, and disarming will not reset the AS or Throttle control to default. Another major problem here is that an RTL will not take control over this either so I've simulated driving this into the ground during an RTL with a low throttle commanded by a do_speed_change This has all been done in simulator runs. I propose a change where by setting either of these to 999 with the do_change_speed would return them to auto control. And an RTL should set throttle and AS to default, as should a disarm. **Version** 3.9 3.8.51 **Platform** [ ] All [ ] AntennaTracker [ ] Copter [ x] Plane [ ] Rover [ ] Submarine

github.com/ArduPilot/ardupilot

Method to avoid bad RC trims

opened 08:23AM - 17 Aug 18 UTC

tridge

This is a proposed method to avoid people having bad RC trims. The idea is to …add an arming check that checks if the roll/pitch and yaw sticks have been within the dead-zone for at least 2 seconds continuously since boot. If they haven't been, then give a RC calibration error on arming. This idea was inspired by issues with people using the trim tabs to trim planes, but this also applies to copters. Related to this is to have an option to behave like we do on the skyviper, which is to automatically set the roll/pitch and yaw trim if a radio is connected and the values don't change at all for 2 seconds, and the values are within 25% of center.

github.com/ArduPilot/ardupilot

Plane: Respect frame type on VTOL_TAKEOFF commands

ArduPilot:master ← WickedShell:wickedshell/qplane-takeoff

opened 08:22PM - 13 Aug 18 UTC

WickedShell

+24 -11

This makes the handling of `MAV_CMD_NAV_VTOL_TAKEOFF` correct to the spec, makes… it consistent with the handling of `MAV_CMD_NAV_TAKEOFF` and reduces the possibility of a perceived fly away if an absolute altitude mission is uploaded at any non sea level location. Probably most importantly: it allows for restarting the mission once in flight without the plane deciding to climb the entire takeoff height again if it is above the height. This has been tested in SITL with absolute, and relative altitude takeoffs.

github.com/ArduPilot/ardupilot

Solo Slew Rate software fix for ESC signal issues

opened 02:11AM - 04 Apr 17 UTC

closed 11:16PM - 22 Jan 19 UTC

proficnc

Enhancement Copter Library Solo

#### Issue details _Please describe the problem, or desired feature_ Solo ESC'…s have a serious flaw that can cause them to stop for a second or two if they fail to receive pulses above 2.4v the overall design of the Solo electrical system causes a ground lift of over 1v in some situations. So at 3.3v signal level coming out of the flight control is only seen as 2.3v at the esc in this situation, causing the ESC to register no pulses, and shut down there are some seriously negative side effects of trying to fix this serious hardware problem in software, however, upgrading the autopilot in the solo is a cost, and some people may be happy with the reduced performance of adding the slew rate limiter. It is also worth noting that even with this limit, a heavily loaded vehicle could still have this symptom regardless of the slew rate setting. #### Version _What version was the issue encountered with_ Any version that may be used on a SOLO or any other airframe with this issue. #### Platform [ ] All [ ] AntennaTracker [ X ] Copter [ ] Plane [ ] Rover #### Airframe type _What type of airframe (flying wing, glider, hex, Y6, octa etc)_ SOLO #### Hardware type _What autopilot hardware was used? (pixhawk, pixracer, PX4FMU etc)_ Pixhawk 2.0 Pixhawk 2.1 (no mod) Doesnt affect the Solo Green cube modified version of the cube. This version outputs 5V on the PWM lines which gives a 2.6v head on the ESC required voltage level.

Plane

Copter

RC Update

Rover

Sailboat https://youtu.be/D9I1MK1teL8

CraigElder · September 6, 2018, 6:28pm

Attendee count (max): 16

UTC2300 - python3 for declination script

Package not even available for Python2?
How is this running for tridge?
- Things are different in Australia?
- Possibly an earlier version of the package
https://github.com/ArduPilot/ardupilot/pull/9344
Should we have a test suite?
Merged!

UTC2302 - waf changes and fixes

https://github.com/ArduPilot/ardupilot/pull/9342
Allows multiple boards to be configured
Don’t need to reconfigure for a different board
Use board flag when doing the build
[9:03 AM] (Channel) MdB: python2 waf --board=fmuv3 plane
How does this configure configuration items based on board type?
- Saves configuration files for each board
Configure phase is same as before
Will fail if you try to build for a board you haven’t configured
- But generally remembers multiple board configurations
How does this affect the bootloader building?
- “Group bootloader not found”?
Stuff that goes across from config to build may be a problem
Speed?
- Shouldn’t change anything
(we still need to fix the removed-file-kills-builds problem)
It’s a good idea…
- Still need more people to test this

UTC2306 - https://github.com/ArduPilot/ardupilot/pull/9336

Cruise target speed handling
A couple of small things
- Addresses the case that in fbwb if you lose RC you will go to minimum airspeed
- Add flight option mode to split the throttle percentage to be two ranges - between trim-and-max and trim-and-min rather than just between min and max
We have a variable somewhere which tells us if we have a pending failsafe
Failsafe.throttle_counter
- Don’t read control switch when this is non-zero
- Fbwb would go to minimum for maybe a fifth of a second
Linear interpolate code should be checked
Rc-channel gets get-range callback
- Want a float
- 1% resolution is an issue
- Good start!
Reverse thrust?
- When we treat as an angle rather than a range?
- Could be an issue…
Quick review added
Copter oddity
- Get-control-mid dead-zone
- Half-throttle input
  - You’ll get 48% not 50%
  - Heli pilots might notice this

UTC2320 - https://github.com/ArduPilot/ardupilot/pull/9290

Just fixes a warning in compilation
UNUSED_RESULT
We should probably make these fatal

UTC2323 - https://github.com/ArduPilot/ardupilot/pull/9289

Tridge is OK with changing the API
Need to track down users and make sure they’re not going to complain
Might be a bug in the terrain handling
Someone needs to do a pass through the code to check existing use cases before such a merge is done

UTC2323 - https://github.com/ArduPilot/ardupilot/pull/9281

Patrick’s going back and looking at it
Tridge likes the calling syntax
Two different ways of doing things
Peter will

UTC2337 - https://github.com/ArduPilot/ardupilot/pull/9275

Can go in after Francisco’s reviewed it

UTC2345 - https://github.com/ArduPilot/ardupilot/pull/9270

Things that were irking MdB
Shouldn’t be any functional changes
Missing failure path spotted
- Contiguous memory
  - Most boards have two non-contiguous heaps
Tridge wants to think about the should_relax change to make sure it’s non-functional

UTC2350 - https://github.com/ArduPilot/ardupilot/issues/9266

Board safety enable
Tridge needs a way to reproduce this
- Turning vehicle on and off is only way
Power-cycling required
If you set the board safety mask on a Plane
- You should get outputs in board safe state
- Separate issue
Tridge couldn’t find a code path that would allow it to fail
Also tested it on a Cube Black with 40 or 50 power-cycles
- Using USB switch
- And also software reboot using repeat in mavproxy
Might need to be battery
Also seen on cuav v5 by CO
Tlog would clear this up nicely
Tridge needs to be able to reproduce this
- Or at least get a tlog
No persistent state on the IO
Note put on issue requesting a tlog
Maybe another piece of code is asking for the safety to be enabled again
- Race condition between the two
Offboard controls on Solo?
- Happens on non-Solo devices
FF has seen this
- Sporadic
The in-flight happening is what tridge is worried about
- Needs logs!
- CO has looked at the dataflash
  - Can provide it to tridge
BRD_SAFETY_ENABLE naming
- No persistence on the IO processor means this setting can’t be pushed into the IO processor without recompilation
  - Checksumming means using a flash counter isn’t easy
Any power issues possible?
- Anything sufficient to reset the IO controller?
- Possibly
Possibly something to do with being in an inconsistent state and that continuing even past the safety switch being pressed
Chris will look for logs
Require valid communication with FMU before safety switch becomes operative?
What would the Heli patch do?
- CO hasn’t started work on it
- Tridge is keen to see such a patch
- When safetenable is set to 0 then the circuit is disable globally
  - Hard since there’s a separate microcontroller involved

UTC0021 - https://github.com/ArduPilot/ardupilot/pull/9255

Why not implement it?
ChibiOS already magically does what the call is doing
The nowait thing was to workaround a limitation in the PX4 API
Merged!

UTC0023 - https://github.com/ArduPilot/ardupilot/issues/9218

Plane do-change-speed
Musing on using buttons for speedup and slowdown
Implementation to be done so we have something to discuss

UTC0027 - https://github.com/ArduPilot/ardupilot/pull/9211

Looks good
Randy likes how the new pattern means both vehicles
This pattern means the rc channel stuff becomes a nexus for other bits of the code
Merged!
Having a generic function-failed handler might be nice

UTC0034 - https://github.com/ArduPilot/ardupilot/pull/9186

Adds option to skip mission items if they make no sense
E.g. above takeoff you would still do the transition logic including holding course
Keeps you going through the loop until you load something
There’s a 255 command limit so infinite loop should be avoided
Interface had a boolean which we didn’t honour
Tridge wants to double-check
- Most places are quite straight-forward

UTC0040 - autotest server having issues

UTC0044 - Plane report

Facebook user chased OLED problem with bisection search which was great
Pretty smooth in 3.9.1 release

UTC0044 - Copter rc8 is going OK

Little behind on support
Rc9 out later this week
Solo gimbal is twitching
Every 10 20 30 seconds it yaws randomly
It goes Clang!
https://github.com/ArduPilot/ardupilot/issues/9181
Only on ChibiOS?
Does happen on NuttX but in a much lower form
Could be related to uart size changes
- That code isn’t in the code being run
No flow control on serial4
Maybe packets are being sent too frequently?
Add some sleeps?
https://www.youtube.com/watch?v=hezXabfObvo

UTC0055 - next Rover release is close

A few weeks
Randy’s time is otherwise occupied a little soon
Still working on hector-slam stuff
- Still very bad outdoors on rough surfaces

UTC0056 - Sailing!

Iampete is on call but no micro
https://www.youtube.com/watch?v=2y7xPfMAq4Q
Arm-on-switch vs arming on stick
- We do have parameters to control the ARMING_RUDDER behaviour….
Windvane also working
- Blue thing on the front of the boat
- 360-degree pot
Sailboat Support
Acro mode?
- Heading hold / speed control
- Maintains turn rate based on user input
  - E.g. full right is 90-degrees/second
- In manual throttle contains mainsail sheet
  - Pulling it in or not
- In acro mode no sail control so you go at max speed

UTC0104 - slew rate?

Leonard’s been working on redundancy improvements
Slew rate stuff afterwards, hopefully