LLM-Based System for Autonomous Intersection Management
Researchers propose LISA (LLM-Based Intersection Signal-free Arbitration), a cognitive arbitration system using large language models for autonomous intersection management. The system addresses challenges in cooperative driving where rule-based approaches fail due to conflicting right-of-way demands, heterogeneous vehicle priorities, and kinematic constraints. Unlike existing methods that rely on signal infrastructure or reservation-based systems, LISA uses LLMs as primary decision-makers without traffic signals. The approach aims to overcome limitations of current LLM-based systems that still depend on signal controllers and face inference latency issues for sub-second control. The paper is published on arXiv with ID 2605.12321.
Key facts
- LISA stands for LLM-Based Intersection Signal-free Arbitration
- System uses LLMs as primary decision-makers for intersection management
- Addresses conflicting right-of-way demands and heterogeneous vehicle priorities
- Existing approaches use LLMs as auxiliary components on top of signal-based systems
- Signal controllers are vehicle-agnostic
- Reservation-based methods lack intent awareness
- Recent LLM-based systems still depend on signal infrastructure
- LLM inference latency limits sub-second control settings
Entities
Institutions
- arXiv