Skip to content


SINERGI is delighted to announce a seminar series, showcasing cutting-edge research in the realm of transport and logistics. Join us on April 11, 2024, at Lecture Hall G, within the esteemed Civil Engineering Department at Delft University of Technology. Discover a diverse array of enlightening topics presented by esteemed domain experts from around the globe.

Research Cases on the Applications of Data-Driven Methods in Smart Cities

Speaker: Dr. Hai Wang
Time: 11:00 – 11:45 am

Abstract: The rapid development and widespread adoption of mobile devices, sensors, IoT, and communication technology have led to the generation of vast volumes of multi-source, high-dimensional data in various systems within the broader framework of smart cities, including transportation, logistics, e-commerce, healthcare, etc. Consequently, numerous data-driven methods have been developed and implemented to address research challenges related to the design and operations of these systems. In this talk, we will briefly discuss several research cases on the applications of data-driven methods in smart cities. These cases include: (1) Descriptive methods for mobile transaction digits distribution and crowd-sourcing food delivery operations; (2) Predictive methods for ICU patient condition evaluation and freelance platform service quality prediction; (3). Prescriptive method for multi-objective matching optimization in ride-sourcing transportation. Through these cases, we aim to showcase the diverse applications of data-driven methods in addressing some key challenges in smart cities.

Joint Assortment-Price-Position Optimization Problem under the Exponomial Choice Model

Speaker: Dr. Hai Jiang
Time: 11:45 – 12:30 pm

Abstract: We study the joint assortment-price-position optimization problem under the Exponomial Choice Model (ECM). The goal is to determine the revenue-maximizing subset of products with their corresponding selling prices and display positions. We formulate this problem as a non-linear mixed integer program. We identify structural properties of the optimal solution and develop an exact decomposition-based algorithm to obtain the optimal solution.

Enablement of Last-Mile Delivery with Autonomous Delivery Vehicles and Electric Two-wheelers

Speaker: Dr. Miaojia Lu
Time: 13:15 – 14:00 pm

Abstract: This study presents a novel approach for enhancing last-mile delivery, enabling the optimization of delivery processes based on system costs, greenhouse gas (GHG) emissions, and autonomous delivery risk. Given the immaturity of technologies and regulations in the logistics industry, completely replacing traditional vehicles with unmanned vehicles to perform last-mile delivery remains a challenge. A potential solution identified in this study is to implement a two-echelon delivery method that combines autonomous delivery vehicles (ADVs) and electric two-wheelers (E2Ws). However, the success of this solution hinges on the synchronization of the two-echelon vehicles. Ensuring such synchronization requires avoiding excessive wait times, which is a key component of maintaining the overall reliability of the solution. To achieve synchronization, we calculate and configure arrival intervals of ADVs and E2Ws at transshipment satellite locations to achieve times under a predefined threshold (10 min) and predefined probability level (80%). Additionally, recognizing the potential accident risk associated with ADV delivery, an enhanced method based on kernel density estimation (KDE) is proposed for risk-aware route planning. To tackle the challenge comprehensively, we present a multi-objective two-echelon vehicle routing problem (2EVRP) with the aims of curtailing system costs, GHG emissions, and delivery risk. A hybrid MOGA-ALNS algorithm, which integrates an adaptive large neighborhood search heuristic with a multi-objective genetic algorithm, is designed to address the problem. Computational results reveal a substantial decrease in GHG emissions and risk mitigation when compared to a model solely focused on cost optimization. The synchronization reliability constraint significantly curtails the amount of vehicle desynchronization in stochastic scenarios. We discuss Pareto-optimal solutions for the three objectives, offering decision-makers insights into the feasibility of mitigating GHG emission and delivery risk without a substantial increase in costs.

Incorporating congestion effects in line planning

Speaker: Dr. Rolf Van Lieshout
Time: 14:15 – 15:00 pm

Abstract: This talk considers the problem of determining lines and frequencies in a public transport system. In contrast to existing approaches, we explicitly consider congestion and assume that passengers may choose different routes to reduce discomfort due to crowding. Our solution approach targets at finding a system-optimal solution by generating passenger routes in a dynamic fashion, whilst also adding cutting planes to deal with the non-linearity introduced by the congestion terms. Since in practice passengers may deviate from system-optimal routes, line plans are evaluated by computing a user equilibrium routing based on Wardrop’s principle. A case study shows that incorporating congestion leads to fundamentally different line plans that achieve a lower perceived travel time, both for the system-optimal routing and for the user equilibrium.

Group Counterfactual Analysis for Supervised Classification and Beyond

Speaker: Dr. Dolores Romero Morales
Time: 15:00 – 15:45 pm

Abstract: Counterfactual Analysis has shown to be a powerful tool in the burgeoning field of Explainable Artificial Intelligence. In Supervised Classification, this means associating with each record a so-called counterfactual explanation: an instance that is close to the record and whose probability of being classified in the opposite class by a given classifier is high. While the literature mostly focuses on the problem of finding one counterfactual for one record, in this presentation we take a stakeholder perspective, and we address the more general setting in which a group of counterfactual explanations is sought for a group of instances. We introduce some mathematical optimization models as illustration of each possible allocation rule between counterfactuals and instances, for tabular as well as functional data, and tasks beyond Supervised Classification.