No abstract

Embedded systems are becoming increasingly complex and heterogeneous.Formal methods have proven effective in ensuring reliability and safety. However, they are hard to scale up. Modeling and Simulation (M&S)-based methods, on the other hand, deal effectively with scalability issues and provide the benefits of a risk-free testing environment. Yet, they are usually at most semi-formal, and models are not directly executed on the target hardware. To address the above challenges, we present a formal M&S-based kernel that runs on bare-metal and execute the original simulation models on the target hardware.

Development of Embedded Real-Time Systems is prone to error, and developing bug-free applications is expensive and no guarantees can be provided. We introduce the concept of Digital Quadruplet which includes: a 3D virtual representation of the physical world (a Digital Twin), a Discrete-Event formal model of the system of interest (called the “Digital Triplet”), which can be used for formal analysis as well as simulation studies, and a physical model of the real system under study for experimentation (called the “Digital Quadruplet”). We focus on the definition of the idea of a Digital Quadruplet and how to make these four apparati consistent and reusable. To do so, we use the Discrete-Event formal model as a center for both simulation and execution of the real-time embedded components with timing constraints, as well as a common mechanism for interfacing with the digital counterparts, providing model continuity throughout the process. Here we focus on a principal part of the Digital Quadruplet idea: the provision of an environment to allow models to be used for simulation (in virtual time), visualization, or execution in real-time. A Discrete-EVent Systems specifications (DEVS) kernel runs on bare-metal hardware platforms, avoiding the use of an Operating RTOS in the platform, and the combination with discrete-event modeling engineering.

With the rising popularity of embedded systems came new challenges due to evergrowing market application demands, increasing complexity and widening productivity gap. To deal with these issues, model-driven development promotes a higher level of abstraction during design and uses models as the primary artifacts that guide the product development. In fact, the fundamental principle is to construct a model of a system and then transform it into the real system. We focus on DEMES, a model-driven development methodology based on the Discrete Event System Specification (DEVS)-that defines a formal Modeling and Simulation framework for discrete event dynamic systems-, and especially the transition from simulated platform to execution platform, i.e. the embedded hardware. In this dissertation, we present bare-metal real-time executives that allow DEVS models to be executed on a target platform without the need of an operating system. This is particularly important for target platforms with limited resources. In addition to the real-time executives, we introduce a hardware abstract layer that supports several hardware peripheral libraries and fosters fast prototyping. We also illustrate the DEMES-driven development cycle with a particular case study: a line tracking robot application. Our contributions have resulted in a reduced footprint, increased performance and enhanced platform portability. I thank Dr Gabriel Wainer whose invaluable advice and ongoing support helped through my graduate studies and research. One of my objectives in graduate school was to work with and learn from some of the best in the field, and I had this privilege under his supervision. The past two years have been both a challenging and rewarding experience. I also thank the members of the Advanced Real-Time Simulation Laboratory for their help and comradeship. v

No abstract