Algorithm recipes
Start with one of our four algorithms, then use the planner catalogue to choose comparisons. Each recipe is a configuration you can inspect and adapt.
For a small simulator-free integration, run the direct Python example. The recipes below use the experiment runner to save configurations, metrics and trajectories.
MDOC
Planar constrained planning:
genedynamics-run configs/single_2d/mdoc.yaml \
--device cpu --seed 0 --level 0 \
--development-root results/_development/mdoc
The V1 configuration exposes MDOC's single-robot planning component. The original MDOC project contains the paper's multi-robot coordination application.
MD-COAS
Start with configs/single_2d/mdcoas.yaml for constrained diffusion with
adaptive scheduling. Install the optimization extra for the QP integrations.
genedynamics-run configs/single_2d/mdcoas.yaml \
--device cpu --seed 0 --level 0 \
--development-root results/_development/mdcoas
Then explore configs/d3il_avoiding/mdcoas.yaml for seven-joint arm avoidance;
follow the D3IL setup guide first. The historical
cfsmbd_full key in these files identifies MD-COAS, not a separate planner.
2GO
| Task | Starting configuration | Setup |
|---|---|---|
| Quadruped stepping stones | configs/quadruped/stepping_stones_2d/main/twogo.yaml |
Simulation + optimization extras |
| Humanoid corridor | configs/humanoid/corridor_2d/main/twogo_zone_a.yaml |
Core + optimization for planning; deploy extras for execution |
The humanoid corridor recipe explains a complete runner
command and its output contract. Configuration filenames use twogo; the
registered solver key is 2go.
MGA
| Task | Starting configuration | Setup |
|---|---|---|
| Peg insertion | configs/arm/peg_insert/main/mga.yaml |
Simulation extras; frozen learned assets for formal runs |
| Compliant surface scan | configs/arm/surface_scan/main/mga.yaml |
Simulation extras; prior assets depend on the selected settings |
See installation for simulator and asset setup. V1 qualifies the CPU path; GPU qualification and humanoid walking/push are follow-up priorities.
Compare and scale
The planner catalogue lists the baseline configurations and their dependency requirements. For a runner introduction, the planar CPU smoke pins the CLI output and artifact contract.
Use --dry-run to inspect a configuration, then a development root and one
seed for an initial execution. Inspect status, metrics and trajectory artifacts
before scaling to the formal matrix. The first-run guide
shows how to resume multi-seed runs.