quad_landing.yaml

env:
   num_envs: 1
   num_threads: 1

common:
   verbose: false
   visualization: true

training_env:
   type_input: "ground_truth"  # At the moment, only 'ground_truth' is supported

   sim_dt: 0.25 # if we are not using a sleep in between actions: 0.10 
   max_t: 90.0

   # Action space dimensions
   act_dim: 1
   
   # Observation space dimensions
   img_width: 128
   img_height: 128
   img_channels: 2 # depth + semantics
   
   # Reward parameters
   reward:
      # Configuration
      use_early_stop_semantics: true

      # Final
      landing_good_class: 50.0
      landing_bad_class: -20.0
      max_time: -50.0
      target_height: 5.0

      # Step
      step: -0.2
      sem_good: 1.0
      sem_bad: 0.0

      # Action reward
      lateral_mov: -0.05
      stand_still: -0.5
      down_mov: 1.0

      # Collision
      coll_survival: 0.0

      # Weights step reward
      sem_weight: 1.0
      coll_weight: 0.5
      distance_weight: 0.0
      
      # If robot moves out of environment limits
      out_of_bounds: -5.0
   
   # Collision
   collision:
      min_dist: 3.0 # [m] Min dist from obstacles
      robot_radius: 0.5 # [m]
      
   # Policy learning parameters
   use_dropout: false
   dropout_rate: 0.1
   
   # Environment limits
   env_lim_min_x: -300.0
   env_lim_max_x: 300.0
   
   env_lim_min_y: -300.0
   env_lim_max_y: 300.0
   
   env_lim_min_z: -10.0
   env_lim_max_z: 80.0

quadrotor:
   vel_lateral: 1.0  # [m/s]
   vel_vertical: 2.0 # [m/s]