Towards the development of digital twins for fruit trees

Context and challenges

Model parametrization represents a significant hurdle for FS modelling, currently acting as a brake on its use as a decision-making tool for orchard management. The parametrization process calls for large quantities of data that are both complex and time-consuming to acquire manually, especially for large populations of individuals.

To overcome this hurdle, the DTwin4FruitTrees project will explore the possibility of parametrizing FS models by making use of imaging data from high-throughput phenotyping.

The project will therefore set out to forge closer links between different scientific communities, building connections between groups working on the development of FSPM models for fruit trees and those whose interest lies in the acquisition and analysis of high-throughput phenotyping data, establishing a two-way exchange between the two approaches.

Goals

This project will combine emerging phenotyping methods that use LiDAR and imaging data with methods for the creation and parametrization of FS models. The work will be organised into 4 stages:

1. Literature review of traits that can be accessed through phenotyping to feed the models.
2. Assimilation methods for branching rules: machine learning of meristem development and branching rules using Lidar data.
3. Inference of tree shape and organ geometry: optimisation of FSPM outputs from photogrammetric data, down to the shape and distribution of the organs.
4. Exploration of morphospaces: development of an FSPM prototype for apricot and use of apple and apricot FSPMs to explore the 'morphospaces' created by different genotypes.

This project will open up novel opportunities for the optimisation of orchard management and could enable the development of virtual teaching tools. Its ultimate purpose is to apply the concept of digital twins to the study of fruit trees.

Contact - Coordination :

Project participants

INRAE structures

Non-INRAE partners