GRIDS: A Geospatial AI System for Land Boundary Demarcation and Scheduling

Accurate land boundary demarcation is critical for defining legal property rights and ensuring effective land management in Taiwan. Traditional manual workflows face significant challenges due to imprecise forecasting of fieldwork duration and inefficient, non-digitalized task allocation, leading to substantial case backlogs. To address these issues, we present GRIDS (Government Resource Intelligent Digital Services), a fully implemented, end-to-end geospatial AI system that transforms this process. At its core, GRIDS integrates a sophisticated hybrid deep learning model, combining Periodic Linear Encoding and Graph Convolutional Networks with attention to provide highly accurate survey task duration forecasts. This is coupled with an automated scheduling module designed to balance workloads and optimize resource allocation for field personnel. GRIDS is currently deployed within the Taiwan land administration system and has demonstrated substantial real-world impact. Empirical validation in real cadastral cases shows that the system reduces the prediction error of task duration to within 10 %, representing more than 20 % improvement over traditional heuristic methods. Crucially, by intelligently balancing workloads, GRIDS shortens overall case handling time by approximately 55 % to 70 % and promotes more equitable task distribution. This work offers a production-ready digital transformation solution that leverages accurate AI forecasting, practical scheduling, and interactive visualization. The design blueprint of GRIDS is readily extendable to other resource-intensive public-sector field services.

Recommended citation: Min-Huan Tsai, Chia-Chen Tsai, En-Zuo Huang, Pei-Xuan Li, Hsun-Ping Hsieh, Yu-Yin Wang, Tzu-Chang Lee. GRIDS: A Geospatial AI System for Land Boundary Demarcation and Scheduling. In 2025 IEEE International Conference on Data Mining Workshops (ICDMW)
Download Paper | Download Slides