Towards Intelligent Control of Vehicle Depot Parking and Dispatching

Abstract

Vehicle depots are a challenging real-world environment for online decision-making. Vehicles are often parked in lanes, one behind another, which allows space-efficient use of the depot but creates blocking constraints: only the front vehicle in a lane can leave directly. As a result, each parking decision affects whether suitable vehicles will later be available for scheduled tours. This becomes even more complex when vehicle types, battery state of charge, charging processes, heterogeneous lane layouts, and operational disturbances such as delays must also be taken into account.

The goal of this thesis is to investigate robust decision strategies for real-time depot control. The focus lies on selecting suitable actions for arriving and departing vehicles, as well as necessary moves within the depot, in order to avoid operational deadlocks. Building on an existing Java-based framework and simulator, the thesis will explore whether more general and transferable strategies can be developed that do not depend on depot-specific heuristics. Possible directions include interpretable rule-based or search-based methods, machine-learning or AI-based approaches, or hybrid combinations of both.

Background

Efficient depot operations are essential for reliable public transport, especially with the increasing use of electric vehicles. In many depots, vehicles are parked in lanes, one behind another. This parking structure is space-efficient, but it imposes strong operational constraints, since only the first vehicle in a lane is immediately available for departure. A poor parking decision can therefore lead to situations in which an arriving vehicle cannot be parked, or a scheduled outgoing tour cannot be served by any feasible vehicle.

In addition to these accessibility constraints, depot control must account for vehicle type compatibility, battery state of charge, charging processes, varying vehicle lengths, heterogeneous parking layouts, and disturbances such as delays or temporarily unavailable parking slots. Since decisions must be made online and within strict runtime limits, the problem is a practically relevant and computationally difficult discrete optimization task.

An existing software framework and benchmarker in Java are already available. Current baseline approaches work well for a specific depot setting, but rely on heuristics tailored to this individual depot and may therefore generalize poorly. The purpose of this thesis is to investigate alternative action-selection strategies that are more robust and transferable across different depots and operating conditions.

Information

• Master’s thesis for one person (Bachelor’s thesis also possible)
• 30% theory, 70% practical application

Prerequisites

Ideally, familiarity with some of the following topics:

• algorithms and data structures
• discrete optimization / operations research
• search methods
• machine learning / AI
• Java programming

Scope of the Thesis

The goal of this thesis is to develop and evaluate a robust strategy for online vehicle depot control that generalizes better across different depots and operating conditions than existing heuristic approaches. The expected outcome is an implemented and experimentally validated method that can reliably support real-time decisions for parking, dispatching, and in-lane vehicle movements while taking charging and operational constraints into account. The work should contribute to the practical automation of depot operations and provide a basis for more scalable and transferable decision-support systems in public transport.

Literature Review
Investigation of existing approaches in discrete optimization, online decision-making, search algorithms, and machine learning / AI for solving constrained operational planning problems.

Method Development
Analysis of the existing framework and baseline approaches. Design, implementation, and improvement of a new strategy for selecting actions in real time. Possible directions include white-box rule-based or search-based methods, black-box AI-based approaches, or hybrid methods combining both. After an initial analysis, the student may choose the most promising direction.

Evaluation
Comparative analysis of the developed method against existing baseline approaches using the available simulator and benchmarker. Evaluation should consider multi-day simulations, real depot instances, and, if useful, additional synthetic depot layouts. Performance metrics may include feasibility, robustness under disturbances, runtime behavior, and the ability to avoid deadlocks.

Bewerben

Interested? Then please send us your CV, along with a brief statement explaining why you’re interested in this thesis, to thesis@scs.ch.