<< back to topics

Cyber Physical Systems (CPS)

A selection of papers from Bernhard Rumpe and the Software Engineering Group

Cyber physical systems (CPS) are software controlled, collaborating physical machines.

In [KRS12] we discuss that this new term arises mainly due to the increased ability of computers to sense their environment and to interact with their contexts in various ways. As consequence, CPS are usually designed as distributed networks of interacting nodes and physical devices (machines) that carry out certain tasks. Often some of these devices are mobile (robots or autonomous cars, but also smartphones, airplaines and drones) and interaction with humans is essential. CPS are therefore complex in several dimensions: they embody characteristics of physical, networked, computational-intensive, and of human-interactive systems. Furthermore, they typically cannot be developed as monolithic systems, but need to be developed as open, composable, evolving, and scalable architectures.

Nowadays, CPS are found in many domains, including aerospace, automotive, energy, healthcare, manufacturing, and robotics. Many distributed CPS use a virtual communication network mapped to the Internet or telecommunication infrastructure.

At the heart of CPS engineering suffers from the problem that control theory, built on integration and differentiation calculus used by almost any engineering discipline, and the digital theory of state machines are not very well integrated and thus do not allow us to describe CPS in an integrated way. Many attempts have been made, but a good standard yet has to emerge.

The complexity and heterogeneity of CPS introduces a wide conceptual gap between problem and solution domains. Model-driven engineering of such systems can decrease this gap by using models as abstractions and thus facilitate a more efficient development of robust CPS.

Modeling CPS

For the aviation domain, we have developed a modeling language [ZPK+11] that allows to specify flight conditions including trajectories, status of the airplanes and their devices, weather conditions, and pilot capabilities. This modeling language allows EuroControl to operationalize correct flight behavior as well as specify and detect "interesting events". As long term interest, we intensively do research on how to improve the engineering for distributed automotive systems as well. For example [HRR12] outlines our proposal for an architecture centric development approach, which we apply to robotics in [RRW13c] and [RRW14a].

CPS & Automotive

Automotive is a highly innovative CPS subdomain. Therefore we discuss in [GRJA12] what an OEMs needs to understand about costs arising from requirements complexities and from cross-plattform dependencies in their automotive development projects. Transforming a set of individual projects with similar requirements and technology into a product line for a central part of a car is discussed in [HRRW12].

In [BR12b] we discuss current and future processes and tools for development of autonomous driving cars based on our experiences in building such a car and using sophisticated simulation techniques for the context of autonomous robots (cars). In [BBR07] we describe that fully automatic simulation of the cars' cyber physical contexts' and fully automatic checking of the robots behavior leads to an highly efficient development process with high quality results.

CPS & Robotics

Robotics is another highly innovative CPS subdomain. It is characterized by an inherent heterogeneity of involved domains, platforms, and increasing set of challenges. Engineering of robotics applications requires composition and interaction of complex, distributed systems as well. We developed a component & connector architecture description language suitable for the specific challenges in robotics [RRW13c] as well as in [RRW14a] and partially position it as a requirements modeling language family in [RRW12].

CPS & Buildings

Smart and energy efficient buildings embody a lot of IT technology. There is a multitude of networked systems and sensors to continuously control the building's "behavior". We have built the Energy Navigator described in [KPR12] and [FPPR12]) to be able to model the specifications of such buildings in order to control the measured actual data against the desired specification, e.g to save energy. In [KLPR12] we discuss how such a specification approach improves development quality in the energy subdomain of CPS.


  1. CPS tackles two core challenges:
    • Lack of integration of calculus and automaton theory, and
    • Heterogeneity of domains need integration of heterogeneous modeling technologies as CPS requires cross domain solutions and techniques.
  2. Furthermore, CPS tend to be complex in functionality and because of distribution and quality needs.
  3. CPS are typically not built from scratch, but evolve as new components (services, devices, machines) are added.
  4. We have developed architectural modeling techniques to describe CPS and applied those to cars, robots and building infrastructures.

Further Topics:

Selected(!) Publications:

<< list of all topics list of all publications >>