Neurips Binaries - Windows

Release notes:

Environment Pak Files

• Same as v0.0-windows. No need to re-download.

airsimneurips

• Please upgrade airsimneurips python package version to 0.0.11 to use the below-mentioned APIs.
  • $ pip install airsimneurips --upgrade
• New: The pip package is now documented at https://microsoft.github.io/AirSim-NeurIPS2019-Drone-Racing/

API

• This patch is concerned with the low level control APIs. Specifically, updating and extending old angle set-point APIs in the previous version, and addition of new angle-rate APIs.

• All of the below-mentioned (old / new) APIs work in the body frame of the drone.
  Set points are given in radians or radians / second.
  The body frame follows the Front Left Up (FLU) convention, and right-handedness.

  • X axis is along the "Front" direction of the quadrotor.
    Clockwise rotation about this axis defines a positive roll angle.
    Hence, rolling with a positive angle is equivalent to translating in the "right" direction, w.r.t. our FLU body frame.

  • Y axis is along the "Left" direction of the quadrotor.
    Clockwise rotation about this axis defines a positive pitch angle.
    Hence, pitching with a positive angle is equivalent to translating in the "front" direction, w.r.t. our FLU body frame.

  • Z axis is along the "Up" direction.
    Clockwise rotation about this axis defines a positive yaw angle.
    Hence, yawing with a positive angle is equivalent to rotated towards the "left" direction wrt our FLU body frame. Or in an anticlockwise fashion in the body XY / FL plane.

  • moveByAngleRatesThrottleAsync() (new)

  • moveByAngleRatesZAsync() (new)

  • moveByRollPitchYawThrottleAsync() (new)

  • moveByRollPitchYawZAsync()
    (was moveByAngleZAsync )

  • moveByRollPitchYawrateThrottleAsync()
    (was moveByAngleThrottleAsync )

  • moveByRollPitchYawrateZAsync() (new)

  • moveByYawRateAsync()
    (was rotateByYawRateAsync )

  • moveToYawAsync()
    (was rotateToYawAsync )

• Minor API changes

  • enableApiControl(bool, vehicle_name) is now broken into enableApiControl(vehicle_name), disableApiControl(vehicle_name)

    • enableApiControl
    • disableApiControl

  • armDisarm(bool, vehicle_name) is now broken into arm(vehicle_name), disarm(vehicle_name)

    • arm
    • disarm

Neurips Binaries - Linux

Release Notes:

• Environment Pak files are same as v0.0.0-linux. No need to re-download.

• Please upgrade airsimneurips python package version to 0.0.11 to use the below-mentioned APIs.
  New: The pip package is now documented at https://microsoft.github.io/AirSim-NeurIPS2019-Drone-Racing/

  • $ pip install airsimneurips --upgrade

• This patch is concerned with the low level control APIs. Specifically, updating and extending old angle set-point APIs in the previous version, and addition of new angle-rate APIs.

• All of the below-mentioned (old / new) APIs work in the body frame of the drone.
  Set points are given in radians or radians / second.
  The body frame follows the Front Left Up (FLU) convention, and right-handedness.

  • X axis is along the "Front" direction of the quadrotor.
    Clockwise rotation about this axis defines a positive roll angle.
    Hence, rolling with a positive angle is equivalent to translating in the "right" direction, w.r.t. our FLU body frame.

  • Y axis is along the "Left" direction of the quadrotor.
    Clockwise rotation about this axis defines a positive pitch angle.
    Hence, pitching with a positive angle is equivalent to translating in the "front" direction, w.r.t. our FLU body frame.

  • Z axis is along the "Up" direction.
    Clockwise rotation about this axis defines a positive yaw angle.
    Hence, yawing with a positive angle is equivalent to rotated towards the "left" direction wrt our FLU body frame. Or in an anticlockwise fashion in the body XY / FL plane.

  • moveByAngleRatesThrottleAsync() (new)

  • moveByAngleRatesZAsync() (new)

  • moveByRollPitchYawThrottleAsync() (new)

  • moveByRollPitchYawZAsync()
    (was moveByAngleZAsync )

  • moveByRollPitchYawrateThrottleAsync()
    (was moveByAngleThrottleAsync )

  • moveByRollPitchYawrateZAsync() (new)

  • moveByYawRateAsync()
    (was rotateByYawRateAsync )

  • moveToYawAsync()
    (was rotateToYawAsync )

• Minor API changes

  • enableApiControl(bool, vehicle_name) is now broken into enableApiControl(vehicle_name), disableApiControl(vehicle_name)

    • enableApiControl
    • disableApiControl

  • armDisarm(bool, vehicle_name) is now broken into arm(vehicle_name), disarm(vehicle_name)

    • arm
    • disarm

Neurips Binaries - Windows

Release Notes:

• Environment Pak files are same as v0.0-windows. No need to re-download.

• Please upgrade airsimneurips python package version to 0.0.7 to use the below-mentioned APIs

  • $ pip install airsimneurips --upgrade

• New APIs():

  • setTrajectoryTrackerGains()

    traj_tracker_gains = airsim.TrajectoryTrackerGains(kp_cross_track = 7.5, kd_cross_track = 0.0, 
                                                        kp_vel_cross_track = 5.0, kd_vel_cross_track = 0.0, 
                                                        kp_along_track = 0.4, kd_along_track = 0.0, 
                                                        kp_vel_along_track = 0.04, kd_vel_along_track = 0.0, 
                                                        kp_z_track = 2.0, kd_z_track = 0.0, 
                                                        kp_vel_z = 0.4, kd_vel_z = 0.0, 
                                                        kp_yaw = 3.0, kd_yaw = 0.1)
    traj_tracker_gains_list = traj_tracker_gains.to_list()
    client.setTrajectoryTrackerGains(gains=traj_tracker_gains_list, vehicle_name='')

  • moveOnSplineAsync()

    • Fits a minimum jerk trajectory to the list of given 3D waypoints (specified by the waypoint_vector3r_list parameter).
      Uses ETHZ-ASL's mav_trajectory_generation as the trajectory planning backend.
    • Tracks the references positions and velocities using a pure pursuit tracking controller. The gains of the pure pursuit tracking controller are set by setTrajectoryTrackerGains() as shown above.
    • Note: setTrajectoryTrackerGains() must be called once before calling moveOnSpline()

    moveOnSplineAsync(waypoint_vector3r_list, vel_max=15.0, acc_max=7.5, add_curr_odom_position_constraint=True, add_curr_odom_velocity_constraint=True, viz_traj=True, vehicle_name = '')

  • moveOnSplineVelConstraintsAsync()

    • Fits a minimum jerk trajectory to the list of given 3D waypoints (specified by the waypoint_vector3r_list parameter), with corresponding 3D velocity vector constraints (specified by the velocity_vector3r_list parameter).
      Uses ETHZ-ASL's mav_trajectory_generation as the trajectory planning backend.

    • Tracks the references positions and velocities using a pure pursuit tracking controllers. The gains of the pure pursuit tracking controllers are set by setTrajectoryTrackerGains() as shown above.

    • Note: setTrajectoryTrackerGains() must be called once before calling moveOnSplineVelConstraintsAsync()

    moveOnSplineVelConstraintsAsync(waypoint_vector3r_list, velocity_vector3r_list, vel_max=15.0, acc_max=7.5, add_curr_odom_position_constraint=True, add_curr_odom_velocity_constraint=True, viz_traj=True, vehicle_name = ''):

  • plot_transform()

        plot_transform(pose_list, vehicle_name='')

Neurips Binaries - Linux

Release Notes:

• Environment Pak files are same as v0.0-linux. No need to re-download.

• Please upgrade airsimneurips python package version to 0.0.7 to use the below-mentioned APIs

  • $ pip install airsimneurips --upgrade

• New APIs():

  • setTrajectoryTrackerGains()

    traj_tracker_gains = airsim.TrajectoryTrackerGains(kp_cross_track = 7.5, kd_cross_track = 0.0, 
                                                        kp_vel_cross_track = 5.0, kd_vel_cross_track = 0.0, 
                                                        kp_along_track = 0.4, kd_along_track = 0.0, 
                                                        kp_vel_along_track = 0.04, kd_vel_along_track = 0.0, 
                                                        kp_z_track = 2.0, kd_z_track = 0.0, 
                                                        kp_vel_z = 0.4, kd_vel_z = 0.0, 
                                                        kp_yaw = 3.0, kd_yaw = 0.1)
    traj_tracker_gains_list = traj_tracker_gains.to_list()
    client.setTrajectoryTrackerGains(gains=traj_tracker_gains_list, vehicle_name='')

  • moveOnSpline()

    • Fits a minimum jerk trajectory to the list of given 3D waypoints (specified by the waypoint_vector3r_list parameter).
      Uses ETHZ-ASL's mav_trajectory_generation as the trajectory planning backend.
    • Tracks the references positions and velocities using a pure pursuit tracking controller. The gains of the pure pursuit tracking controller are set by setTrajectoryTrackerGains() as shown above.
    • Note: setTrajectoryTrackerGains() must be called once before calling moveOnSpline()

    moveOnSplineAsync(waypoint_vector3r_list, vel_max=15.0, acc_max=7.5, add_curr_odom_position_constraint=True, add_curr_odom_velocity_constraint=True, viz_traj=True, vehicle_name = '')

  • moveOnSplineVelConstraintsAsync()

    • Fits a minimum jerk trajectory to the list of given 3D waypoints (specified by the waypoint_vector3r_list parameter), with corresponding 3D velocity vector constraints (specified by the velocity_vector3r_list parameter).
      Uses ETHZ-ASL's mav_trajectory_generation as the trajectory planning backend.

    • Tracks the references positions and velocities using a pure pursuit tracking controllers. The gains of the pure pursuit tracking controllers are set by setTrajectoryTrackerGains() as shown above.

    • Note: setTrajectoryTrackerGains() must be called once before calling moveOnSplineVelConstraintsAsync()

    moveOnSplineVelConstraintsAsync(waypoint_vector3r_list, velocity_vector3r_list, vel_max=15.0, acc_max=7.5, add_curr_odom_position_constraint=True, add_curr_odom_velocity_constraint=True, viz_traj=True, vehicle_name = ''):

  • plot_transform()

        plot_transform(pose_list, vehicle_name='')

Neurips Environment Paks - Windows

List of Levels by Difficulty (Least to Most Difficult)

1. Soccer Field - A simple outdoors environment with few obstacles, and an easy to follow course.
2. ZhangJiaJie - A mountainous landscape based on a national park in the Hunan province of China.
3. Building99 - A tight race course designed inside one of Microsoft's very own buildings.

How to Use

1. Download AirSim.zip and unzip it into a place of your choosing. This is your main executable file.
2. Download your desired DLC package(s).
3. Copy your downloaded pak file, and paste it into Airsim/AirSimExe/Content/Paks.

Neurips Environment Paks - Linux

List of Environments

1. SoccerField - A simple soccer field environment with few obstacles, and an easy to follow course.
2. ZhangJiaJie - A mountainous landscape based on a national park in the Hunan province of China.
3. Building99 - A tight race course designed inside one of Microsoft's very own buildings.

How to Use

1. Download AirSim.tar.gz and extract it into a place of your choosing. This is your main executable file.
2. Download your desired DLC package(s).
3. Copy your downloaded pak file, and paste it into Airsim/AirSimExe/Content/Paks.