Integrate modern optimization and machine learning methods to improve the efficiency and resilience of supply chains and mobility systems.

Vision

In a world shaped by ever-changing global uncertainties, this Regroupement aims to rethink the management of supply chains and mobility systems by integrating operational research and machine learning. It aims to make these systems more robust, responsive and sustainable, while reducing their environmental footprint and strengthening Quebec’s leadership in the integration of these fields.

Objectives

The objectives of this Regroupement are to integrate and improve modern optimization and machine learning methods to improve the efficiency and resilience of supply chains and mobility systems, while reducing their environmental footprint. The goal is to exploit machine learning:

  • In the design of optimization models that will make these systems more reactive to changing conditions.To accelerate the optimization of very large systems that account for economic, societal, environmental and human contexts.
  • To quantify uncertainty and make supply chains more resistant to disruptions caused by natural disasters, pandemics, wars, and other global events.

Research Axes

Axis 1: Complex uncertainty quantification

Out-of-distribution robustness, count data predictions, complex correlation structure, learning under partial observability.

Axis 2: Integrating prediction and optimization

Simulated environments for contextual optimization, robust supply chain optimization, cost-effective vessel routing and scheduling.

Axis 3: Accelerating the solution of multi-stage stochastic decision problems

Machine learning-assisted Bender decomposition, reinforcement learning for inventory management, fleet management for rail transportation.

Axis 4: Developing decision-aware end-to-end optimization

Fair decision-making, learning of risk-averse policies, integration of human behavior.

Axis 5: Integrating endogeneity in decision models

Integrated learning and optimization for demand forecasting, mitigating epistemic uncertainty in offline learning.

Challenges

Supply chains and mobility systems are facing unprecedented volatility due to global crises, such as pandemics or natural disasters. How can we develop solutions to better anticipate and respond to these disruptions? The challenge also lies in striking a balance between operational efficiency, transparent decision-making, and stakeholder satisfaction, while minimizing environmental impact.

Anticipated Impact

The Regroupement ‘s activities will contribute to three main dimensions of R³AI initiative by developing robust, reasoning, and responsible artificial intelligence (AI) systems. There is a strong need to have robust systems that can react quickly to unforeseen changes and mitigate the effects of these disruptions on downstream operations.

Reasoning systems, creating systems that are more stable and less sensitive to data variations, that are more similar in their behavior to how humans reason and make decisions. The Regroupement will contribute to the responsible adoption of AI by developing models and algorithms that lead to decisions that achieve good trade-offs between the objectives and constraints of multiple stakeholders.

  • Scientific progress: significant scientific contributions to the related disciplines and diffusion with publications in top-tier scientific journals, conferences, etc.
  • Methodological developments: tests on real data in order to draw relevant practical and managerial insights.
  • Industrial collaboration: implementing innovations with strategic players that will ensure methods can be adopted in a real-world setting.

Ongoing Projects

AI-Driven Optimization for Multimodal Transportation Networks: Integrating Theoretical Insights with Data-Driven Decision Modeling
(Axis 4)

Gamifying online stochastic optimization problems
(Axis 2)

Uncertainty Quantification for Spatiotemporal and Multivariate Demand in Supply Chain and Mobility System
(Axis 1)

Dynamic Data-Driven Resource Planning for Wildfire Management under Uncertainty
(Axis 1)

Learning-based sequential decision-making under uncertainty
(Axis 3)

RETRO: Optimization of Orbital Debris Remediation through Machine Learning and Combinatorial Optimization
(Axis 2, Axis 4)

Development of machine learning methods and stochastic short-term production planning to reduce environmental footprint of dynamic fleet management systems in mining complexes
(Axis 2)

Efficient Risk-averse Policy Learning for Supply Chain Management
(Axis 4)

Counterfactual explanations for contextual optimization problems subject to endogenous uncertainty
(Axis 2, Axis 5)

Research Team

Co-leaders

Yossiri Adulyasak
HEC Montréal
Jean-François Cordeau
HEC Montréal
Erick Delage
HEC Montréal
Emma Frejinger
Université de Montréal

Researchers

Research Advisor

Danielle Maia de Souza : danielle.maia.de.souza@ivado.ca