Hybridize planning and optimization approaches to improve organizational performance

  1. Home
  2. Scientific programs
  3. Hybridize planning and optimization approaches to improve organizational performance

Hybridizing planning and optimization: an innovative approach to industrial engineering

Scientific program Hybridization for Operations & Planning, Organizations & Performance, Optimization & Problem-solving (HOPoPoP)

The Hybridization for Operations & Planning, Organizations & Performance, Optimization & Problem-solving (HOPoPoP) scientific program works to improve organizational performance by proposing hybrid decision-support methods. These methods combine several approaches from different fields such as operations research, simulation and learning methods, hoping to overcome the limitations of individual methods.

 

HOPoPoP program objectives

The aim of this scientific program is to propose new hybrid methods, and to assess the relevance of using them to effectively address today's challenges, characterized by an explosion in system complexity, massive data availability, and growing uncertainties in highly dynamic environments.

 

Methodologies and tools developed

It offers contributions and animations on models, methods and tools capable of dealing with complex problems. It draws on contributions in discrete optimization, scenario evaluation by simulation, multi-criteria decision making and machine learning. Through this program, members of the research team share their contributions with the industrial engineering (IFAC, IEEE, SAGIP) and computer science (IJCAI, CP, ROADEF) communities. They are also active in the GDR MACS and SAGIP, the GDR ROD and RADIA through various working groups. The scientific program also benefits from a high degree of synergy to position its contributions in relation to the laboratory's applied research.

 

Industrial applications and use cases

The HOPOPOP scientific program contributes to :

  • FLOWS axis, to exploring the performance of supply chains subject to multiple uncertainties,
  • DiSCS axis, to the integration of human agents in crisis decision-making processes,
  • TRACE axis, the coordination of players in a multi-scale system located in a heterogeneous territory.
  • WHOPS axis, to health system design and management issues.

 

The program's scientific challenges

The program follows the decision-making process: model, solve and decide. It will address the following challenges:

  • Modeling aspect

Design models capable of faithfully representing real-life situations, while adapting to the need to use hybrid methods.

Expected contributions: creation of models combining several structures adapted to hybrid resolution and exploration of hybrid models combining different methods such as neural networks, multi-agent systems and classical mathematical models.

  • Resolution aspect

Solve large-scale problems in a reasonable time while maintaining good solution quality, exploit combinations of approaches and identify the landscape where hybridization is of real interest.

Expected contributions: development of hybrid algorithms that significantly reduce computation times while maintaining and/or improving solution quality, and comparative analysis of approaches to identify the best trade-offs between quality, time and data consumption.

  • Decision-making aspect

Adapt recommendations to user profiles and the decision-making context, and provide information on the quality of the solutions returned.

Expected contributions: implementation of adaptable and context-sensitive recommendation methodologies and tools, and study of the impact of proposed solutions on the quality of final decisions, and empirical validation of their relevance.