Learning Infrastructure

Embodied learning trains agents to act through a robot’s observations and actions. This section covers policy training and evaluation in EmbodiChain; the built-in training framework currently focuses on reinforcement learning.

• Reinforcement Learning

Related Workflows

Use the Environment Framework to define task observations, actions, and rewards. For demonstration-based learning workflows, see Expert Data Expansion and Online Data Streaming for collecting and supplying training data. The optional Using EmbodiChain with RLinf integration connects EmbodiChain environments to an external training backend.