Humanoid corridor planning

The Zone A recipe exercises the 2GO constrained planner on a multimodal corridor. V1 supports the planning configuration; humanoid walking and push-to-line execution remain deferred.

For G1 model rendering or simulator execution paths, prepare the robot assets with the MuJoCo Menagerie integration. Asset setup is separate from the corridor planning configuration below.

Inspect the resolved robot task

genedynamics-run \
  configs/humanoid/corridor_2d/main/twogo_zone_a.yaml \
  --device cpu --seed 0 --level 0 --dry-run

The command must exit with status 0 and identify:

Configuration is valid (dry run mode)
Environment: humanoid_corridor_2d
Method: 2go
Levels: [0]
Seeds: [0]

Run one development case

Install the optimization extra, then write results outside the formal matrix:

python -m pip install -e ".[optimization]"
genedynamics-run \
  configs/humanoid/corridor_2d/main/twogo_zone_a.yaml \
  --device cpu --seed 0 --level 0 \
  --development-root results/_development/v1-corridor

A complete run records the resolved configuration, the selected trajectory, constraint metrics, and execution status. The planner may return different valid corridor modes across platforms because floating-point reductions can break symmetric ties; the output schema and safety checks are the stable V1 contract.