Distributed Abductive Reasoning for Self-Explaining Digital Twins

We are seeking a motivated PhD candidate to contribute to the Engineering Digital Twins (EDT) program within Catalyst: the Reliable Hybrid Model Forge . The research focuses on developing explainable and distributed reasoning mechanisms for hybrid digital twins , with the goal of interpreting discrepancies between model predictions/simulation and real-world observations.

Digital Twins are increasingly used to design, operate, and maintain complex Cyber-Physical Systems (CPS) . Modern implementations rely on hybrid models that combine physics-based models with data-driven components. This hybridization enables digital twins to operate in environments where physical models are incomplete, evolving, or partially unknown.

However, as digital twins interact with their physical counterparts, discrepancies inevitably arise between predicted and observed behaviors . This phenomenon, known as the reality gap , may result from model abstraction, sensor drift, environmental variability, or system degradation.

The goal of this PhD is to develop methods enabling digital twins to interpret and explain these discrepancies , transforming them from simple prediction errors into actionable knowledge about the system and its requirements.

Research Focus

Current approaches often treat discrepancies between digital twins and physical systems as numerical errors that must be minimized through reactive model adaptation. While effective in some cases, these methods present several limitations.

First, they rarely provide human-interpretable explanations for the causes of the observed mismatches. Second, reasoning processes are typically implemented in centralized architectures , preventing contextual reasoning near data sources. Third, adaptation mechanisms often ignore the broader system requirements , potentially improving local model accuracy at the expense of global objectives.

This PhD proposes to address these challenges by introducing abductive reasoning as a central mechanism for interpreting the reality gap in hybrid digital twins.

Unlike deduction (deriving predictions from physical models) or induction (learning models from data), abduction focuses on generating plausible explanations for unexpected observations . It is therefore well suited to identifying the causes of discrepancies between model predictions and real-world data.

Explanation Architecture

We will explore a hierarchical and distributed explanation architecture where:

• Local explanations are generated at the edge , close to sensors and data sources
• Intermediate reasoning layers at the fog level refine and combine explanations
• Global reasoning at the cloud or system-level twin validates explanations and evaluates their impact on system requirements

This hierarchical process creates a traceable chain of explanations , linking low-level sensor anomalies to high-level system behaviors.

Research Directions

Abductive reasoning for hybrid digital twins

Designing reasoning mechanisms capable of explaining discrepancies between physics-based and data-driven models.

Distributed explanation architectures

Developing reasoning frameworks operating across edge–fog–cloud infrastructures.

Hierarchical explanation generation

Building multi-level explanation processes connecting local observations to system-level interpretations.

Requirement-aware reasoning

Ensuring that model corrections and explanations preserve global system requirements.

Ultimately, the goal is to transform digital twins from reactive error-correction systems into proactive, self-explaining companions capable of supporting trustworthy decision-making in complex cyber-physical environments.

Key Responsibilities

• Conduct research on abductive reasoning methods for digital twins
• Develop distributed architectures for explanation generation
• Design algorithms linking local anomalies to global system interpretations
• Implement and evaluate prototypes in cyber-physical system scenarios
• Collaborate with interdisciplinary researchers in AI and digital twin engineering
• Publish research results in international conferences and journals
• Participate in EDT program meetings and collaborative activities

Research Environment

The PhD will be conducted at Université Côte d’Azur , within a research environment specializing in artificial intelligence, cyber-physical systems, and digital twin technologies .

The candidate will collaborate with researchers working on:

• hybrid digital twin architectures
• distributed systems and edge computing
• cyber-physical system modeling

Facilities and Resources

• Access to computing infrastructure for distributed experimentation
• Collaboration with researchers in digital twin engineering
• Participation in interdisciplinary EDT research activities
• Opportunities for international collaborations and conference participation

Qualifications

Required

• Master’s degree in Computer Science, Artificial Intelligence, or related field
• Strong background reasoning systems, or distributed computing
• Programming experience
• Interest in cyber-physical systems and digital twin technologies
• Strong analytical and problem-solving skills
• Good written and oral communication skills in English

Preferred

• Knowledge of explainable AI or symbolic reasoning
• Experience with distributed architectures or edge computing
• Familiarity with cyber-physical systems or digital twin concepts
• Background in knowledge representation or reasoning systems

Funding and Benefits

• Duration : 3 years
• Salary : Competitive PhD stipend according to French standards
• Benefits : Social security, health coverage, research travel support
• Travel : Support for conference attendance and research collaborations