level1.yaml

# Level 1

# | Evaluation Scenario | Constraints  | Rand. Inertial Properties | Randomized Obstacles, Gates | Rand. Between Episodes | Notes    |
# | :-----------------: | :----------: | :-----------------------: | :-------------------------: | :--------------------: | :------: |
# | `level1.yaml`       | **Yes**      | **Yes**                   | *No*                        | *No**                  | Adaptive |

# Configuration
num_episodes: 5 # Customizable number of episodes (NOTE: solutions that require fewer episode will receive higher scores)
use_firmware: True # Whether to use pycffirmware or not (NOTE: only set to False for debugging purposes)
verbose: True # Boolean passed to the Controller class, can be used to turn on and off additional printouts

# Note: DSL's flyable/usable space in Toronto extends from -3 to +3 meters in x and y and 0 and 2 meters in z
# Thus, only scenarios that meet these constraints are suitable for sim2real experiments

# Environment
quadrotor_config:
  seed: 1337 # Random seed
  reseed_on_reset: True # Whether to re-seed the random number generator between episodes
  gui: False # Whether to spawn PyBullet's GUI
  camera_view: [5, -40, -40, 0.5, -1, 0.5] # Distance, yaw, pitch, x, y, z target position

  ctrl_freq: 30 # Frequency of the controller's decision making
  pyb_freq: 500 # If `use_firmware` == True, must be multiple of 500, otherwise it must be multiple of ctrl_freq
  episode_len_sec: 33  # Maximum episode duration, in seconds

  done_on_violation: True # Whether `done` becomes True when a constraint is violated
  done_on_completion: True # Whether `done` becomes True when the task is completed (traveled all gates and 2'' hover at stabilization_goal)
  done_on_collision: True # Whether `done` becomes True when a collision happens
  cost: competition  # Sparse reward function, if you desire, re-implement it in Quadrotor._get_reward()
                     # github.com/utiasDSL/safe-control-gym/blob/beta-iros-competition/safe_control_gym/envs/gym_pybullet_drones/quadrotor.py
  
  task: stabilization
  task_info:
    stabilization_goal: [-0.5, 2.9, 0.75]  # Final goal (target position in x-y-z)
    stabilization_goal_tolerance: 0.15

  # Nominal (pre-randomization) initial state
  init_state:
    init_x: -0.9
    init_x_dot: 0
    init_y: -2.9
    init_y_dot: 0
    init_z: 0.03
    init_z_dot: 0
    init_phi: 0
    init_theta: 0
    init_psi: 0
    init_p: 0
    init_q: 0
    init_r: 0

  # Nominal (pre-randomization) positions of gates and obstacles
  gates:
    [  # x, y, z, r, p, y, type (0: `tall` obstacle, 1: `low` obstacle)
      [0.5, -2.5, 0, 0, 0, -1.57, 0],
      [2, -1.5, 0, 0, 0, 0, 1],
      [0, 0.2, 0, 0, 0, 1.57, 1],
      [-0.5, 1.5, 0, 0, 0, 0, 0]
    ]
  obstacles:
    [  # x, y, z, r, p, y
      [1.5, -2.5, 0, 0, 0, 0],
      [0.5, -1, 0, 0, 0, 0],
      [1.5, 0, 0, 0, 0, 0],
      [-1, 0, 0, 0, 0, 0]
    ]

  # Randomization of the initial state
  randomized_init: True
  init_state_randomization_info:
    init_x: 
      distrib: "uniform"
      low: -0.1
      high: 0.1
    init_y: 
      distrib: "uniform"
      low: -0.1
      high: 0.1
    init_z:
      distrib: "uniform"
      low: 0.0
      high: 0.02
    init_phi:
      distrib: "uniform"
      low: -0.1
      high: 0.1
    init_theta:
      distrib: "uniform"
      low: -0.1
      high: 0.1
    init_psi:
      distrib: "uniform"
      low: -0.1
      high: 0.1

  # Randomization of the quadrotor inertial properties
  randomized_inertial_prop: True
  inertial_prop_randomization_info:
    M:
      distrib: "uniform"
      low: -0.01
      high: 0.01
    Ixx:
      distrib: "uniform"
      low: -0.000001
      high: 0.000001
    Iyy:
      distrib: "uniform"
      low: -0.000001
      high: 0.000001
    Izz:
      distrib: "uniform"
      low: -0.000001
      high: 0.000001

  # Randomization of the gates and obstacles positions
  randomized_gates_and_obstacles: False
  # gates_and_obstacles_randomization_info:
  #   gates:
  #     distrib: "uniform"
  #     low: -0.15
  #     high: 0.15
  #   obstacles:
  #     distrib: "uniform"
  #     low: -0.15
  #     high: 0.15

  # Constraints
  constraints:
    # Input constraints
    - constraint_form: default_constraint
      constrained_variable: input
    # State constraints
    - constraint_form: bounded_constraint
      constrained_variable: state
      active_dims: [0, 2, 4] # The uncommented entries in the upper_bounds/lower_bounds vectors below
      upper_bounds:
      - 3     # x
      # - 100 # x_dot
      - 3     # y
      # - 100 # y_dot
      - 2     # z
      # - 100 # z_dot
      # - 3   # phi
      # - 3   # theta
      # - 3   # psi
      # - 10  # p
      # - 10  # q
      # - 10  # r
      lower_bounds:
      - -3     # x
      # - -100 # x_dot
      - -3     # y
      # - -100 # y_dot
      - -0.1   # z
      # - -100 # z_dot
      # - -3   # phi
      # - -3   # theta
      # - -3   # psi
      # - -10  # p
      # - -10  # q
      # - -10  # r

  # Disturbances
  disturbances:
    # Additive noise on the commanded input action
    action:
      - disturbance_func: white_noise
        std: 0.001
    # Random external force applied to the quadrotor (e.g. a push or wind)
    dynamics:
      - disturbance_func: uniform
        low: [-0.1, -0.1, -0.1]
        high: [0.1, 0.1, 0.1]



  # Do NOT edit
  quad_type: 3  # 3D quadrotor
  info_in_reset: True # Info dictionary as second return value (after the initial observation) of env.reset()
  normalized_rl_action_space: False
  done_on_out_of_bound: False