microsoft/AirSim-NeurIPS2019-Drone-Racing v0.3.0-linux

Neurips Training Binaries - Linux

on GitHub

Release notes:

• We are introducing a validation submission pipeline so that participants can see how they are performing w.r.t. others, and for us to get feedback on how easy or hard the race tracks are currently.
• Please note that this leaderboard is not for the qualification rounds. This is effectively a dry run for you to gauge your performance and for us to gather feedback. The qualification binaries with new race tracks will be released next month.
• Website updates:
  • https://microsoft.github.io/AirSim-NeurIPS2019-Drone-Racing/ lists all relevant pages
  • API doc page
  • Upload validation log files
  • Validation Leaderboard

Environment Pak Files

• LoadLevel() is neater in that drones don't fall from the sky.
• Fixed disqualify bug - #43
• Optimized performance of ZhangJiaJie and Building 99 environments

airsimneurips

• v1.1.0
• Documentation

APIs

• New APIs

  • Sim / Race APIs:

    • simDisableRaceLog() - #46
    • simGetNominalGateInnerDimensions() - #28
    • simGetNominalGateOuterDimensions() - #28

  • Set controller gain APIs:

    • setVelocityControllerGains()
    • setAngleLevelControllerGains()
    • setAngleRateControllerGains()
    • setPositionControllerGains()
    • clearTrajectory()

• Changes in existing APIs:

  • setTrajectoryTrackerGains():
    Earlier, the signature was setTrajectoryTrackerGains(gains, vehicle_name=''), where gains was a list of floats, obtained by calling TrajectoryTrackerGains().to_list().
    Now, you just need to pass a TrajectoryTrackerGains object, akin to the above set*ControllerGains() APIs:
    setTrajectoryTrackerGains(gains=<airsimneurips.types.TrajectoryTrackerGains instance>, vehicle_name='')
  • moveOnSpline() and moveOnSplineVelConstraintsAsync() have a bug fix, some new features, and some param name changes
    • Fixed a bug which was causing moveOnSpline*() APIs to fail randomly while replanning with multiple waypoints.
    • Plan from lookahead point - see replan_from_lookahead(bool) and replan_lookahead_sec(float) params.
    • Trajectory visualization
      • works for multiple drones (viz_traj bool param)
      • plots requested waypoints automatically (no need for extraneous plot_transform() calls.
      • supports user-defined color with the new viz_traj_color_rgba param, which takes a list of four floats to specify RGBA.
    • Trajectory stitching constraints:
      • Stitch trajectories with acceleration using new add_acceleration_constraint bool param.
      • add_curr_odom_position_constraint has been renamed to add_position_constraint(bool).
      • add_curr_odom_velocity_constraint has been renamed to add_velocity_constraint(bool)
    • path param has been renamed to waypoints param