IP4: Forecasting Phenotypes based on Management Decisions

关系到种植制度日常管理的战术管理决策，如浇水、施肥或间伐，对园艺作物的植物生长和果实的数量和质量有很大影响。为了支持像农民这样的决策者，该项目旨在开发基于不同战术管理决策的人工智能方法来预测植物的未来生长。通过这种方式，人工智能模型提供了更好的理解，战术管理决策将如何影响园艺作物的产量。这将通过预测与决策者相关的表型性状，如植株高度或水果的数量和大小来实现。对于预测，我们将专注于深度神经网络，它从一致的植物几何-语义模型序列中预测表型，同时考虑过去和计划的战术管理决策。由于每天可以做出的战术管理决策数量非常大，我们的目标也是开发一个能够独立提出未来战术管理决策的网络。虽然该项目的重点是开发新的方法，但开发的技术有可能成为农业决策者的强大工具。

Tactical management decisions that concern the day-to-day management of the cropping system like watering, fertilizing, or thinning, have a large impact on the growth of the plants and the quantity and quality of fruits for horticultural crops. In order to support decision makers like farmers, this project aims to develop AI approaches to forecast the future growth of plants based on different tactical management decisions. In this way, the AI models provide a better understanding how tactical management decisions will affect the yield of a horticultural crop. This will be done by forecasting phenotypic traits like height of plants or the quantity and size of fruits that are relevant for the decision maker. For forecasting, we will focus on deep neural networks that forecast phenotypes from sequences of consistent geometric-semantic models of plants, while taking past and planned tactical management decisions into account. Since the amount of tactical management decisions that can be made every day is very large, we also aim to develop a network that is able to propose future tactical management decisions by itself. While the focus of this project is on developing new approaches, the developed techniques have the potential to become powerful tools for agricultural decision makers.