protomotions.envs package

Environments for reinforcement learning tasks. All environments extend BaseEnv and implement task-specific reward functions and observations.

BaseEnv

class BaseEnv(config, robot_config, simulator_config, device, terrain=None, scene_lib=None, motion_lib=None)

Base class for all reinforcement learning environments.

Provides core functionality including simulator integration, terrain management, scene and object handling, motion library integration, observation computation, and episode management.

Parameters:

  • config – Environment configuration

  • robot_config – Robot morphology configuration

  • simulator_config – Simulator backend configuration

  • device – PyTorch device for computations

  • terrain – Optional terrain instance

  • scene_lib – Optional scene library

  • motion_lib – Optional motion library

Key Methods:

  • reset() - Reset environments

  • step(actions) - Step simulation with actions

  • compute_observations() - Compute observations

  • compute_reward() - Compute rewards

Example:

config = SteeringEnvConfig()
env = Steering(config, robot_config, simulator_config, device)
obs, _ = env.reset()
next_obs, rewards, dones, info = env.step(actions)

Mimic

class Mimic(config, robot_config, simulator_config, device)

Motion imitation environment for full-body tracking.

Trains agents to imitate reference motions from a motion library with detailed tracking rewards for joint positions, velocities, and end-effector locations.

Key Features:

  • Full-body pose tracking with per-body-part weights

  • Contact-aware foot placement rewards

  • Sync/async motion playback modes

  • Early termination on tracking errors

  • Terrain-aware tracking

Parameters:

  • config – MimicEnvConfig with reward weights

  • robot_config – Robot configuration

  • simulator_config – Simulator configuration

  • device – PyTorch device

Example:

config = MimicEnvConfig()
env = Mimic(config, robot_config, simulator_config, device)
obs, _ = env.reset()

Steering

class Steering(config, robot_config, simulator_config, device)

Steering task environment for humanoid locomotion.

Trains agents to walk in a target direction at a target speed. Target direction and speed change periodically to encourage versatile locomotion.

Key Features:

  • Variable target speeds (including stopping)

  • Periodic heading changes with random variations

  • Visual markers for target direction

  • Rewards for velocity and heading matching

Parameters:

  • config – SteeringEnvConfig

  • robot_config – Robot configuration

  • simulator_config – Simulator configuration

  • device – PyTorch device

PathFollowing

class PathFollowing(config, robot_config, simulator_config, device)

Path following task environment for navigation.

Trains agents to follow predefined paths with observations of future waypoints. Rewards staying close to the path and making forward progress.

Key Features:

  • Multiple future waypoint observations

  • Path generation (lines, curves, complex trajectories)

  • Visual path markers

  • Distance and progress rewards

Parameters:

  • config – PathFollowerEnvConfig

  • robot_config – Robot configuration

  • simulator_config – Simulator configuration

  • device – PyTorch device