Skip to content

RL Tasks

This section covers reinforcement learning task definitions and training-facing configuration surfaces built on top of the core simulation stack.

Current Coverage

Understand why long-lived robot tasks are usually better served by a manager-based architecture than a Gym-style direct env.

Connect lav2.tasks.isaaclab, configure task imports, and work with the main manager-based RL environment setup.

Run the managed-environment task definitions under lav2.tasks.genesis_forge.

Follow the manager-based RL task wiring under lav2.tasks.mjlab.

Scope

This section is intended for tuning manager-based task configs, training agent setups, or RL-specific runtime parameters. The pages here build on the core simulation stack and focus on backend-specific training workflows.