Explore a 51-minute seminar from GERAD Research Center on enhancing simulation-based optimization algorithms through mixed integer linear programming, focusing on autonomous ridesharing. Delve into the research of Claudia Bongiovanni from HEC Montréal as she presents innovative approaches to improve computational efficiency in large-scale discrete optimization problems. Learn about dynamic partitioning of search spaces, problem-specific partitioning rules, and their application to complex stochastic dynamics in urban mobility. Discover how this methodology addresses unpredictable environmental changes affecting service level costs in ridesharing systems. Gain insights into the Dial-a-Ride Problem, event-based simulation, and preliminary results of this novel approach combining simulation-based optimization with mixed integer linear programming techniques.
Enhancing General-Purpose Simulation-Based Optimization Algorithms Via Mixed Integer Linear Programming: A Case Study in Autonomous Ridesharing
GERAD Research Center via YouTube
Overview
Syllabus
Intro
Urban Mobility and Logistics
Handling Uncertainty
Data-driven Decision Making Under Uncertainty
Discrete Simulation-based Optimization (DSO)
DSO Algorithms
A Nested Partitions (NP) Algorithm
Benchmark Partitioning Rules!
The Dial-a-Ride Problem (DARP)12
The Electric Autonomous Dial-a-Ride Problem13
Event-based DARP for Hardly Constrained Problems
DARP DSO
Event-based Simulator
Partitioning Ideas
Implementation & Benchmark Dataset
DSO Settings
Simulation Example
Solutions from the B&B Tree
Generic Partitioning
Compute Time per Node
Preliminary Results
Next Steps
Taught by
GERAD Research Center
