Model-Driven Engineering (MDE) enables the automatic generation of performance models from software design models by model transformations. The performance models are used for performance analysis of the software under development, to guide the design choices from an early stage and to ensure that the system will meet its performance requirements. The software model evolves during development, so inconsistencies may appear between the software and performance models. This research aims at keeping the software and performance models synchronized. An important concept in model transformations is the mapping between the source and target (meta)models, which can be specified in a reusable manner with the help of mapping patterns. In this research we propose a subclass of such patterns, called containment-based mapping patterns. used to map a group of containment-related source model elements into a group of target model elements. We focus on these patterns because the containment relationship is frequently found in metamodel specifications. The containment mapping patterns are applied in the context of a non-trivial transformation from UML software models extended with MARTE performance annotations into Layered Queueing Network (LQN) performance models. We show how the mapping patterns can be applied for designing the transformation rules for a batch transformation implemented in a specific language. (The complete implementation of the batch transformation was done in separate work). In this research, we extend the batch transformation to generate, beside the target model, a traceability model between the mapped source and target elements. After solving the generated LQN model with an existing solver, the performance results are fed back to the software model by following the cross-model trace links. The next objective of the research is to design (based on the mapping patterns), implement and evaluate an incremental change propagation (ICP) approach to re-synchronize the software and performance models. During the development process, when the software model evolves, we detect the changes with the Eclipse EMF Compare tool, then incrementally propagate them to the LQN model. The proposed ICP is implemented with the Epsilon Object Language (EOL) and is evaluated by applying it to a set of case studies.

This paper proposes a higher-order transformation (HOT) for realizing Incremental Change Propagation (ICP) from software UML models extended with performance annotations to performance Layered Queueing Network (LQN) models. Such a transformation is necessary for integrating quantitative performance analysis into the model-driven engineering of real-time systems. The entire process starts by automatically generating an LQN and a trace model from a UML model extended with MARTE annotations, with a batch Epsilon ETL transformation previously developed by the authors. The textual ETL transformation definition is translated to an ETL transformation model using the Epsilon Haetae tool. The ETL transformation model conforms to the ETL metamodel and represents the mapping between source and target models at a high level of abstraction. We use it to answer the question: what needs to be changed in the target model upon detecting changes in the source model? During the development process, when the UML model evolves, we detect such changes with the Eclipse EMF Compare tool, then incrementally propagate them to the LQN model to keep it synchronized. The extended approach is illustrated by applying it to an e-commerce model from the literature. The execution time of ICP is measured and compared to the traditional batch transformation.

This paper presents an intelligent, sensor-driven framework that integrates emerging technologies to deliver smart, reliable, and accessible transit assistance for visually impaired people (VIP). The proposed system leverages Internet of Things (IoT), Internet of Devices (IoD), GPS, and crowdsensing to collect multimodal data, comprising audio, video, and environmental signals, used to characterize and respond to users’ real-time mobility needs. A cloud-based architecture performs centralized data fusion and decision-making using artificial intelligence (AI) and machine learning (ML) algorithms, enabling rapid interpretation of sensor inputs and generation of personalized navigation guidance. The framework is implemented through a mobile application that coordinates data exchange between edge devices and cloud services, providing context-aware navigation, obstacle alerts, and two-way communication with transit operators. Unlike existing assistive mobility solutions, which rely primarily on static location services and lack cross-sensor integration, the proposed system introduces a unified AI-enabled sensing layer that supports dynamic adaptation to complex urban environments. The results demonstrate the framework’s potential to enhance autonomy, safety, and situational awareness for VIPs, offering a scalable foundation for inclusive smart transportation systems.