Sizing a fleet of service vehicles remains a major challenge for many companies, even with the widespread use of ERP systems that provide access to large amounts of data. This difficulty is largely explained by the stochastic nature of the problem: service vehicles are dispatched on demand to handle incidents that are difficult to predict. Thus, unlike interventions that can be planned in advance (by constructing vehicle routes beforehand), when a vehicle is dispatched—where the objective is to minimize travel times—the times of occurrence and the locations of the interventions to be handled are not known precisely.
Traditional sizing approaches found in the literature often struggle to effectively capture the stochastic nature of this problem, as well as to establish a link between the temporal and spatial components of intervention demand. Finally, the transition to electric vehicles presents a major challenge for the industry that has been little studied in this specific context.
In this perspective, this master’s thesis proposes a data-driven approach to sizing a fleet of electric service vehicles facing stochastic demand. This methodology is divided into three main parts: 1) forecasting future interventions, 2) developing a simulation model to test new strategies for assigning service vehicles to interventions, and 3) developing heuristics to construct vehicle routes in cases where intervention forecasts are sufficiently reliable. Particular attention was also given to the correction of collected data. To this end, a methodology was developed, based on the simulation model, to scale the number of interventions.
What distinguishes this approach from existing work is the application of methods used in related fields to a domain that has been little explored: electric service vehicles responding to (hard-to-predict) interventions that affect the level of public transportation service. Its originality lies in the processing of data to mimic the spatiotemporal occurrence of intervention requests, combined first with simulation and subsequently with heuristics used for sizing the service vehicle fleet.
Our results show that the transition to electric vehicles can represent an opportunity to reduce fleet size at the STM, depending on the assumptions considered, with reductions of up to 68% in the number of vehicles compared to the current fleet. Some of the tested strategies also resulted in lower average intervention times. Finally, centralizing vehicles within a single depot would not substantially increase intervention times, which would lead to savings in charging infrastructure through its shared use across the entire fleet.
| Date | 19 Dec 2025 |
|---|
| Original language | French |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Tasseda Boukherroub (Supervisor) & Sana Berraf-Belmokhtar (Co-supervisor) |
|---|
Jezequel, S. (Author),
Boukherroub (Supervisor) & Berraf-Belmokhtar (Co-supervisor),
19 Dec 2025Student thesis: Master's thesis › Master in Engineering: Automated Manufacturing Engineering