WholeBodyVLA
Unified VLA for full humanoid locomotion and manipulation
Most humanoid robot models handle walking and arm movements separately. WholeBodyVLA combines them — so the robot can walk across a room and pick something up in one seamless flow, the way a human does, without switching between two separate systems.
A new release with growing documentation. Requires serious GPU hardware and robotics infrastructure. Follow OpenDriveLab's GitHub closely as tooling matures.
Framework for unified latent VLA controlling both locomotion and manipulation in humanoid robots simultaneously — one model, whole body.
Humanoid locomotion and bimanual manipulation combined dataset
Walk to a table and pick up an object in one smooth motion
Navigate and manipulate in unstructured home environments
Whole-body humanoid control benchmark
Lab-grade infrastructure needed
RELATED DISCUSSIONS
Community →Seeded and planned prompts — not live forum activity yet.
- Discussion room plannedBeginnerBest beginner robotics kit for learning sensors and motors?
Ask the community for honest kit recommendations — simulation-first options welcome. Not live forum data.
- Discussion room plannedAll levelsWhat should go into a first build log?
Build Logs room — photos, BOM, failures, and simulation vs hardware notes. Community thread coming soon.
- Discussion room plannedBeginnerAsk the community: which RoboU track should I start with?
Beginner Robotics + Learn — placement questions without pretending threads are live on Discourse yet.