On September 25, 2025, the ACROSS Laboratory hosted a defense committee for five research projects funded by IRD, following the Call for Projects 2024/2025 launched by ACROSS/IRD to support the laboratory’s ongoing research activities.
The Defense Committee welcomed the participation of a Scientific Council composed of representatives from IRD/ACROSS and Thuyloi University (TLU). The selected research projects will also be considered for presentation at TLU’s Annual Scientific Conference.
Those projects are not only a testimony for the strong collaboration at stake between IRD/ACROSS and TLU, but also a joint initiative that generates innovative knowledge for both institutions.
Developing a cloud-based visualization interface for GAMA to simplify the simulation of complex socio-environmental systems
Principal Investigator: Dr. Le Nguyen Tuan Thanh
The project presents a novel cloud-based architecture for multi-agent simulations, built on top of the GAMA platform, to address computational limitations of the traditional desktop-based version. By leveraging the cloud computing model, pub/sub messaging, and containerized deployment, the system enables scalable, parallel execution of complex socio-environmental simulations.
Results demonstrate significant performance improvements, with the system supporting concurrent simulations across multiple worker nodes. The solution reduces infrastructure costs by 40% compared to physical implementation while providing researchers with an accessible web interface for scenario execution. This project establishes a reusable framework for cloud-based agent-based modeling, with particular applicability to smart agriculture and epidemic management.
Deep learning-based forecasting of groundwater levels: A case study of a monitoring station in Hanoi
Principal Investigator: Dr. Ta Quang Chieu
Members: MSc. Nguyen Dac Phuong Thao, MSc. Hoang Van Hung, Ngo Quang Vinh, Nguyen Ha Linh, Nguyen Anh Duc
The project focuses on gathering and analyzing groundwater from a monitoring station in Hanoi. It will develop and validate a deep-learning model capable of producing reliable groundwater level forecasts. Additionally, the project provides a foundation for future research on groundwater forecasting, demonstrating how advanced AI techniques can be leveraged for environmental sustainability. The contribution includes:
- The deep learning models (e.g., RNN, LSTM, Transformer, Autoformer) specifically trained to predict groundwater levels in the station of Hanoi (Q64).
- Data and data processing of several monitoring stations in the Hanoi area (Q64, Q66, Q69).
The research forecast not only short-term (48 hours) but also medium-term (120 hours) and long-term (360 hours) of groundwater levels, helping to overcome the short-term forecasting limitation that earlier research has encountered for groundwater level forecasting. Experimental results reveal that RNN does well in short-term forecasting, while LSTM performs well in the medium term, and Autoformer is clearly superior in long-term forecasting situations. This demonstrates that the model based on an attention-based architecture can capture long-term properties of groundwater level time series. These findings support the use of deep learning in groundwater level forecasting, paving the way for the creation of intelligent forecasting systems, aiding decision-making in water resource management, and developing ways to adapt to climate change in large cities
Designing an IoT system in the platform of UAV and Web visualization for gathering water quality data for a large area of reservoir
Principal Investigator: Assoc. Prof. Dr. Pham Duc Dai
Member: MSc. Nguyen Thi Thuy Hang
This project tackles the challenges of surface water pollution caused by industrialization and human activity. It develops an IoT-integrated UAV system equipped with Total Suspended Solids (TSS) sensors to collect and transmit real-time water quality data to a web-based monitoring platform.
Field tests conducted at Thanh Nhàn Lake revealed high TSS concentrations and elevated pH levels, indicating significant algal presence and alkaline conditions potentially harmful to aquatic ecosystems. This innovative system demonstrates how UAV and IoT technologies can expand monitoring coverage and improve environmental management efficiency.
Evaluating erosion dynamics and determinants in the Da river basin, Vietnam
Principal Investigator: Dr. Le Van Thinh
This study offers a quantitative analysis of soil erosion in the Đà River Basin by integrating multiple environmental and geological factors into the RUSLE model. Using Principal Component Analysis (PCA), the research identified rainfall and soil composition (particularly sand content) as dominant determinants of erosion, with slope, slope length, and silt content also contributing.
The project produced an erosion map that helps managers identify erosion-prone areas and develop strategies to enhance soil conservation, agricultural productivity, and regional planning. The findings provide a scientific basis for future watershed management and sustainable land-use policies.
Building a real time soil NPK monitoring system toward smart agriculture and climate change adaptation
Principal Investigator: MSc. Vu Thanh Vinh
Members: Pham Thanh Binh, Truong Xuan Nam, Do Van Hai
To address the environmental impact of fertilizer overuse, this project designed a portable device capable of measuring soil nutrients (N, P, K) and environmental indicators such as temperature, light, and humidity in real time. The device collects sensor data and transmits it via RF waves to a computer, where it is stored and visualized on a dynamic dashboard.
The team evaluated multiple machine learning models-Random Forest, XGBoost, LightGBM, Neural Network, and Transformer-identifying the Neural Network as the most accurate for predicting crop yield. This model has been integrated into a web-based prediction platform, contributing to the advancement of smart agriculture and climate-resilient farming.
