Env

BaseEnv

class gibson2.envs.base_env.BaseEnv(config_file, model_id=None, mode='headless', action_timestep=0.1, physics_timestep=0.004166666666666667, render_to_tensor=False, device_idx=0)

a basic environment, step, observation and reward not implemented

__init__(config_file, model_id=None, mode='headless', action_timestep=0.1, physics_timestep=0.004166666666666667, render_to_tensor=False, device_idx=0)

Parameters

• config_file – config_file path

• model_id – override model_id in config file

• mode – headless or gui mode

• action_timestep – environment executes action per action_timestep second

• physics_timestep – physics timestep for pybullet

• device_idx – device_idx: which GPU to run the simulation and rendering on

reload(config_file)

Reload another config file, this allows one to change the envrionment on the fly

Parameters

config_file – new config file path

load()

Load the scene and robot

clean()

Clean up

simulator_step()

Step the simulation, this is different from environment step where one can get observation and reward

NavigateEnv

class gibson2.envs.locomotor_env.NavigateEnv(config_file, model_id=None, mode='headless', action_timestep=0.1, physics_timestep=0.004166666666666667, automatic_reset=False, device_idx=0, render_to_tensor=False)

We define navigation environments following Anderson, Peter, et al. ‘On evaluation of embodied navigation agents.’ arXiv preprint arXiv:1807.06757 (2018). (https://arxiv.org/pdf/1807.06757.pdf)

__init__(config_file, model_id=None, mode='headless', action_timestep=0.1, physics_timestep=0.004166666666666667, automatic_reset=False, device_idx=0, render_to_tensor=False)

Parameters

• config_file – config_file path

• model_id – override model_id in config file

• mode – headless or gui mode

• action_timestep – environment executes action per action_timestep second

• physics_timestep – physics timestep for pybullet

• automatic_reset – whether to automatic reset after an episode finishes

• device_idx – device_idx: which GPU to run the simulation and rendering on

step(action)

apply robot’s action and get state, reward, done and info, following OpenAI gym’s convention :param action: a list of control signals :return: state, reward, done, info

reset()

Reset episode