Normalize action to in [-1, 1]. Optionally normalize observations and scale reward.

Represent a grid and operations on it

CARLA agent, we will wrap this in a proxy env to get a gym env

Context manager to synchronize output from different sensors. Synchronous mode is enabled as long as we are inside this context with

Context manager to synchronize output from different sensors. Synchronous mode is enabled as long as we are inside this context with

Normalize action to in [-1, 1]. Optionally normalize observations and scale reward.

Ant navigating a maze.

Configuration class to store constructor arguments. This is used to store parameters to pass to Gym environments at init time.

Class that encapsulates a graphical window.

Class for rendering DM Control Physics objects.

Colored floor tile the agent can walk over

Class for rendering mujoco_py simulations.

Wrapper that forwards mjlib calls.

Class that encapsulates a MuJoCo simulation. This class exposes methods that are agnostic to the simulation backend. Two backends are support

RoamingAgent implements a basic agent that navigates scenes making random choices when facing an intersection. This agent respects traff

Wrapper which adds an exploration bonus. This is a reward to encourage exploration of less visited (state,action) pairs.

Basic ant locomotion environment.

Base class for all robot classes.

Context manager to synchronize output from different sensors. Synchronous mode is enabled as long as we are inside this context with

A simple agent for the town driving task. If the car is currently facing on a path towards the goal, drive forward. If the car would sta

Empty grid environment, no obstacles, sparse reward

Encode mission strings using a one-hot scheme, and combine these with observed images into one flat array

Classic 4 rooms gridworld environment. Can specify agent and goal position, if not it set at random.

Fully observable gridworld using a compact grid encoding

Ant locomotion rewarded for goal-reaching.

General goal-reaching environment.

Ant locomotion for goal reaching in a disctionary compatible format.

Point locomotion rewarded for goal-reaching.

Swimmer locomotion rewarded for goal-reaching.

Use the image as the only observation output, no language/mission.

Kitchen environment with proper camera and goal setup

Base class for mujoco .xml based agents.

2D grid world game environment

Superclass for all MuJoCo environments.

SuperClass for all MuJoCo goal reaching environments

Base class for offline RL envs. Args: dataset_url: URL pointing to the dataset. ref_max_score: Maximum score (for score norm