Abstract

Abstract
This paper presents an interactive game designed to improve users’ experience related to driving behaviour, as well as to provide decision support in this context. This paper explores machine learning (ML) methods to enhance the decision-making and automation in a gaming environment. It examines various ML strategies, including supervised, unsupervised, and Reinforcement Learning (RL), emphasizing RL’s effectiveness in interactive environments and its combination with Deep Learning, culminating in Deep Reinforcement Learning (DRL) for intricate decision-making processes. By leveraging these concepts, a practical application considering a gaming scenario is presented, which replicates vehicle behaviour simulations from real-world driving scenarios. Ultimately, the objective of this research is to contribute to the ML and artificial intelligence (AI) fields by introducing methods that could transform the way player agents adapt and interact with the environment and other agents decisions, leading to more authentic and fluid gaming experiences. Additionally, by considering recreational and serious games as case studies, this work aims to demonstrate the versatility of these methods, providing a rich, dynamic environment for testing the adaptability and responsiveness, while can also offer a context for applying these advancements to simulate and solve real-world problems in the complex and dynamic domain of mobility. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
Artificial Intelligence (AI) is now ubiquitous in daily life, significantly impacting society by supporting decision-making. However, in many application areas, understanding the rationale behind AI decisions is crucial, highlighting the need for explainable AI (XAI). AI algorithms often lack transparency, making it hard to understand their inner workings. This work presents an overview of XAI solutions for decision support in mobility context. It addresses the complexity of explaining decision support models by offering explanations in various formats tailored to different user profiles. By integrating language models, XAI models may generate texts with varying technical detail levels, aiding ethical AI deployment and bridging the gap between complex models and human interpretability. This work explores the need for flexible explanation formats, supporting varied user profiles with graphical, textual, and tabular explanations. By integrating natural language processing models personalized explanations that are accurate, understandable, and accessible to a diverse audience can be generated. This study ultimately aims to support the task of making XAI robust and user-friendly, boosting its widespread use and application. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
The adoption of residential photovoltaic (PV) systems to mitigate the effects of climate change has been incentivized in recent years by government policies. Due to the impacts of these systems on the energy mix and the electrical grid, it is essential to understand how these technologies will expand in urban areas. To fulfill that need, this article presents an innovative method for modeling the diffusion of residential PV systems in urban environments that employs spatial analysis and urban characteristics to identify residences at the subarea level with the potential for installing PV systems, along with temporal analysis to project the adoption growth of these systems over time. This approach integrates urban characteristics such as population density, socioeconomic data, public environmental awareness, rooftop space availability, and population interest in new technologies. Results for the diffusion of PV systems in a Brazilian city are compared with real adoption data. The results are presented in thematic maps showing the spatiotemporal distribution of potential adopters of PV systems. This information is essential for creating efficient decarbonization plans because, while many households can afford these systems, interest in new technologies and knowledge of the benefits of clean energy are also necessary for their adoption.

Abstract
The number of electric vehicles has been increasing significantly due to various factors, such as the higher prices of fossil fuels, concerns about the increasing pollution, and the resulting incentive to use energy from renewable sources. There are currently a few charging facilities, which are still quite scattered, and several are still experimental, requiring appropriate planning of this infrastructure in order to support the growing number of electric vehicles adequately. Thus, optimising the location of charging stations becomes a critical issue, which can be achieved through the application of mathematical models and data analysis tools. An example is genetic algorithms, which have demonstrated their versatility in solving complex optimisation problems, especially those involving multiple variables. This work presents a proposal for a more comprehensive genetic algorithm model that encompasses all variables from the perspectives of all entities involved. Its experimentation was conducted using real data, with the aim of finding the best combination of locations, minimising the total number of stations and maximising the coverage of the area under study. Thus, it is essential to carefully consider user preferences, accessibility, energy demand, and existing electrical infrastructure to ensure an effective and sustainable installation. The findings highlight the crucial role of these computing tools in addressing complex problems from various viewpoints, leading to solutions that cater to the needs of all parties involved. While not necessarily perfect, these solutions represent a balanced compromise across multiple dimensions of the problem. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
The effectiveness of optimizing complex problems is closely linked to the configuration of parameters in search algorithms, especially when considering metaheuristic optimization models. Although various automated methods for algorithm configuration have been proposed to alleviate users from manually tuning parameters, there is still unexplored potential in dynamically adjusting certain algorithm parameters during execution, which can lead to enhanced performance. The main objective is to comparatively analyze the effectiveness of manual parameter tuning compared to a dynamic online configuration approach based on reinforcement learning. To this end, the State-Action-Reward-State-Action (SARSA) algorithm is adapted to adjust the parameters of a genetic algorithm, namely population size, crossover rate, mutation rate, and number of generations. Tests are conducted with these two methods on benchmark functions commonly used in the literature. Additionally, the proposed model has been evaluated in a practical problem of optimizing energy trading portfolios in the electricity market. Results indicate that the reinforcement learning-based algorithm tends to achieve seemingly better results than manual configuration, while maintaining very similar execution times. This result suggests that online parameter tuning approaches may be more effective and offer a viable alternative for optimization in metaheuristic algorithms. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.