İnci Sarıçiçek

Otonom elektrikli araçların, sıfır emisyonlu yapıları ve gerçek zamanlı karar verme yetenekleri sayesinde taşımacılık sistemlerindeki kullanımı hızla artmaktadır. Bununla birlikte, batarya kapasitesinin sınırlı olması, şarj süreçlerinin zaman alıcı yapısı ve şarj altyapısındaki yetersizlikler, klasik araç rotalama yaklaşımlarının doğrudan uygulanmasını güçleştirmekte, dinamik ve kısıt duyarlı çözüm yöntemlerine olan ihtiyacı artırmaktadır. Bu çalışmada, otonom elektrikli araçlar için gerçek zamanlı koşullara uyum sağlayabilen dinamik bir rota planlama yaklaşımı önerilmektedir. Depo merkezli dağıtım yapısı altında, her bir aracın başlangıç rotası Yasaklı Arama (TS) metasezgisel yöntemi kullanılarak oluşturulmuştur. Rotalama süreci boyunca araçların batarya seviyeleri sürekli olarak izlenmiş ve enerji yetersizliği durumunda araçlar en uygun şarj istasyonlarına yönlendirilmiştir. Şarj süreleri ile zaman penceresi ihlalleri, toplam maliyet fonksiyonuna entegre edilerek çözümün operasyonel açıdan uygulanabilirliği artırılmıştır. Önerilen yaklaşım, dinamik ortamlarda ortaya çıkan yeni müşteri talepleri, talep iptalleri, trafik yoğunluğundaki değişimler ve yol kapanma veya açılma gibi beklenmedik durumlara gerçek zamanlı olarak uyum sağlayacak biçimde tasarlanmıştır. Bu adaptif yapı sayesinde sistem, hizmet sürekliliğini korurken toplam operasyonel maliyetin kontrol altında tutulmasını mümkün kılmaktadır. Sonuç olarak geliştirilen metasezgisel tabanlı yöntem, çoklu araçlı ve gerçek zamanlı değişkenlere duyarlı otonom elektrikli araç rotalama problemleri için esnek, ölçeklenebilir ve etkili bir çözüm alternatifi sunmaktadır.

Designing fleet management systems with built-in fault tolerance across vehicles, operators, environment, communication, and software components is critical to ensuring the long-term sustainability of operations. A key design priority is to minimise the negative impacts of vehicle failures on operational processes. Ensuring operational continuity in the transport sector requires proactive strategies for anticipating and resolving such failures. In this study, rerouting is carried out using recovery plans that are proactively generated based on existing vehicle routes to address potential electric vehicle failures. The proposed system performs rerouting while considering time window constraints through the Regret-2 and 2-opt approaches. If a charging need arises on the newly created routes, the vehicle is directed to a charging station suitable for the route. Simulation results conducted within the scope of test scenarios demonstrate that the proposed approach not only ensures service continuity but also addresses battery, load, and time constraints. This study creates potential for a fault-tolerant planning system by offering proactive solutions for vehicle failures within the scope of dynamic rerouting in fleet management software.

This study introduces a lightweight yet systematic simulation-based verification framework for evaluating electric vehicle routing (EVR) solutions under realistic traffic and charging conditions. Rather than proposing new EVR optimization algorithm like heuristics, meta-heuristics or dynamics, our contribution is a framework that ingests routing outputs produced by any EVR algorithm, exposes these outputs to microscopic traffic and energy models, and conducts distortion testing (e.g., traffic jams, customer removal, charging-station relocation) to assess robustness. The framework reports comparable performance metrics—distance, travel time, energy consumption, state of charge (SoC) trajectories, charging waits—and quantifies plan–execution gaps via statistical measures. We demonstrate the framework on a campus-scale case study. Results show how distance-optimal plans may become energy-suboptimal once regenerative braking limits, auxiliary loads, and charging constraints are enforced, and how seemingly favorable changes (e.g., moving a charger closer) can paradoxically reduce final SoC due to capacity and timing effects. The framework is algorithm-agnostic, transparent, and reproducible; it complements, rather than replaces, classical EVR models by providing an execution-level lens. We discuss threats to validity arising from scale and parameterization and outline how the framework generalizes to larger settings. Overall, the work provides a verification analysis for stress-testing EVR outputs before deployment.

The Dynamic Electric Vehicle Routing Problem (D-EVRP) presents a framework that requires electric vehicles to meet demand with limited energy capacity. When dynamic demand flows and charging requirements are considered together, traditional methods cannot provide sufficient adaptation for real-time decision-making. Therefore, a learning-based approach was chosen to ensure that decision-making processes respond quickly to changing conditions. The solution utilizes a model with a Double Deep Q-Network (DDQN) architecture and a discrete valuation structure. Prioritized Experience Replay (PER) was implemented to increase model stability, allowing infrequent but effective experiments to contribute more to the learning process. The state representation is constructed using the vehicle’s location, battery level, load status, and current customer demands. Scalability is ensured by dividing customer locations into clusters using the K-means method, with each cluster handled by an independent representative. The approach was tested with real-world road data obtained from the Meşelik Campus of Osmangazi University in Eskişehir. Experiments conducted under different demand levels and data sizes have shown that the PER-assisted DDQN structure produces more stable and shorter route lengths in dynamic scenarios, but random selection, greedy method and genetic algorithm experience significant performance losses as dynamicity increases.

Sustainable transportation and green logistics are becoming increasingly important, and the efficient use of electric vehicles (EVs) plays a critical role. However, efficient path planning for EVs remains a major challenge due to limited driving range and the need for optimised charging strategies. Usually, path recommendations are made based on a single criterion. However, drivers may want to consider multiple criteria for path selection. This study focuses on building a multi-criteria path planning algorithm that incorporates driver preferences by considering total travel time, energy consumption and travelling distance. To obtain the appropriate recommendation, these three criteria are evaluated using the Analytic Hierarchy Process (AHP) and Dijkstra algorithm is used to identify roads that take into account driver preferences. Johnson technique was used to remove negative energy weights due to energy recovery and solved the incompatibility problem of the Dijkstra algorithm with negative edge weights. The results have shown the proposed algorithm can efficiently generate solutions designed based on driver preferences and is suitable for EV routing applications. This study presents a method to increase user satisfaction by aiming at the widespread adoption of EVs and emphasizes the importance of multi-criteria decision making in addressing the unique challenges of EVs.

Dial-a-Ride problem (DARP) is a specialized variant of the vehicle routing problem that focuses on designing optimal routes for users who specify pick-up and delivery requests between origins and destinations. This problem involves determining the most effective and efficient routes using limited vehicles and resources to offer responsive door-to-door transportation services for users with specific needs. Logistics service providers require systems that can deliver optimal solutions to these combinatorial problems within a reasonable time. Recently, there has been a significant increase in the use of artificial intelligence optimization algorithms such as meta-heuristics or learning-based approaches to solve such problems. Among the learning-based approaches, reinforcement learning has gained prominence for routing and scheduling tasks, owing to its ability to adaptively learn from complex state spaces and dynamically changing environments. In this study, a novel transformer-based deep reinforcement learning method is proposed to solve the Dial and Ride problem for a single service vehicle. In the proposed model, we adopt a modified transformer architecture instead of employing traditional linear layers, we integrate convolutional layers. To validate our approach, we conduct comprehensive experiments comparing our method against four well-known metaheuristic algorithms and a Deep Q-Network algorithm. The results indicate that proposed approach outperforms these techniques in terms of shorter total travel distances. Additionally, the proposed method is tested on a real-world scenario generated in the Buyukdere neighborhood of Eskisehir. The results demonstrate that the proposed method makes it possible to solve the problem within a reasonable time. This study confirms that the proposed deep reinforcement learning method can effectively address the Dial-a-Ride Problems.

The use of electric vehicles in urban transportation is increasing daily due to their energy efficiency and environmental friendliness. In last-mile logistics, route optimization must consider charging station locations while balancing operational costs and customer satisfaction. In this context, solutions for cost-oriented route optimization have been presented in the literature. On the other hand, customer satisfaction is also important for third-party logistics companies. This study discusses the Capacitated Electric Vehicle Routing Problem with Time Windows (CEVRPTW) encountered in last-mile logistics. This article defines the objective function of minimizing total tardiness and compares the routes between the service provider logistics company and the customer receiving the service. In this study, the CEVRPTW was solved for the minimum total tardiness objective function with the hybrid adaptive large neighborhood search (ALNS) algorithm. The success of ALNS was proven by comparing the differences between the optimal solutions obtained with the CPLEX Solver and the ALNS solutions. Tardiness objective function-specific operators for ALNS are proposed and supported by local search and VNS algorithms. The findings of this study contribute to the literature by analyzing the balance trade-offs between customer-oriented and cost-oriented and the effect of time windows on the number of vehicles.

In recent years, electric vehicles have become increasingly widespread, both in the logistics sector and in personal use. This increase, together with factors such as environmental concerns and government incentives, has brought energy consumption and range estimation issues to the forefront. In this study, the energy consumption of an electric cargo vehicle under different speed and load conditions is examined with an experimental and data-driven approach, and then used for range estimation. The raw data collected from the vehicle on the selected ~2 km route in Eskisehir Osmangazi University campus are combined into per-second samples with time synchronization and data cleaning. The route is divided into average of 150 m segments, and variables such as slope, energy consumption, and acceleration are calculated for each segment. Then, the data are used to train various machine learning models, such as Extra Trees, CatBoost, LightGBM, Voting Regressor, and XGBoost, and their performances regarding energy consumption-based range estimation are compared. The findings show that driving dynamics such as high speed and sudden acceleration, as well as road slope and load conditions, significantly shape the energy consumption and thus the remaining range. In particular, Extra Trees outperforms other machine learning models in terms of metrics such as R², RMSE and, MAE, with a reasonable computational time. The results provide applicable guidance in areas such as route optimization, smart battery management, and charging infrastructure to reduce range anxiety and increase the operational efficiency of electric vehicles.

The growth of the urban population and the increase in e-commerce activities have resulted in challenges for last-mile delivery. On the other hand, electric vehicles (EVs) have been introduced to last-mile delivery as an alternative to fossil fuel vehicles. Electric vehicles (EVs) not only play a pivotal role in reducing greenhouse gas emissions and air pollution but also contribute significantly to the development of more energy-efficient and environmentally sustainable urban transportation systems. Within these dynamics, the Electric Vehicle Routing Problem (EVRP) has begun to replace the Vehicle Routing Problem (VRP) in last-mile delivery. While classic vehicle routing ignores fueling, both the location of charging stations and charging time should be included in the Electric Vehicle Routing Problem due to the long recharging time. This study addresses the Capacitated EVRP (CEVRP) with a novel Q-learning algorithm. Q-learning is a model-free reinforcement learning algorithm designed to maximize an agent’s cumulative reward over time by selecting optimal actions. Additionally, a new dataset is also published for the EVRP considering field constraints. For the design of the dataset, real geographical positions have been used, located in the province of Eskisehir, Türkiye. It also includes environmental information, such as streets, intersections, and traffic density, unlike classical EVRP datasets. Optimal solutions are obtained for each instance of the EVRP by using the mathematical model. The results of the proposed Q-learning algorithm are compared with the optimal solutions of the presented dataset. Test results show that the proposed algorithm provides remarkable advantages in obtaining routes in a shorter time for EVs.