Examples

Example code and tutorials for EmbodiK.

Not sure where to start?

New to the API → Quickstart and Basic IK. Teleop stalls or weak tracking → Solver Robustness & Recovery. Collision latency or safety → Collision Constraints & Tuning. Full guide map → Guides overview.

🧭 Core Guides

• Basic IK Example — Minimal solve_position_step() IK bring-up
• Multi-Task Multi-Constraints IK Example — Hierarchical tasks with CoM support-polygon constraint
• Collision-Aware IK Example — Self-collision avoidance with velocity-damper constraints
• CoM Constraint Example Overview — Support-polygon inequality constraints and visualization
• Dual-Arm ECTS Example Overview — Coordinated bimanual control modes and constraints

⭐ Highlighted Scripts

These are the public examples that are kept closest to release quality and are the best starting points for users.

• 01_basic_ik_simple.py — Minimal fixed-base IK loop with interactive target
• 02_collision_aware_IK.py — Collision-aware IK with optional GPU mode
• 03_teleop_ik.py — Minimal teleop input adapter into EmbodiK IK
• 04_com_constraint_example.py — CoM support-polygon constraint demo in Viser
• 05_dual_arm_ects.py — Dual-arm ECTS/Orthogonal coordination with collision handling
• 06_bimanual_whole_body_ik.py — Bimanual whole-body teleop for AI Worker and RB-Y1 with CoM, collision handling, adaptive tuning, and optional Seer input
• 07_unitree_g1_retargeting_ik.py — Unitree G1 retargeting IK with palm/foot/pelvis targets, CoM visualization, and optional self-collision constraints
• 08_spot_full_body_ik_viser.py — Spot full-body IK in regular Viser with arm+torso, torso-only, full-body, and two-stage modes
• 09_spot_locomanip_mjviser.py — Spot locomanipulation ONNX policy rollout in MuJoCo through mjviser

Highlighted interactive examples use a shared solver runtime policy so example code stays focused on tasks, targets, and visualization while robust constraint handling remains in C++. Viser examples share the same default browser endpoint, http://localhost:8080; pass --port when running multiple viewers at once.

🧪 Specialized References

These scripts remain available in a repository checkout or copied example bundle, but they are narrower benchmarks or API walkthroughs rather than the main demo path.

• gpu_batch_ik.py — GPU batched velocity IK benchmark
• gpu_collision_batch.py — GPU batch collision detection benchmark
• gpu_solver_demo.py — CPU vs GPU solver scaling demo
• parallel_trajectory_tracking.py — 100+ robots tracking trajectories in parallel
• robot_model_example.py — RobotModel API walkthrough (FK/Jacobians/CoM)
• visualization_example.py — Visualization and interactive marker usage

🚀 Running Examples

Install and copy the example bundle once using the Installation Guide. Then run scripts from the copied embodik_examples directory:

cd embodik_examples
python3 01_basic_ik_simple.py
python3 03_teleop_ik.py

If pip downloads embodik-*.tar.gz or fails while finding native CMake packages, use the source-build fallback in the Installation Guide.

Examples default to the Panda preset; use --robot <key> to switch models.

For repository development, use Pixi:

pixi run python examples/01_basic_ik_simple.py
pixi run python examples/03_teleop_ik.py

Pass --robot <key> when you want a non-default robot preset.

🧰 Example Helpers

The examples/example_helpers/ directory contains reusable utilities:

• ik_common.py — Shared defaults and small IK/collision helper functions
• teleop_ik_backend.py — Reusable registered-task IK backend for the teleop example
• common_bimanual_model_utils.py / common_bimanual_teleop_app.py — Shared bimanual whole-body IK helpers for public robot examples
• dual_arm_ik_helper.py — Dual-arm IK utilities
• g1_model_utils.py / g1_ik_runtime.py — Unitree G1 model, retargeting, and IK runtime helpers
• limit_profiles/ — Joint limit profile configurations