4 Presentations

• 
  03:30 PM - 03:55 PM

  Optimal Menu Pricing Strategies for On-Demand Service Platforms

  • Asadi Melina, presenter, Hec Montreal

  The rise of on-demand service platforms like Uber has led to an increase in premium
  membership plans. These platforms typically charge per-transaction fees while offering
  premium plans with enhanced features and discounted fees for members. This paper presents a
  profit maximization model for such platforms, examining how optimal pricing decisions change
  with menu pricing. It explores the impact of customer behavior and platform improvements,
  such as service quality and discounts, on pricing strategies. Our findings indicate that the
  optimal pricing approach varies between a membership fee, a per-transaction fee, or a
  combination of both. When premium features are limited, platforms rely on higher pertransaction fees with minimal membership fees. However, when discount levels exceed a
  certain threshold, the membership fee surpasses the per-transaction fee. Using analytical
  models and numerical simulations, this study offers insights into customer-platform
  interactions, guiding platforms in designing effective pricing strategies to maximize
  profitability.

• 
  03:55 PM - 04:20 PM

  Decision aid for design for assembly

  • Joseph Thompson, presenter, Polytechnique Montreal

  Assembly lines are a widely used manufacturing system in which a sequence of items passes through a series of stations with a different set of tasks performed at each station. The Type-1 Simple Assembly Line Balancing Problem (SALBP-1) involves assigning tasks to stations in a way that minimizes the total number of stations, given task durations and precedence constraints. While most existing literature assumes that task precedence relations are fully known in advance, this assumption often does not hold in real-world industrial settings. Moreover, precedence constraints can significantly influence the efficiency and feasibility of the assembly line design. To address this challenge, we propose a system to identify and analyze precedence relations that critically impact the optimal solution. Due to the need for real-time decision-making, the system must be computationally efficient and capable of performing rapid sensitivity analysis. To this end, we explore several machine learning techniques to stand in for the use of an exact solver.

• 
  04:20 PM - 04:45 PM

  Understanding Distribution Centre Costs: A Case Study on Automation’s Impact

  • Maude Gagné, presenter, Cirrelt
  • Maryam Darvish, Université Laval
  • Jacques Renaud, Université Laval, CIRRELT

  Effective cost management in distribution centres is essential for profitability and overall supply chain performance. While prior research on warehouse costs predates widespread automation, this study reevaluates cost structures by comparing a traditional distribution centre with a partially automated facility utilizing an A-Frame dispenser. A case study of a Canadian pharmaceutical wholesaler analyzes cost allocation across key activities: receiving, storage, replenishment, order picking, and shipping, with costs categorized into labour, supplies, equipment, infrastructure, and operational support.
  Findings indicate that automation reduces per-unit costs and enhances order-picking productivity but shifts expenses toward replenishment. The automated facility processes three times more orders while achieving a 56.52% reduction in per-unit costs. Despite improving efficiency, automation still requires substantial labour and requires investment in fixed equipment and IT, leading to shifts in cost dynamics. This study’s reproducible methodology provides a practical framework for assessing cost structures, offering valuable insights for decision-makers evaluating automation strategies.

• 
  04:45 PM - 05:10 PM

  Optimization of Human-Robot Task Allocation in Collaborative Assembly/Disassembly Cells using Fuzzy Expert Systems and Reinforcement Learning

  • Elham Ghorbani, Department of Mathematics and Industrial Engineering, Polytechnique Montréal, Canada
  • Ashkan Amirnia, presenter, Department of Mathematics and Industrial Engineering, Polytechnique Montréal, Canada
  • Samira Keivanpour, Department of Mathematics and Industrial Engineering, Polytechnique Montréal, Canada

  The rapid evolution of manufacturing systems towards Industry 5.0 emphasizes the significance of integrating ergonomic and operational efficiency within human-robot collaborative (HRC) environments. Ergonomic risks, predominantly resulting from poor posture and repetitive motions, significantly impact on worker health and organizational productivity. Traditional ergonomic assessments often fail to adequately capture dynamic human movements and the uncertainties inherent in real-time industrial tasks.
  To bridge these gaps, this study proposes an innovative optimization framework for task allocation between humans and collaborative robots (cobots) within assembly and disassembly cells. The approach employs real-time video processing to dynamically capture worker movements, extracting critical ergonomic risk indicators such as joint angles and motion magnitudes. A fuzzy inference system integrates posture and fatigue risks into cumulative ergonomic risk evaluations, accounting for uncertainty and variability in human performance.
  Recognizing the complexity and combinatorial nature of task allocation problems, the proposed framework incorporates a reinforcement learning (RL) strategy, specifically leveraging a Deep Q-Network (DQN) algorithm. The DQN method dynamically optimizes the allocation of tasks to humans or cobots based on minimizing cumulative ergonomic risk alongside operational cost considerations. This approach is tested and validated through an illustrative case study employing real-time video data processed via Python and the Mediapipe library.
  This research contributes to optimization-oriented studies by offering a holistic, real-time method for ergonomically safe and cost-efficient task distribution in collaborative manufacturing environments. The proposed framework stands as a critical advancement toward human-centric design, ensuring sustainable productivity and enhanced worker well-being in modern manufacturing systems.