Servers by jDrones

Dev Call July 29 2019 2300 UTC

(Craig Elder) #1


Scrimmage Simulator

Issues & Pull Requests


  • Update



(Craig Elder) #2

Attendee count (max): 20

UTC2304 - AirSim and MAVProxy GSoC updates

  • Param editor not merged yet
    • Need to meet with Akshay
  • Rajat has been improving AirSim support
    • the LIDAR support
    • Also working on the documentation
    • AirSim takes a bit to get working
      • Rewarding once done
      • Good documentation will reduce frustration
    • Initial environment is just “blocks”
      • Want a cityscape or town environment by default
      • Extra steps are required to get these
      • Need to walk people right through to the point they have one of these environments set up
      • Better environments are pay-for
        • Artists tend to charge for the better environment
        • Could we fund someone to create us an environment?
        • Tools are available
          • Quite a bit of skill involved
        • Pre-built binaries of airsim off server which has embedded copy of licensed environment
          • We need a patched airsim so this won’t work for us
      • [9:10 AM] (Channel) kd: The Carla simulator (Unreal-based) can generate Unreal environments: I haven’t tried it yet though.
    • Car support is moving along
      • Rover!

UTC2310 - Pete Hall and task planning

  • Basic implementation coming in
    • Maybe come back to the circle thing
    • Still needs to consider the turn radius of a vehicle
      • Gimballed camera would make quite a difference here
  • Some work on motor-sailing
    • Working on choosing between the motor and sail

UTC2316 - Matt and autorotation

  • Has gone out and tested entry and guide phases
  • Mixed results

UTC2324 - no video project update

UTC2325 - GSoC end of July

  • All 5 students passed their evaluations
  • 20th August is end of coding time
    • Into the final part of the run
  • Documentation should be a focus later on!
    • Need to bring new features to be available to users
      • Not always appropriate, however
  • Autorotation may not be ready
    • Maybe bring it in SITL-only?

UTC2332 - Scrimmage

  • Created at Georgia tech
  • Needed to be able to simulate 100 vehicles in formation flight
    • Reached ROS/Gazebo limits
  • Made some assumptions about how to do collision detection
    • Abstracted out the detection so you can do high-fidelity detection if required
  • Used a few other software packages
  • Gazebo’s suffered due to too many owners over time
  • Scrimmage plugin communicates with Ardupilot SITL
  • Used for about 6 months or so
  • Each instance of vehicle brings up own instance of scrimmage so no interaction yet
  • Final commit should have sim_vehicle launch physical environment for all vehicles
    • Morse currently does this
    • Doesn’t have to be done the same way necessarily
    • [9:35 AM] (Channel) AT:
    • Morse is written in Python
    • Multi-cast UDP for mesh networking simulation
    • Set MAVProxy to talk on same multi-cast mesh network
    • Menu-item in MAVProxy to choose between
    • db: I run multiple arduplane a lot and typically use one mavproxy per sim, and have all of them using --out to forward all aircraft data to one “real” vehicle… this is different to what tridge just suggested and does not need multicast.
  • Needs to pass CI
  • Starting to use std::string?
  • Has lock-step scheduling

UTC2346 -

  • Originally for matternet
  • Take max of all variances
  • If other lanes are unhappy it is nice to know
  • Primary lane may be fine and vehicle whinging about
  • MdB is unhappy with munging all of them together
    • Makes situation look worse than it is/was
    • Would prefer extensions
    • Instance field in extensions?
    • Telemetry ports are default mavlink1
      • So you would bounce between the two instances
    • Maybe in packing we switch between the max code based on mavlink1 vs mavlink2
  • Use a compatibility bit to indicate primary or not sensor
  • If we did the max() thing then someone running an EKF on a third hard-mounted IMU might swamp the other IMUs
  • New mavlink2 message?
  • Randy likes the idea of mavlink2 extension
  • Peter likes the idea of a new mavlink2 message
  • Tridge likes using a status bit to indicate primary/secondary in status flags

UTC0000 -

  • Fix compilation warnings
  • ADSB library may be using wrong type

UTC0002 -

  • Fixes for alpine linux
  • Our Code assumes glibc
  • May break on Apple machines
  • Should we run CI on Apple?
    • Current CI systems are capable

UTC0006 -

  • Merged
  • Remove wrapper around send_text

UTC0009 -

  • Frame not checked in MissionPlanner
  • Peter to test in MAVProxy then merge

UTC0011 -

  • Fix weird API
    • And some bugs
  • Merged!

UTC0012 -

  • Not quite ready
  • ADSB message streamrate
  • Fills the buffer with as many messages as it can possibly fit
    • Like parameters
  • [10:19 AM] (Channel) MdB: If you are setting the individual message rates I’m guessing you are doing it very programattacly and the calculation is valubale
  • The 50Hz loop rate in Rover can be a problem
  • This PR has been tested on Rover and Copter
  • Some sort of memory issue
  • Is this feature important enough to go into fmuv2?
    • Tested on Copter and Rover to fit
    • Currently under minimise features

UTC0030 -

  • Need to get FF to review this again
    • And preferably someone else
  • Adding a parameter later might be an option
  • Having a per-channel debounce time seems unnecessary at this point
  • 6-position mode switches might be nice to have
  • Do not want complications on this PR e.g. extra bits for halving

UTC0038 - rangefinder arming checks

  • Really old
  • Ready to go again
  • All vehicles
  • If you specify a rangefinder type then it must exist
  • Arming check becomes 32-bit!
    • Only 24 bits of checks, ‘though
  • Shouldn’t be any change in behaviour
  • Need to fix the parameter documentation problem

UTC0040 - Copter mode auto retract landing gear automatically

  • Maybe use a parameter to specify behaviour
    • We have a series of landing gear parameters already
    • LGR_OPTIONS bitmask

UTC0046 - Plane update

  • Origin changes in
    • Prevents a fly-away
    • No tlog
    • Onboard log, ‘though
    • Accidental setting of EKF origin way too low
      • User error in GCS
      • Combined with no SRTM data
        • And a lack of checking for no SRTM data
    • Beacons aligned with world coordinate system and then going outside will no longer work
      • Fixes coming for this
    • Beta3 is out with the change in it
    • Several EKF-origin fixes have also gone in
      • Not in Copter yet
    • 3.10 coming soon

UTC0050 - Copter updates

UTC0053 - end