GeoAI for Driving Risk Assessment via Vision-Language Models: A Knowledge Guided RAG System and Dual Evaluation
Abstract
Accurate and reliable assessment of driving risks enhances road safety and supports efficient urban transportation management. This type of work utilizes diverse data sources and evaluates driving risks from multiple aspects. However, it remains challenging due to the complexity of heterogeneous data and the difficulty in interpreting varied results. This study introduces a novel GeoAI framework that employs Vision-Language Models (VLMs) with a knowledge-guided Retrieval-Augmented Generation (RAG) system to better estimate and evaluate driving risk from the geospatial environment perspective. Specifically, the study proposes a systematic prompt design framework that explicitly instructs VLMs for precise reasoning in driving risk scenarios. The expert knowledge-guided RAG system significantly improves the interpretability and accuracy of the initial VLM-generated results (with about 17% average improvement in few-shot learning). Furthermore, a dual evaluation incorporating both objective performance metrics and subjective human rankings is applied to evaluate the results, revealing the robustness and practicality of our framework. We demonstrate that this framework provides an accurate identification and evaluation of driving risk under the real-world geospatial environment context. This research contributes to the advancements of traffic safety studies with VLMs and RAG and offers insights into geospatial environment reasoning tasks and multimodal GeoAI applications.
Read the paper
Related studies
- Knowledge Management Perspective of Generative Artificial Intelligence
- Navigating Generative Artificial Intelligence Promises and Perils for Knowledge and Creative Work
- Characterizing the evolving research on enterprise content management
- Large language models and multimodal foundation models for precision oncology
- From Generative AI to Extended Reality: Multidisciplinary Perspectives on the Challenges, Opportunities, and Future of Educational Computing
- Generative Artificial Intelligence: A New Engine for Advancing Environmental Science and Engineering
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.