EDENSPACE: Engineering Dependable Cyber-Physical Spaces

Innovations in various technological fields have led to a cyber-physical world where the boundary between the physical and the computational aspects gets increasingly blurry. This phenomenon is reflected in the notion of a spatially-distributed and software-intensive system where computational elements heavily interact with physical entities, especially with Internet of Things devices and along active human agents. The emergence of such systems which incorporate functions of sensing, actuation, and control while making decisions in a predictive or adaptive manner, is a key characteristic of today`s environments, which often permeate critical aspects of our society: How can a smart hospital be designed such that response of medical personnel is guaranteed? How can quality of service be maximized by a smart bike-sharing system in operation? How can assurances be obtained about a swarm of unmanned aerial vehicles providing autonomously emergency response in a city within a disaster scenario? Answering the above questions requires a first-class conception of a cyber- physical space (CPSp) - a spatial environment comprised of both computational and physical elements which are interrelated forming some composite structure. From a software engineering perspective, spatially- distributed software-intensive systems live within a dynamic environment populated with devices, human agents, changing context and/or localized resources. This implies having analyzable models in place, used to observe, evaluate and react to a constantly changing space. EDENSPACE addresses the systematic design and operation of dependable space-dependent systems, arguing that the dynamic spatial environments they indicate bring many challenges which arise in a similar way in classical software engineering when considering requirements such as security, safety, or reliability. Thus, formal analysis and validation techniques such as model checking, or the ability of systems to self-adapt at runtime reacting to environmental changes, can be adopted for systematic engineering of spatially-distributed systems that reliably satisfy their requirements, rendering them dependable.

We increasingly live in spaces that are both physical and computational, and where these two aspects are intertwined: EDENSPACE's objective was to enable reasoning on such contemporary environments, where computational and physical elements interact -- for example, internet-of-things devices, robots and humans may coexist within a smart city. Those systems that emerge are software-intensive, as software dynamically defines and controls them. Systems can be complex, and both their design and operational management should be performed in a systematic way. Think of a bike-sharing fleet within a city, thousands of sensor devices within a forest monitoring for fires, or a manufacturing floor co-habited by humans and robots. Those represent important classes of systems, referred to as cyber-physical, robotic or internet-of-things systems. Their diversity and dynamicity, our desire for them to behave timely and in a correct way, as well as the sheer size and complexity that they exhibit represent challenges. For example, shared bikes should be available when needed, sensing data should be analyzed correctly, or robots should not pose safety issues for humans. Such challenges are exacerbated by the fact that computing features are increasingly integrated into the physical world. This calls for new paradigms, techniques and methods. Those should address how to design them, how should they handle data and computation, or how to ensure that they behave in a correct way. To design such new kinds of complex systems, it is crucial to analyze, specify, and then verify their expected properties. Often properties are classied in functional vs non-functional ones, where the former capture the expected results of the system, whereas the latter correspond to its complementary properties - such as space, time, safety, security, fault tolerance, continuous adaptation, communication, or cost - and are no less relevant than the functional ones. Model-based techniques can considerably simplify and add rigor to the design process. Moreover, such systems call for automated management facilities when they are in operation, in order to observe their environment and possibly react when it changes, towards satisfaction of system properties. To do so, EDENSPACE employed principles from software engineering and distributed systems that have sound mathematical foundations, to ensure that systems we build are dependable. EDENSPACE addressed a wide range of problems within that domain -- ranging from developing appropriate representations, to techniques for moving computation close to users in order to enable responsive applications, or methods to ensure that infrastructures composed of cloud and internet-of-things devices operate in the correct way.