Learning path · embodied-agents
Embodied Agents
A systems path from language-conditioned planning through skill accumulation, grounding, control, and safe deployment.
Working outcome
Design and critique an embodied-agent stack with explicit interfaces for planning, memory, feedback, verification, control, and safety.
Sequence
Build the stack in dependency order.
Planned steps show the shape of the path without pretending the material is already available. Published steps link directly into the full lesson.
Start the available lesson- 01
Language models as planners
PlannedEstablish the planning and tool-use primitives that Voyager builds upon.
- 02
Voyager: compounding capability
AvailableConnect adaptive task selection, executable memory, feedback, and verification in one system.
- 03
Multimodal grounding and world models
PlannedExtend beyond Voyager's text-state abstraction into perception and learned representations.
- 04
Hierarchical robot control
PlannedConnect high-level generated programs to robust low-level policies and physical dynamics.
- 05
Safety for open-ended agents
PlannedGovern authority, verification, memory, and irreversible effects outside simulation.