Task and Motion Planning Library (TAMPL)

Build instructions for development (temporary)

setup

$ git clone --recursive https://github.com/mlsdpk/tampl.git
$ python -m venv tampl_venv
$ source tampl_venv/bin/activate
$ pip install -r requirements.txt

build and install pytampl

$ pip install . --config-settings=build.tool-args=-j8

build and install pytampl-extensions

$ cd extensions && pip install .

Use Cases / Examples

Task Planning in PDDLGym

You can use TAMPL’s Python bindings to perform task planning directly within PDDLGym environments. The provided example script demonstrates how to solve tasks in a PDDLGym environment and generate a visual solution as a GIF. Simply modify the script to adjust the environment settings as needed.

To run the example:

$ python examples/pddlgym_task_planning.py

Motion Planning in PyBullet

TAMPL’s Python bindings provide an out-of-the-box experience for motion planning in PyBullet-based environments. The underlying motion planners are integrated with OMPL, allowing you to leverage all of OMPL's powerful C++ features while executing high-level behaviors in Python through PyBullet.

Supported Robotic Platforms

Manipulators

Brand	Model
Franka	Franka Research 3

Mobile Platforms

Brand	Model
Willow Garage	PR2		Coming soon...
Clearpath Robotics	Jackal		Coming soon...

Legged Robots

Brand	Model
Boston Dynamics	Spot		Coming soon...

Issues

• Example python script cannot find dylib at runtime. Manually setting is still required at the moment as: export DYLD_LIBRARY_PATH=<path-to-dylib>:$DYLD_LIBRARY_PATH
• Fix FF planner build failure on Unix
• yaml-cpp not found on mac and requires setting this before compiling at the moment as: export LIBRARY_PATH=$(brew --prefix yaml-cpp)/lib:$LIBRARY_PATH
• pybullet rendering not using material properties from .dae files yet (consider using .obj, .mtl files with pybullet.URDF_USE_MATERIAL_COLORS_FROM_MTL flag upon loadURDF call)