Feature-oriented requirements and architecture
Model-based product-line engineering relies on features—user-visible requirements—as basic units of product variability. Feature-orientation promotes independent development of features, enabling incremental and parallel development, and outsourcing. A fundamental challenge in feature-oriented development is feature interactions (FIs). They arise when integrating separately developed features, and can result in unintended FIs are unexpected or undesirable behaviors that result from feature composition. For example, the software controllers for three braking features of the 2010 Toyota Prius interacted badly, reducing drivers’ overall ability to brake and leading to multiple crashes and injuries. Bad FIs can also occur in human-computer interaction, such as when the driver potentially confuses similar warning signals from different driver-assistance systems.
The objective of this project is to provide an effective method and architecture to resolve feature interactions in automotive and aerospace applications.
Mechatronics systems modelling, design exploration, and optimization
Cars and airplanes contain complex mechatronic systems with hundreds of sensors, actuators, and processors. Their design must consider concerns and qualities such as real-time schedulability, resource utilization, power consumption, dependability, and the cost and mass of the resulting products. Major challenges in the design of such systems are: the huge size of the design space, induced by alternative sensor and actuator technologies, computing platforms, system deployments and the variability of customer requirements; uncertainty and complexity in the modelling of the qualities in early design; and the limited scalability of current design exploration and multi-objective optimization technologies. As a result, engineers may explore a subset of suboptimal designs.
The objectives of this project are to provide a method to model the design space of cyber-physical systems (CPS) and deal with the uncertainty in early design phases; and techniques to scale design exploration and optimization to large industrial CPS.
Model-based continuous verification and validation (V&V) of CPS product-lines
The current approach to engineering cyber-physical systems (CPS) has been described as "massively parallel development" of subsystems followed by a "big-bang integration," which is very costly because problems are discovered late in the development process when they are expensive to fix. Further, the final integration involves very extensive testing in test vehicles, which is also very expensive. Model-based engineering has been proposed as an approach to address these problems by enabling continuous integration and V&V during development by building a "virtual vehicle" comprising models of controllers, physical systems, and the vehicle environment. A major challenge is the large size of such models, which can easily comprise millions of elements, leading to slow simulation, analysis, and development cycles. This problem is exacerbated by product-line variability, which requires simulating and analyzing many system variants.
The objective of this project is to create methods and tools for scalable and continuous V&V using virtual vehicles.
Runtime assurance and data mining for CPS
The growing complexity of CPS requires solutions that ensure dependable operation even in the presence of defects. Runtime monitoring, verification, and recovery together with fault-tolerant designs are necessary techniques for building such systems. The research will investigate how to monitor and trace large mission-critical applications and how to establish whether the system is in a safe state. It will include work on observability of systems with temporal constraints, mining of data traces, and self-healing and reconfiguration. We will also investigate the definition and analysis of fault-tolerance mechanisms and study fault-tolerant software architectures, such as asymmetric architectures, that separate safety from control. We will build on our previous work on time-triggered runtime verification and low-cost fault detection.
The objective of this project is to advance the state of the art in debugging for security properties and develop architectural principles and techniques for precluding vulnerabilities such as code injection attacks and leakage of private information and to improve scalability of collection, cloud-based storage, and mining of vehicle data for diagnostic and design optimization.