Integrating Intent Understanding and Optimal Behavior Planning for Behavior Tree Generation from Human instructions (IJCAI 2024)

[Website] [arXiv] [Code]

📰 News

• [2026-06] 🎉 MRBTP-demo is now online! A slim, tutorial-friendly version of MRBTP has been released at https://github.com/Caiyishuai/MRBTP-demo. Great for quickly learning multi-robot behavior-tree planning and porting it to your own tasks.
• [2025-02] 🏆 MRBTP (AAAI 2025 Oral) has been released. MRBTP extends our single-robot behavior-tree planning line of work to multi-robot collaboration with theoretical soundness & completeness guarantees.
  • Project page: https://dids-ei.github.io/Project/MRBTP/
  • Code: https://github.com/DIDS-EI/MRBTP
  • Demo / Tutorial: https://github.com/Caiyishuai/MRBTP-demo
  • Paper: https://arxiv.org/abs/2502.18072
• [2025-01] ✨ The slim demo version of this project is available at https://github.com/DIDS-EI/OBTEA-demo.

This repository contains the code for our project, which is divided into three main sections:

1. LLM Folder: This directory contains the code for the Large Language Models (LLM). It encompasses all the scripts and modules necessary to run and implement the large-scale language model functionalities of our project.

2. OBTEA Folder: Here, you will find the code for the Behavior Tree Backward Expansion Algorithm (OBTEA). This section includes the implementation details and the necessary scripts for executing the behavior tree-based algorithms.

3. RoboWaiter Folder: It contains code for deploying our algorithms in a café's digital twin scenario, showcasing the practical application of our research in simulating and enhancing café service operations.

   👉 Download Simulator

We are committed to continuously updating and improving this repository, so stay tuned for future enhancements and additions.

✨️ We have uploaded the slim version at https://github.com/DIDS-EI/OBTEA-demo. If you want to generate a BT file for a custom task, you can refer to OBTEA-demo/test_demo/run_demo_task.py.

1. First, create your own environment under OBTEA-demo\btpg\envs, such as DemoEasy. The key is to establish the action classes and condition classes in exec_lib. Pay attention to the preconditions (pre), additions (add), deletions (del), and their optional parameters for each action.
2. In the main function run_demo_task.py, provide the path to exec_lib to import behavior_lib.
3. Before running the BT algorithm, specify the goal and the current state cur_cond_set.
4. To draw the BT, you need the .btml file and the imported behavior_lib. We will continue to update and maintain this project, so stay tuned!

Related Projects

• BTExpansion-demo: A concise, well-documented demo that reproduces the Behavior Tree Expansion algorithm (AAAI 2021), with PTML export and Graphviz visualization. Useful as a minimal entry point to understand the backward expansion algorithm.
  • https://github.com/Caiyishuai/BTExpansion-demo

License

This repository is released under the MIT license as found in the LICENSE file.