France, the second-largest wine producer in the world, is adapting to climate change and societal pressures. Winemaking fermentation requires low temperatures, making it energy-intensive. To address this, INRAe's digital twin1 project (DigitWine) uses state-of-the-art sensors to monitor fermentation and aroma compounds. Through multiobjective optimization, it balances energy consumption and aroma targets. High-dimensional visualization and decision-support tools are needed to help operators choose optimal strategies, which are then implemented using dynamic temperature profiles and nitrogen additions.
Objectives of the internship
The goal of this internship is to design and implement an online decision support tool that helps operators in the winemaking industry quickly choose optimal fermentation strategies generated by a multiobjective optimization procedure. The selected student will perform the following tasks:
- Conduct a literature review on visualization methods for digital twins, focusing on visualization techniques for Pareto Front2 (PF) datasets and high-dimensional data visualization3.
- Follow a user-centered design methodology to design and implement interactive visualizations, including: (a) dynamically changing and time-varying visualizations of a PF dataset (≈60 dimensions, thousands of data points); (b) design interaction techniques to effectively explore the PF (e.g., select, filter, aggregate, compare, etc)
- Get feedback on the developed prototype from domain experts and partners in the DigitWine project.
- Write a report documenting the design process, the implemented system and findings from the internship.
The selected student will focus on producing a visualization tool for a standard desktop application. The following resources are currently available for an immediate start of this internship: PF datasets for the wine fermentation process (red and white wine); existing scatterplot-matrix based tools for inspiration and to study their limitations4; video recordings of PF exploration sessions with domain experts; and access to domain experts to gather user requirements and feedback.
Required and desirable skills
Required skills include: web development; programming (JavaScript/D3.js, Python, other). Interest in working with real-world data and domain experts. Knowledge in machine learning or multi-objective optimization is not required but experience is a plus. Experience with user-centered design & information visualization is a plus.
Work environment
- Supervisor: send CV & motivation letter to nadia.boukhelifa@inrae.fr and evelyne.lutton@inrae.fr (INRAE,Univ.Paris-Saclay)
- Internship duration: 6 months, ideally starting in March 2025
- Location: AgroParistech Saclay Campus, 22 place de l'agronomie 91120 Palaiseau
- Allowance: around 650 euros per month
References
1. Fuller, A et al. (2020). Digital twin: enabling technologies, challenges and open research. IEEE Access, 8, 108952–108971.
2. Kung, H-T et al. a. (1975). On finding the maxima of a set of vectors. Journal of the ACM (JACM) 22, 4 (1975), 469–476. 18
3. Liu, S et al. (2016). Visualizing high-dimensional data: Advances in the past decade. IEEE transactions on visualization and computer graphics 23.3 (2016): 1249-1268.
4. Boukhelifa, N, et al. (2019). An exploratory study on visual exploration of model simulations by multiple types of experts." Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. 2019.