Explaining and interpreting AI
Explainable Artificial Intelligence (XAI) is one of the most effervescent areas of research in Artificial Intelligence (AI) and Machine Learning (ML), which is due to the fact that in general the internal mechanisms of AI and ML systems are difficult to understand and their outputs are difficult to explain. Stakeholders expect some sort of explanation when decisions or results from an AI system affect them. Furthermore, legislation is being introduced, demanding for more Ethical AI systems. A five-year research project, funded in 2023 by NSERC-GD (Canada), proposes different approaches to explaining how AI arrives at a conclusion, namely applications of concepts and techniques related to causality or the introduction and analysis of score-based explanations that value the various attributes in a study to reflect their relevance. Causality is beginning to play an important role in machine learning, as the correlations implicit in the data may not be sufficient to build accurate, generalizable, and robust AI models.
Blockchain in supply chain management
Blockchain technology (BCT) has emerged as an enabling technology that can provide traceability, provenance and transparency in business operations, across complex global supply chain ecosystems, where leanness, agility, and speed are crucial, in addition to achieving social sustainability. It is considered one of the most disruptive technologies representing decentralised environment for transactions, self-executing digital contracts (smart contracts) and intelligent asset management over the Internet, providing a single-view to the entities (users) involved in the transaction. Therefore, the key characteristics of BCT will significantly impact the organisational governance, supply-chain relationships, operations strategy, digital transformation pathway and existing supply-chain business models. BCT when integrated with other technologies such as the Internet of Things, big data analytics, and artificial intelligence, will help to increase the efficiency of supply chain through agile data-driven decision-making based on high quality data (stored in Blockchain) and further facilitating supply chain transparency that will also afford product traceability, authenticity and legitimacy, and enhance sub-supplier transparency that will alleviate social sustainability problem in multi-tier supply networks.
Stochastic bilevel optimization for transportation network design and service pricing
Mobility and logistics systems are vital for societies, and their performance is of primordial importance to the economy. Nevertheless, transportation networks in major cities are frequently congested and disrupted, and the forecasted growth of the worldwide urban population poses a natural threat to the welfare of mobility and logistics systems. With the emergence of on-demand services, user demands for travel and goods are evolving rapidly with substantial economic consequences for society. This is reinforced by environmental challenges: the prospect of climate change poses greater pressure on the need to develop more sustainable, low-emission and resilient mobility and logistics ecosystems . To address these challenges, this project takes a stochastic optimization perspective wherein perturbations and disruptions in mobility and logistics systems are modeled as sources of uncertainties. The goal of this project is to conceive, design and develop novel methodologies for sustainable on-demand service planning and operations.
Resilience and viability in supply chains during epidemics
A pandemic can wreak havoc in supply chains, as witnessed in the COVID-19 context. As workers get infected, production level drops and demand from customers goes unfulfilled. Combining in a novel way an epidemic model with optimal control theory, various models provide a plant manager with the optimal level of prophylactic effort she needs to deploy over a planning horizon to protect the workforce from a pandemic in its early stage and so maintain production levels.
Emerging roles and career paths of project managers
The research aims to delineate changing priorities in the competencies currently recognized by the project management bodies of knowledge and to explore emerging project manager role challenges that require new or adapted competencies fueled by the demands of modern day social, technological and ecological changes in society. We take stock of the roles and competencies required of project professionals in the light of new and recent trends:
Our research questions are framed in the above “post-COVID" contexts and take advantage of existing work done at the authors' institution in applying Theory U concepts to early career project managers in a higher education setting. We adapt this theoretical framework to the context of early career project managers in a higher education setting through 4 principal steps: intention setting, observation phase, presencing phase, crystallizing vision and enacting.
Reducing congestion and carbon footprint in last mile delivery of e-commerce in cities
The members of the Centre will develop new business models for logistic service providers delivering inside city centres. They will produce demand prediction models using both historical and real-time data. The purpose is also to develop an electronic auction system to assign parcel deliveries in real time to available delivery vehicles while minimizing travel and maximizing service. New delivery methods will be tested and deployed in 6 European cities including Hub&Spoke delivery, Hyperlocal on-demand delivery, Collaborative delivery, Digital-as-a-service delivery, and Containerisation delivery. Demonstration activities will generate new products and services that increase the efficiency of the operations of last mile deliveries which on the one hand create new local jobs and on the other will decrease costs and increase profits for the industrial players of the URBANE urban logistics value chain.
Artificial Intelligence in Health: Modeling Advanced Detection of Fear of Recurrence
Today, the main challenge faced by healthcare systems worldwide is the management of chronic conditions. Indeed many diseases could be treated, but they still impact survivors' quality of life and health, requiring continual care even after recovery; moreover, additional chronic health and well-being issues could be generated by treatment. For example, breast cancer survivors have to deal with important challenges daily. Fear of recurrence (FOR), namely “fear, worry, or concern that cancer will come back or progress" is an important feature of cancer survivorship and literature shows that it relates to important factors in post-treatment overall health and quality of life. This project aims at modelling FOR based on fine-grained, multicomponent tracking of the experience along with individual characteristics and well-being/quality of life in breast cancer survivors. We aim at building an integrated system that results in a mobile application used by breast cancer survivors and by professionals to assess patient reported outcome.
Evolution of Sparse and Explainable Neural Topologies
This project aims to develop a multi-criteria framework for the co-evolution of feature subsets with neural topologies to design efficacious yet sparse and interpretable neural networks. To the best of our knowledge, such co-evolution has not been explored so far, especially, from the perspective of interpretability. Further, the part of this research focuses on the development of new search algorithms which are crucial to solving the multi-criteria co-evolution problem, which is known to be an NP-Hard problem. In particular, this research focuses on a day-ahead movement prediction of a stock index as a benchmark problem classification problem for neural architecture design.
Optimal Control Scenarios for Contact-Network Epidemics