Document Type

Thesis

Degree Name

Master of Applied Computing

Department

Physics and Computer Science

Program Name/Specialization

Applied Computing

Faculty/School

Faculty of Science

First Advisor

Usama Mir

Advisor Role

Main supervisor

Second Advisor

Dariush Ebrahimi

Advisor Role

Second supervisor

Abstract

Parking facility search is a major challenge in sustainable and intelligent urban transportation systems. In dense urban road networks, the time spent looking for an available parking space adds to traffic congestion, travel delay, inefficient road usage, higher fuel consumption, and driver frustration. Smart Parking Systems (SPS) have been developed to support parking recommendation through sensing, communication, and intelligent decision making. However, giving a reliable dynamic recommendation is still difficult, because traffic conditions and parking availability can change while the driver is still traveling toward the destination. As a result, a recommendation that is suitable at the start of a route may not remain suitable later in the same route. Another limitation is that many existing systems perform the recommendation only at the parking lot level. In these approaches, a parking lot is treated as a single available parking option, and the individual parking spots inside the lot are not considered. However, parking lots differ in availability, parking cost, driving time, and walking distance to the destination. Parking spots inside the same lot also differ in availability, spot-specific cost, internal driving time from the entrance, internal walking time to the pedestrian exit, and external walking time to the final destination. Therefore, a complete parking recommendation needs to cover both levels, which include both the parking lot and the individual parking spots within the parking lot. This thesis addresses the above problem by developing two related frameworks. The first framework focuses on parking lot level recommendation and is referred to as the Dynamic Parking Lot Optimization Framework (DPLOF). DPLOF recomputes the driving route and the recommended parking lot from driver’s current location using the latest traffic information available at that point. The framework considers driving time, walking time, and parking cost, together with user-defined constraints such as maximum walking distance and maximum parking cost. The second framework focuses on parking spot level recommendation and is referred to as the Dynamic Parking Spot Optimization Framework (DPSOF). DPSOF continues the decision after a lot has been selected and assigns an individual parking spot inside that lot. In this framework, each available spot is modeled as a separate parking option with its own availability, spot-specific cost, internal driving time, internal walking time, external walking time, and proximity to the pedestrian exit. For both frameworks, this thesis develops optimization-based models and evaluates different solution methods for small-, medium-, and large- scale urban road networks. For parking lot level recommendation, the optimization model is first applied to small-scale networks to obtain optimal results and to check whether the formulation works as intended. Since solving the optimization model repeatedly becomes computationally difficult for larger networks, a heuristic method is developed for the medium- and large-scale cases. This heuristic is then compared with meta-heuristic methods such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Ant Colony Optimization (ACO). For parking spot level recommendation, an optimization model is also formulated, and the same three meta-heuristic methods are applied as scalable comparative methods, so that both problems are studied using a consistent set of solution approaches. The proposed frameworks are evaluated using synthetic urban-like datasets and a real road network case study built in Simulation of Urban MObility (SUMO). The synthetic datasets enable controlled variation of network size and experimental parameters, while the SUMO-based network provides a real-world urban road network with directed road links and actual road connectivity. The results for the lot level framework show that the recommendation can be updated from the current vehicle position when traffic conditions change during travel, and that the proposed heuristic gives solution quality close to the optimization model at a much lower computational cost. The spot level framework then narrows the decision further by selecting a suitable parking spot inside the recommended parking lot. This thesis presents two dynamic parking recommendation frameworks, one at the parking lot level and the other at the parking spot level. By considering both levels together instead of treating them separately, the proposed work aims to reduce parking facility search time and to support better use of the parking facilities that already exist in urban road networks.

Convocation Year

2026

Convocation Season

Fall

Download

Available for download on Saturday, February 27, 2027

Share

COinS