Humanoid Task Studio

Python framework for defining, running, and evaluating humanoid robotics tasks in simulation.

HTS provides a structured way to describe humanoid tasks, load task packs, run baseline training/evaluation flows, and produce reproducible reports.

What It Includes

• Task specifications with task.yaml
• Task discovery and loading
• Low-level and high-level control interfaces
• PPO and behavior-cloning baseline structure
• Deterministic seeding for reproducible experiments
• Evaluation reports in JSON and Markdown
• Smoke, reporting, loader, wrapper, and seed-determinism tests
• Safety policy for non-violent locomotion, manipulation, and choreography tasks

Tech Stack

Python · PyTorch · Gymnasium · MuJoCo · Hydra · pytest

Quickstart

git clone https://github.com/Build4mBottom/HTS.git
cd HTS
python -m venv .venv
source .venv/bin/activate
pip install -e .[dev]
pre-commit install

Windows PowerShell:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -e .[dev]
pre-commit install

Train a baseline:

hts train task=starter_pack/pick_and_place algo=ppo

Evaluate a checkpoint:

hts eval task=starter_pack/pick_and_place checkpoint=artifacts/ppo_latest.pt

Run a scripted demo:

hts demo task=starter_pack/reach_object

Project Layout

src/
  hts_core/        # env API, simulator abstraction, metrics, wrappers, CLI
  hts_tasks/       # task packs, task loader, task logic
  hts_baselines/   # PPO and behavior cloning
  hts_datasets/    # demonstration format and converters
  hts_eval/        # evaluation reports
  hts_bridge/      # sim-to-real bridge experiments
configs/           # Hydra configs
examples/          # end-to-end scripts
docs/              # architecture, safety, and task-authoring docs
tests/             # smoke and determinism tests

Add a Task

Task packs live under:

src/hts_tasks/packs/<pack_name>/<task_name>/task.yaml

To create a new task:

1. Copy task_template/.
2. Update task.yaml with metadata, observations, actions, and reward details.
3. Add a scripted expert in expert_policy.py when useful.
4. Add matching config under configs/task/.
5. Run the relevant tests.

See docs/add_task_5min.md.

What This Demonstrates

• Python package structure
• Simulation and ML workflow design
• Reproducibility discipline
• Testable robotics/AI infrastructure
• Technical documentation for extensible systems

Status

Early alpha research/prototype project. The repository is useful as proof of Python architecture, simulation workflow design, and testable AI infrastructure.

License

Apache-2.0