Configurations

The library comes with a default set of configurations:

Robot Configurations

[dimensions]
width = 1280
height = 720
robot_radius = 20
goal_radius = 10

[timesteps]
delta_t = 1
# 1e3
max_episode_steps = 100
# 1e5
max_session_steps = 1000

[lidar]
n_angles = 1080
lidar_angle_increment = 005817764
lidar_min_angle = 0
lidar_max_angle = 6.283185307

[reward]
collision_score = -25
reached_goal_score = 100
minimum_velocity = 0.1
minimum_distance = 0.1
maximum_distance = 1470
velocity_std = 2.0
alpha = 0.4

[render]
render_each = 1
save_to_file = False

[env]
epsilon = 1

[statistics]
collect_statistics = True
scenario = train

Teacher Configurations

[reward]
alpha = 0.4
terminal_state_reward = 100
max_reward = 3600
base_difficulty = 590
overlap_goal_penality = -100
infinite_difficulty_penality = -1000
too_close_to_goal_penality = -50

[env]
advance_probability = 0.9
max_hard_obstacles_count = 2
max_medium_obstacles_count = 5
max_small_obstacles_count = 7
# {flat: 1D, rings: 2D}
lidar_mode = flat

[statistics]
scenario = train
collect_statistics = True

[timesteps]
max_sessions = 10

Evaluation Configurations

Add Your configs

You can your own custom configurations by creating two files:

robot.config: Robot configurations
teacher.config: Teacher configurations

Include the two files in your current directory, and call the highrl from the terminal with proper args:

(.venv) $ highrl  --robot-config=robot.config --teacher-config=teacher.config