You can deploy this model and start earning money today!

zhijieq/
houseexpo-navigation-env

Name: houseexpo-navigation-env
Availability: InStock

gymnasium

robotics

reinforcement-learning

Views

No views yet

HouseExpo Navigation Env

This is an experimental EnvHub/Gymnasium export of the MuJoCo HouseExpo cardinal-navigation environment. It contains environment/runtime code only, not PPO training code.

Matching policy: zhijieq/houseexpo-navigation-policy.

This repo includes the local HouseExpo JSON subset used by the export.

Load With LeRobot EnvHub

python

from lerobot.envs import make_env

env = make_env("zhijieq/houseexpo-navigation-env", trust_remote_code=True)

Load Directly

python

from env import make_env

env = make_env()
obs, info = env.reset()
obs, reward, terminated, truncated, info = env.step(env.action_space.sample())

Interface

Observation:

observation.lidar: float32[256], 360-degree grid lidar distances in meters.
observation.state: float32[4], [robot_x, robot_y, goal_dx, goal_dy].

Action:

gymnasium.spaces.Discrete(4)
0=move_right, 1=move_up, 2=move_left, 3=move_down.

Rewards match the training rollout convention: navigation reward plus exploration bonus for newly observed free cells.

Status

This is an MVP release for easy pull/eval/finetuning experiments. It is not a grand release or a stable benchmark package.

Citation

HouseExpo: Tingguang Li, Danny Ho, Chenming Li, Delong Zhu, Chaoqun Wang, Max Q.-H. Meng. "HouseExpo: A Large-scale 2D Indoor Layout Dataset for Learning-based Algorithms on Mobile Robots", 2019.