SafeNet: A Neural-Symbolic Network for Safe Planning in Robotic Systems Using Formal Method-Guided LLM Fine-Tuning
Zifan Wang, Jialiang Fan, Rui Zuo, Qinru Qiu, Fanxin Kong
AI summary
Problem
Traditional robotic safety planning either struggles with modern system complexity or lacks formal safety guarantees, while pre-trained LLMs lack the specialized domain expertise needed for reliable safety-critical planning.
Approach
SafeNet integrates formal logical constraints and automated reward machines into pre-trained LLMs through a two-phase fine-tuning process, combining supervised learning with preference optimization to generate safe, constraint-compliant trajectories.
Key results
- Block manipulation planning success increased from 1.17% to 91.60%
- Robotic path planning success increased from 7.23% to 90.63%
- Automated reward machine enforces Signal Temporal Logic constraints without manual decomposition
- Comprehensive evaluation of foundation models highlights performance-cost tradeoffs for safe planning
Why it matters
Provides a scalable, safety-guaranteed framework for deploying LLM-driven robots in complex, human-centric environments.
Abstract
Robotic systems present unique safety challenges due to their complex integration of computational and physical processes and direct interaction with humans and environments. Traditional approaches to robot safety planning either rely on conventional methods, which struggle with the complexity of modern robotic systems, or on pure machine learning techniques, which lack formal safety guarantees. While recent advances in Large Language Models (LLMs) offer promising capabilities, pre-trained LLMs alone lack the specific domain expertise required for effective robotic safety planning. This paper introduces SafeNet, a novel neural-symbolic network architecture that enhances LLMs’ safety planning capabilities through formal method-guided fine-tuning for robotic applica- tions. Our approach integrates formal logical knowledge and reward machines into pre-trained LLMs by carefully designed fine-tuning, creating a neural-symbolic approach that combines the flexibility of neural networks with the precision of formal methods for robot trajectory generation and task planning. Experimental results demonstrate significant improvements in safe trajectory generation for robotic systems, with planning success rates increasing from 1.17% to 91.60% for the block manipulation task and from 7.23% to 90.63% for the robotic path planning task.