SCP: Spatial Constraint Planning for Robotic Manipulation in MuJoCo

This project implements Spatial Constraint Planning (SCP) for robotic manipulation tasks within the MuJoCo. It is partially based on the Rekep and supports testing in MimicGen and Robosuite environments.

Installation Guide

1. Create a Conda Environment:

   conda create -n scp python=3.10
   conda activate scp

2. Install SCP Package:

   cd scp/
   pip install -e .

3. Install MimicGen:

   cd ../mimicgen
   pip install -e .

4. Install Robosuite:

   cd ../robosuite
   pip install -e .

Running Demos

cd scp/scp
python main.py --visualize --use_cached_query

• --visualize: Provides visualization of the planned trajectory.
• --use_cached_query: Uses a cached GPT-4o query for faster execution.

Using Your Own API Key

1. Configure API Key:

   • Edit config.yaml in the scp/scp directory.
   • Set your OpenAI API key in the API_KEY field.
   • Optionally, modify API_BASE.

2. Run with Custom Query:

   python main.py --visualize

Running Other Tasks

Use the --task argument to specify tasks.

1. Run a specific task (e.g., Lift):

   python main.py --task Lift --visualize

2. Adding New Task Descriptions: For new tasks, add descriptions to task_descriptions in config.yaml. Example for "Lift":

   task_descriptions:
     Lift:
       description: "lift the red cube and lift it straight up vertically"
       env: Lift

   • description: Task description.
   • env: Environment name.