PhD Position: Human-centric Digital twins for monitoring robotized biostimulants application pr[...]

Organisation/Company Universita degli studi di Milano Research Field Computer science » Informatics Researcher Profile First Stage Researcher (R1) Positions PhD Positions Application Deadline 15 Apr 2026 - 00:00 (Europe/Brussels) Country Italy Type of Contract Temporary Job Status Full-time Offer Starting Date 1 Sep 2026 Is the job funded through the EU Research Framework Programme? Horizon Europe - MSCA Marie Curie Grant Agreement Number Is the Job related to staff position within a Research Infrastructure? No

Offer Description

Exceptional benefits at a glance

• International PhD training excellence (here)
• Interdisciplinary & multi sectoral research
• Competitive MSCA salary & allowances
• Global academic & industrial network
• Non-academic secondments

Salary Gross amount (per month)

Living Allowance EUR 5325

Mobility Allowance EUR 710

Family Allowance EUR 660

GreenFieldData Project at glance : “IoRT Data management and analysis for Sustainable Agriculture ” is a project funded under the action HORIZON Marie Sklodowska-Curie Action (MSCA) Joint Doctoral Network. GreenFieldData will train a new generation of researchers able to tackle digital and green transition challenges using a human-centric approach to ensure the robustness and relevance of the solutions responding to the specific needs of the EU market in a context of climate change and increasing socio-economic constraints.

GreenFieldData will mobilize 14 Doctoral Candidates (DCs) enrolled in Double Degree Doctorate programmes with 12 academic main beneficiary partners, across 7 EU countries. Moreover, 21 non-academic associated partners, and 3 academic associated partners will provide support to the DCs.

PhD Position D – Human-Centric Digital Twins for Monitoring Robotized Biostimulant Application

Context: Modern farms are gradually adopting autonomous robots capable of precision monitoring and management tasks (planting, weeding, spraying etc), supported by IoT-enabled sensor networks that generate continuous streams of environmental and crop-related data. These developments enable the creation of digital representations of crops and their environments, that can simulate growth, predict problems, and guide targeted interventions for improved sustainability and productivity. Among the promising innovations for sustainable crop management are biostimulants, i.e., natural substances or microorganisms that favour nutrient uptake, increase tolerance to stress conditions (nutrient deficiency, draught, soil salinity etc.), improve yield and quality traits of produce. However, their effective use is still under scrutiny depending on a complex interplay of factors such as crop development stage, management practices, plant stress intensity, etc. Integrating multimodal data (pedoclimatic and plant development time series, imagery, and other sensor information) is key to understanding and optimizing these dependencies. Integrating robotic sensing with AI-driven modelling can substantially improve decision-making in biostimulant applications, optimizing treatment timing, dosage, and spatial distribution. This integration relies on high-quality, comprehensive data, the ability to perform near real-time analyses, and preserving human oversight and interpretability throughout the process. Vision-Language-Action (VLA) models which can seamlessly link a robot’s sensory inputs to its actions through a textual interface, offer considerable potential in achieving these goals. often remains opaque to end-users. For AI-based systems to be trusted and effectively adopted by agronomists and farmers, they must be explainable, interactive, and human-centric, with end-users having the possibility to understand the rationale behind recommendations and of exploring “what-if” scenarios to support informed and transparent decision-making

Objectives: This PhD aims to integrate multimodal advanced plant sensing, robotic monitoring and actuation, and augmented crop modelling within an explainable framework for monitoring and optimizing biostimulant application practices in vegetable crops, under real-world stress conditions. Case studies will focus on biostimulant treatments applied to a crop under varying levels of relevant abiotic stress factors, such as drought, salinity, and nutrient deficiencies.

The research will combine greenhouse/field crop sensing, explainable AI, and simulation within a closed-loop system in which sensing, analysis, and action are seamlessly interconnected under human supervision. The specific objectives are as follows:

• Refine and augment plant development models for the selected vegetable crops case study using advanced sensing techniques.
• Develop and deploy a robust phenotyping system integrating 3D and hyperspectral image analysis with air and soil microclimate data for accurate crop monitoring, including automated detection and correction of problematic sensor data (e.g., noise, misalignment) using lightweight AI models suitable for robotic platforms.
• Integrate the multi-sensor phenotyping system on robotic platforms for early detection of stress conditions and for characterization of plant responses to biostimulant treatments.
• Quantify and model the effects of biostimulant applications and integrate these parameters into the crop development model.
• Design explainable decision-support tools for plant treatment recommendations, ensuring human-in-the-loop control, transparency, and trust.

Work plan :

• Review of the state of the art (months 0–6)
• Development and integration of the multi-sensor phenotyping system on the UniMI robotic platform to collect multimodal data for improving crop development models under different abiotic stress levels and biostimulant treatments (months 6–18)
• Construction of a simplified model for the selected vegetable crop to capture plant growth, forecast potential issues, and recommend optimized biostimulant treatment strategies based on the continuously updated robotic data (months 18–30)
• Implementation of explainable AI methods to interpret and validate predictions and recommendations, ensuring transparency and user interaction (months 24–33)

Expected Results

• Crop phenotyping metrics derived from robotic multimodal sensing for detecting abiotic stress conditions (drought, salinity, nutrient deficiencies) and assessing crop recovery levels under biostimulant treatments.
• A reliable prototype AI-based system for the selected vegetable crop, modelling plant development and responses to biostimulant treatments (efficacy threshold, timing etc.).
• Integration of explainability approaches providing interpretable and transparent recommendations for agronomists and farmers, ensuring human-in-the-loop control and trust.

Recruiting and host institutions

• University of Milan, Dept. of Agricultural and Environmental Sciences - Production, Landscape, Agroenergy, Italy (18 Months)(Recruiting institution)
• AAB @ University of Milan, Italy
• Pr. Roberto Oberti (University of Milan, Italy)