Towards Customizable CPS: Composability, Efficiency and Predictability

Abstract

Today, many industrial products are defined by software, and therefore customizable by installing new applications on demand - their functionalities are implemented by software and can be modified and extended by software updates. This trend towards customizable products is extending into all domains of IT, including Cyber-Physical Systems (CPS) such as cars, robotics, and medical devices. However, these systems are often highly safety-critical. The current state-of-practice allows hardly any modifications once safety-critical systems are put in operation. This is due to the lack of techniques to preserve crucial safety conditions for the modified system, which severely restricts the benefits of software. This work aims at new paradigms and technologies for the design and safe software updates of CPS at operation-time – subject to stringent timing constraints, dynamic workloads, and limited resources on complex computing platforms. Essentially there are three key challenges: Composability, Resource-Efficiency and Predictability to enable modular, incremental and safe software updates over system life-time in use. We present research directions to address these challenges: (1) Open architectures and implementation schemes for building composable systems, (2) Fundamental issues in real-time scheduling aiming at a theory of multi-resource (inc. multiprocessor) scheduling, and (3) New-generation techniques and tools for fully separated verification of timing and functional properties of real-time systems with significantly improved efficiency and scalability. The tools shall support not only verification, but also code generation tailored for both co-simulation (interfaced) with existing design tools such as Open Modelica (for modeling and simulation of physical components), and deployment on given computing platforms.