LLM-HBT: Dynamic Behavior Tree Construction for Adaptive Coordination in Heterogeneous Robots
wang chaoran, Jingyuan Sun, Yanhui Zhang, mingyu zhang, Wu Changju
AI summary
Problem
Existing multirobot coordination methods struggle with adaptability and robustness in dynamic environments due to rigid decision frameworks or LLM plans that lack online correction. Traditional behavior trees require manual design and cannot scale to unseen tasks or heterogeneous robot capabilities.
Approach
A centralized virtual allocator monitors execution and detects failures, then uses an LLM to dynamically generate and insert behavior tree subtrees for recovery or task reallocation across heterogeneous robots in a closed-loop cycle.
Key results
- Dynamic behavior tree construction via LLM reasoning eliminates manual node design
- Hybrid centralized-distributed mechanism enables real-time failure recovery and task reassignment
- Validated across 60 simulated tasks and a real-world café scenario with diverse robots
- Consistently outperforms baselines in task success rate, robustness, and scalability
Why it matters
Enables reliable, long-term collaboration among diverse robots in complex, changing environments, advancing practical multirobot deployment.
Abstract
We introduce a novel framework for automatic behavior tree (BT) construction in heterogeneous multirobot systems, designed to address the challenges of adaptability and robustness in dynamic environments. Traditional robots are limited by fixed functional attributes and cannot efficiently reconfigure their strategies in response to task failures or environmental changes. To overcome this limitation, we leverage large language models (LLMs) to generate and extend BTs dynamically, combining the reasoning and generalization power of LLMs with the modularity and recovery capability of BTs. The proposed framework consists of four interconnected modules—task initialization, task assignment, BT update, and failure node detection—which operate in a closed loop. Robots tick their BTs during execution, and upon encountering a failure node, they can either extend the tree locally or invoke a centralized virtual coordinator (Alex) to reassign subtasks and synchronize BTs across peers. This design enables long- term cooperative execution in heterogeneous teams. We validate the framework on 60 tasks across three simulated scenarios and in a real-world caf ́e environment with a robotic arm and a wheeled-legged robot. Results show that our method consistently outperforms baseline approaches in task success rate, robustness, and scalability, demonstrating its effectiveness for multirobot collaboration in complex scenarios.