Intelligent Crop Production: Data-integrative, Multi-task Learning Meets Crop Simulator

Many studies have reported that global crop production needs to double by 2050 due to rising population, and boosting crop production is essential. This project focuses on computational approaches for this problem by using data about yields obtained under different conditions along with soil and weather data and drone and satellite images. Computational approaches can be roughly divided into two: 1) machine learning (ML), where a model is learnt from the data, and 2) simulators, where the data is generated from a model. Our contribution is two-fold: 1) we develop a new ML method, and 2) we combine ML with simulation. The first objective is to attain a high predictive performance for crop production. The second objective is to improve simulation quality with ML, which allows using new types of data, say satellite images, which cannot be used by conventional simulators. The project will develop state-of-the-art ML technology that will increase the crop production.