Abstracts Track 2023

Area 1 - Methodologies and Technologies

Nr:	7
Title:	Clustering Transportation Modes with the Use of Combined Methods of Multi-Criteria Decision-Making, Network Analysis, and Meta-Heuristic Algorithm of Particle Swarm Optimization
Authors:	Paria Sadeghian
Abstract:	Global positioning systems (GPS) have been widely used in transportation, especially in order to identify human behaviours and transportation modes. Some important travel information can easily be extracted from GPS devices. However, finding transportation modes is more difficult to acquire and needs a more complex analytical process. Several methods, which range from rule-based methods to advanced machine learning algorithms, have been applied to detect the correct transportation modes from GPS data. Most of the previous studies rely on large amounts of labelled datasets. This study proposes a general clustering method with a combination of multi-criteria decision-making, network analysis, and a meta-heuristic algorithm of particle swarm optimization to cluster the transportation modes. The model uses the raw GPS data as input and, as a result, the transportation mode as predictive analysis. In this model, network analysis is used to determine the weight of each variable. Moreover, considering the elements of the analytic network process’s (ANP) super matrix and the transportation modes as variables, the particle swarm optimization algorithm is used to identify the weight of the characteristics and the classification thresholds. The results indicate that the proposed model can achieve more than 88% accuracy in clustering the transportation modes.

Nr:	15
Title:	Age of Information for Discrete-Time Queue
Authors:	Yutae Lee
Abstract:	Age of information (AoI) measures how fresh information is. The AoI refers to the amount of time that has elapsed since the generation of the most recently successfully received message. The metrics related to AoI include average AoI, peak AoI. Average AoI means the time average of AoI and peak AoI means the peak value of AoI. The peak AoI corresponds to the AoI immediately before the information is updated. In this paper, we consider the information freshness of a discrete status updating GI/G/1/1 system and derive the probability distribution for the peak AoI and AoI in steady state. Consider a case where the freshness of information needs to be kept below a threshold. Average AoI alone cannot determine how much AoI is below the threshold. Peak AoI is sometimes inefficient because it only considers the peak values and not the time interval between peaks. The freshness ratio of information (FRoI) is defined to be the fraction of time the age does not exceed a predefined freshness threshold. We are interested in high FRoI in order to maintain fresh information. Thus, in this abstract, we also derive the FRoI in steady state. The FRoI can be utilized in time-critical applications where we need to apply a threshold restriction on AoI. This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT). (No. NRF-2021R1A2C1011756)

Area 2 - Applications

Nr:	10
Title:	Sustainable Supplier Selection and Order Allocation Under Subjective and Objective Criteria: The Case of Uncertain Supplier Availability
Authors:	Sadeque Hamdan, Hani Arezki Aitizem and Oualid Jouini
Abstract:	The challenge of deploying a sustainable supply chain strategy has become paramount for companies, and the choice of suppliers who respect these commitments aligns with these companies' strategies and helps achieve a competitive advantage. Although it benefits companies in various aspects such as reducing cost and enhancing quality, sourcing is risky and highly dependent on supplier reliability. Contracting with reliable and sustainable suppliers is crucial, and accounting for disruptive events makes this task challenging. Earthquakes, tsunamis, wars and epidemics are examples of disruptive events that significantly affect the sourcing function among various supply chain functions. Disruptive events add another layer of risk to sourcing, where some suppliers may become unavailability completely. In this work, we propose a two-stage sustainable supplier selection model with uncertain supplier availability. In the first stage, we evaluate suppliers based on subjective and objective criteria. The subjective evaluation is done using a fuzzy analytic hierarchy process and fuzzy technique for order preference by similarity to the ideal solution to obtain the subjective factor measure while accounting for uncertainties in the evaluation. The objective evaluation is done by considering objective criteria, such as cost, lead time and distance. Objective criteria are normalized, and the objective factor measure is obtained. Subjective and objective criteria evaluation are combined in one factor. The output of the first stage is a weight value, combining subjective and objective factors, for each supplier and is used in the second stage. In the second stage, a stochastic mixed integer linear mathematical model is formulated. The model considers the combined evaluation of each supplier, product's demand, suppliers' capacities, product lifecycle emissions and costs parameters. It considers the possibility of supplier unavailability using the decision tree approach. The model consists of three objective functions: the expected total cost, the expected total emissions and the expected total purchasing value. A scalarization approach is used for the multi-objective problem. A modified fix-and-relax heuristic is used to solve the formulated problem. The model output includes the selected suppliers, quantities to be ordered, inventory and shortage levels in each period. The model can be used to illustrate the purchasing strategy in the possibility of disruptive events that restrict the availability of some suppliers and under different inventory conditions. The results show that the difference between the exact and modified heuristic solutions is minimal (less than 5%) and that the heuristic significantly reduces the computational time. The work contributions are as follows. First, the work constructs a novel multi-criteria decision-making model based on objective and subjective criteria for selecting green suppliers. Second, we consider supplier availability uncertainty in a stochastic mathematical model, which is a feature that was not considered in previous works to the best of our knowledge. The model also considers the possibility of advancing the order or delaying it by using the company's storage.

Nr:	11
Title:	Operations-Time-Space Network for Solving Freight Train Scheduling Problems in Port Systems
Authors:	Veronica Asta, Daniela Ambrosino and Luca Abatello
Abstract:	The Port Rail Shunting Scheduling Problem (PRSSP) arises in the so-called rail-sea yard where the modal switch between maritime and rail transportation is performed. A rail-sea yard system, also usually called port area, generally includes one railway station, a shunting zone, and several maritime terminals. The focus is on the trains' transfer operations within the port area, between maritime terminals, on one side, and the national railway network, on the other [1]. Supposing to have only one shunting park inside the system, the physical path that an export train may follow in the port area starts when it arrives and waits (if necessary) in the rail station before going to its maritime terminal through either a unique shunting operation (if allowed from the infrastructure), or by performing two operations passing through the shunting park. An import train performs the same opposite path. The PRSSP consists in defining a schedule of those activities for the trains’ transfer within the system, respecting the time limits imposed by the railway network schedule and by the ships one, and the limits due to the finite resources available in the port area. Such resources include both the physical infrastructure and the locomotives and shunting teams needed to perform the operations. An operations-time-space network [2] representing the rail station and the maritime terminals (either the origin or the destination of the trains) and the operations that might be performed in each zone of the port system is used for modelling and solving the problem. The nodes of the network, representing the zones and the operations to execute on trains, are replicated for each time interval of the considered schedule horizon. Vertical arcs represent the transfer of a train from one zone to another one, i.e., the end of a given operation in t and the simultaneous begin of the following required operation. The horizontal arcs represent the time spent by a train in a given zone for the execution of the required operation. Other than the classical constraints of the flow model, we have constraints related to the capacity of tracks, the limit on the number of simultaneous shunting operation, the required duration of each operation. The proposed network is able to deal with different types of capacity and time constraints that characterized the most part of logistic problems. This work introduces the main characteristics of the proposed network and addresses how it is possible to use this operations-time-space network to model port areas characterized by different layouts and capacity constraints. In fact, one of its strengths is the flexibility which allows to adapt the network to different systems with relative simplicity. In particular, some focus on specificities of real port systems in Italy which can be modelled with this kind on network will be shown. [1] D. Ambrosino, V. Asta, T.G. Crainic. Port Rail Shunting Scheduling Problem, Technical report CIRRELT 2022-02 [2] Ambrosino, D., Asta, V. (2021) An innovative operation-time-space network for solving different logistic problems with capacity and time constraints. Networks, 78 (3), pp. 350-367.

Nr:	13
Title:	Decision Policy for Intraoperative Ultrasound Imaging Technology for Brain Cancer Surgery
Authors:	Manoj Kumar, Narayan Rangaraj, Santosh Noronha, Aliasgar Moiyadi, Prakash Shetty and Vikas K. Singh
Abstract:	This research addresses the choice of intraoperative ultrasound (US) imaging technologies for glioma brain cancer surgery. The supporting imaging technologies available to surgeons are 2D ultrasound (less expensive and faster to use) and 3D ultrasound (more costly and time-consuming). The prime objective of the surgery is to remove the maximum tumor without worsening the post-operative neurological conditions of the patient. We explore the factors that influence these imaging technologies. These factors include the patient and tumor characteristics and the surgeon's experience. We also analyze the eventual effectiveness of surgery depending on the type of US imaging used. The data analysis is performed using statistical and machine learning techniques, such as random forest and logistic regression. The actual surgery data collected from the electronic records of a tertiary cancer referral center is used for our data analysis. The results indicate that seven factors are enough to determine the type of ultrasound. Moreover, our data analysis highlights the regions of patient and tumor characteristics where the aid of either ultrasound type results in a comparable outcome of the surgery.