Install with python3 -m pip install .

Test using:

python test_env.py

python test_env_batch.py

python test_env_visual.py

name: "SingleAgentGatheringEnv" # Name of the environment class

verbose: 0

base_map: "" # Path to a txt map of NxM elements | ".": empty | "w": wall | "o": R | "x" -R

map_size: [32,32] # Size of previously defined map (from base_map path)

map_view_extend: 0 # Add padding to full map ( if more than 0 pixels)

env_max_steps_no: 1000 # Max number of steps that can be played in the env -> Reset after.

no_envs: 1

no_agents: 1

agents_collide: no

hide_other_agents: no

partial_observable: [1, 4]

                            # [0 (no) / 1 (yes), radius] If to use partial observability and
                            # size of view image considering a radius in pixels around the agent

agents_init_pos: [[-1, 5, 5, 4]]

                             # [ [no_agents, x_coord, y_coord, radius] 
                             # Position no_agents around point (x_coord, y_coord)  within
                             # a radius of "radius"

agents_init_colors: [[-1, 0]] # [ [no_agents, colord_id] ]

use_laser: no

agents_laser_size: 2

reward_distance: no # Reward based on distance to first reward found on map

reward_value: [1., -1.] # [ reward_values, ... ] Reward value for each type of reward defined

reward_respawn_time: [4, 4] # [ reward_respawn_time, ...] No steps to wait for each type of

                         # reward to respawn

reward_init_pos: [[0, 10, 25, 25, 5], [1, 50, 16, 16, 20]]

              # [ [Reward type, no of reward elements, x_coord, y_coord, radius] ...]
              # Define zones to spawn / re-spawn rewards

use_cuda: no

visualize_rgb: yes # Render in RGB mode (obs returned by .render()

store_agents_trace: no

visualize: yes # Should be yes if we want to render

record_data: turned_on: no only_coord: no