We are assisting to a massive diffusion of networked ICT devices increasingly entangled with our physical and social world. This is making our everyday life and economic activities increasingly – if not fully – dependent on the functionalities of large-scale distributed software-intensive (i.e., heavily relying on software) systems.Our streets will be soon populated by myriads of self-driving cars, interacting with each other and with traffic control infrastructures to globally improve urban mobility and its sustainability.  Most urban and agricultural activities will be delegated to teams of robots, typically interacting with each other and with the surrounding physical work to perform their tasks in a collectively coordinate way. The Internet of things vision, to unfold its potentials, will require coordinating the sensing and actuating activities of millions of networked physical objects and their interactions with the physical work. Finally, the vision of “smart matter” consider the future possibility of defining novel materials made up of smart, software-defined particles, that will make it possible to produce adaptive self-shaping and self-repairing artifacts other than possibly pave the way for smart drugs and nano-level in-body therapies.The kind of systems are indeed exemplary instantiations of CAS, Collective Adaptive Systems. However, the above classes of CAS exhibit distinguishing characteristics with respect to the simpler classes of CAS that have been so far investigated in mainstream CAS researches (which typically deal with CAS composed of homogeneous components with limited autonomy, situated in rather controllable artificial or biological environments) described above exhibit common characteristics that make them substantially different from the classes of systems typically addressed by traditional programming and software engineering technique. In fact:

• They are inherently heterogeneous and socio-technical, since they require orchestrating the activities of components as diverse as software agents, a variety of sensors and actuators, robots, and – last but not least – they involve the active contributions of humans with their peculiar capabilities and competences.
• They are situated in dynamic and unpredictable physical and social environments, where their components are required to be context-aware and socially-aware in their interaction, and where consequently any coordination scheme has to be adaptive to the context.
• They require the capability of effectively coordinating the activities of a huge number (up to the millions) of decentralized autonomous components. This implies the impossibility of enacting some centralized scheme of coordination of the activities, as well as the impossibility of enforcing full control over the activities and interactions of some (if not most) of the components.

The above characteristics suggest novel challenges for researches in CAS, to build more solid foundations for the understanding and harnessing of the emerging classes of CAS of which we will be components and in which, in some way, we will be forced to live and interact.However, other than understanding the foundations of these novel classes of CAS, this document has emphasized that there is need of novel engineering approaches to pave the way for the systematic development of systems that, despite the limited controllability and dependability of the individuals and their situations, can exhibit predictable and dependable collective behaviour, capable of serving specific purposes. That is, engineering such systems implies the capability of governing their collective behaviour, where such governance has a twofold meaning:

• Shaping the collective adaptive behaviour of the system, that is assuring at the design level that the system will be able to serve its purpose at the global macro level,  despite the impossibility of controlling each individual components and of accurately predicting the dynamics of the environment in which they situates
• Controlling the dynamic collective behaviour of collective systems and its impact on the socio-physical environment in which it situates, to make it possible to enforce constraints (e.g., safety rules) in its behaviours and in its interactions with other systems, and possibly to tune its collective behaviour in order to dynamically meet emerging requirements.

Engineering challenges for governing the behaviour of collective systems imply the definition of an overall framework that should include:

• The definition of novel models for expressing required global behaviours, where functional requirements should express the overall mission of a system, and non-functional requirements can be possibly expressed as constraints over the way such mission is accomplished.
• The definition of qualitative and quantitative models for understanding collective behaviours (e.g., formal models, model checkers, simulators), and tools (e.g., programming languages) facilitating the shaping of specific global behaviours, and the definition of control models and associated decentralized control tools to make it possible to dynamically tune or steer the behaviour of deployed systems.

Specific attention in the definition of the above models and tools should be placed on a number of challenging issues that are peculiar of collective systems dived in complex socio-physical environment and that are particularly critical for the proper functioning and social acceptance. In particular:1) Transparency and polycentric governance. Collective systems will be called to support our everyday activity in a variety of situations, and will be managed by a variety of stakeholders and will interact with a variety of other systems. This raises the issue of understanding how it is possible to guarantee specific purposes in a transparent (understandable and appreciable by humans) way, and in face of multiple and possibly  conflicting goals within interacting systems and within a single system.2) Higher order emergence.  When individual components of a system are situation-aware they can be made perceiving the global behaviour of the system, and this can possibly influence their individual behaviour, affecting in turn the global behaviours. Such phenomena of second-order (higher-order in general) emergence have not been so far investigated in the area of collective adaptive systems.3) Handling dynamic and uncontrollable environments. Due to unavoidable uncertainty in modelling CAS, there will be need to build live testing environments, running in a ``mirror world'' reflecting real-world data, behaviours and interactions while protecting real individuals, continuously providing feedbacks on CAS behaviour and absorbing patches, upgrades, intervention of engineers.

The results of attacking the above challenges should eventually be a comprehensive “intervention framework” for governing complex socio-technical CAS.

At the time this document is being released, we noticed that the above engineering issues have some relations with FET-PROACTIVE in the 2016-2017 Workprogramme, and in particular with some topics associated to call “FETPROACT-01-2016: FET Proactive: emerging themes and communities”. In particular, the call refers to a number of issues and applications/societal scenarios (being human in a technological word, bio-electronic medicines and therapies, new computing paradigms and their technologies, complex bottom-up construction) that are strictly related to the issue of governing collective adaptive systems. That is, all of the systems/technologies of interest to the call somehow involve collective systems and implicitly involve some means to govern some sort of collective systems. Yet, we consider that a specific funding action addressing the engineering and governance foundations of the novel identified classes of CAS is needed and urgent.

Get the complete FoCAS Manifesto at http://www.focas.eu/manifesto/