facebookresearch / OccupancyAnticipation

This repository contains code for our publication "Occupancy Anticipation for Efficient Exploration and Navigation" in ECCV 2020.
MIT License
76 stars 26 forks source link

How is "done" variable being defined in exploration task? #32

Closed AgentEXPL closed 3 years ago

AgentEXPL commented 3 years ago

Hello, I observe the following code in habitat_extensions/exploration_demo.py. And I am confused how is the value of "done" being computed.

while True: obs, reward, done, info = env.step(action) if done: obs = env.reset()

Apart from the maximum number of steps per episode, is there any other condition for an agent to finish one episode? For example, if the agent has fully explored all spaces in the house before the number of experienced steps reaches the allowed maximum number of steps per episode. Will the agent finish this episode and perform env.reset()?

AgentEXPL commented 3 years ago

Sorry for asking. I have found the answer by checking the code in habitat module.