A Decision-Making System for Slope Risk Deduction

The availability of open-access databases offering free access to multi-source data has greatly accelerated global efforts in landslide risk management. Researchers have been actively working to develop various methods for interpreting landslide-related data. However, the effective application of open data in landslide risk reduction is still hindered by the lack of general-purpose data mining methods. While open data holds immense potential for landslide risk management, its transformation into actionable knowledge for decision-making and slope reinforcement remains a significant challenge. This gap underscores the need for further advancements in knowledge extraction techniques to make open data truly usable in landslide risk reduction. 

To tackle the challenge, an autonomous decision-making system is proposed for slope risk assessment and landslide mitigation based on multimodal large models. By utilizing multimodal large models, the system can retrieve multi-source data, analyze unstructured data, integrate information, and intelligently extract disaster factors. It also supports traceability and human-in-the-loop corrections to minimize model hallucinations. The system autonomously applies specialized algorithms for slope stability analysis, 3D reconstruction, and disaster prediction. The integration of engineering insights, industry reports, research papers, and policies into its knowledge base, using multi-agent and chain-of-thought reasoning, consequently provides reliable engineering suggestions, risk assessments, and impact analyses. 

This system serves as an operational example of transforming open data into open knowledge, specifically for geohazard management, providing a practical reference for other domains. Its capability for autonomous data acquisition and information extraction creates a user-friendly environment, enabling users to effectively manage and mitigate geohazards with essential slope-related data input. Additionally, its intelligent reasoning-based autonomous decision-making technique further enhances users’ capabilities. By streamlining data processing and analysis, the system facilitates faster, more accurate response strategies, improving emergency efficiency and enabling effective resource allocation. As such, the system can play a key role in minimizing potential damage and safeguarding communities. 

The project involves a core working group from Beihang University (China), the University of Urbino (Italy), the United Nations University Institute for Water, Environment and Health (Canada), and Srinakharinwirot University (Thailand). The project is structured into three groups: (1) science and engineering team, (2) secretariat, and (3) supervisory committee. 

The first Project Coordination Committee (PCC) meeting will discuss the project’s goals, methodology, expected outcomes, milestones, and deadlines. The PCC will convene biannually to review progress, with overall project supervision led by the Team Lead.