The challenges which the engineering of self coordinating systems faces, are characterized by a number of research questions like: 

• How can optimal strategies be determined in the presence of partial or even unreliable information?
• How can heterogeneous components decide which information is relevant and which need not be considered?
• What algorithms might help us in reaching stable, robust, and desirable behavior in a distributed network?
• How can components find out about their coordination possibilities with cooperating or even competing components?
• What design and operation principles do we need for the underlying technical infrastructure and how can we maintain these principles when resources are restricted and parts can fail?
• How can we manage the secure exchange of information without a central public-key infrastructure and only limited resources?
• When do we need cross-border migrations between hardware and software?
• What modeling formalisms have to be supplied in order to enable the construction of adaptable systems?
• What analysis techniques are required to make software reflect on both its own and its environment’s behaviour and consequently change itself?
• What verification techniques can ensure the correctness of self-coordinating systems despite their inherent potential volatility?
• How can a tight integration between classical feedback controllers and the state-based discrete control be achieved?
  Today’s software engineering research by and large is not really focussing on answering these questions but is rather centered around much smaller systems than the ones mentioned above which, in addition, often consist of software "only" (e.g. information systems). It is also usually done by small, non-interdisciplinary research teams. Of course, that does not mean that we no longer need research on testing and analysis, on program understanding, or on formal modelling and verification approaches, to list only a few examples. However, this research must be combined with e.g. the development and layout of complex networks and their underlying infrastructure as well as research on corresponding data structures and algorithms and control theory.

This  paper outlines the components of a design theory for the development of PIS. The theory consists of a set of meta-requirements, a set of meta-design considerations, and a set of design method considerations.
PIS introduce new elements in multiple dimensions spanning different IS domains, such as Human- Computer Interaction (HCI) and Software Engineering, which admonish us to examine them as a new Information Systems class. In essence, PIS revisit the way we interact with computers by introducing new input modalities and system capabilities. So far, the interaction paradigm for Information Systems has been the desktop. Thus, the design and implementation of Information Systems was based on this paradigm. PIS extend this paradigm by introducing a set of novel characteristics that are summarised in the following

• PIS deal with non-traditional computing devices that merge seamlessly into the physical environment. 
• PIS simulate the way that humans interact with the physical world: PIS may  incorporate elements of ambient interactions with devices or objects from the physical space.
• PIS support a multitude of heterogeneous device types that differ in terms of size, shape (more diverse, ergonomic, and stylistic), and functionality (simple mobile phones, portable laptops, pagers, PDAs, sensors, and so on), providing continuous interaction which moves computing from a localised tool to a constant presence. 
• PIS support nomadic devices which may be carried around by users and present location-based information. 
• PIS need to support spontaneous networking, implying ad-hoc detection and linking of the participating devices into a temporary pervasive network creating dynamic dependencies among the linked devices.
• The participating elements of a pervasive system are highly embedded in the physical environment. 
• PIS emerge a revised viewpoint in the way we perceive system design. 
• in PIS it is highly unlikely for the system designer to know in advance the kinds of users who will be interacting with the system. Users may range from being vaguely familiar with IT to expert users. In addition, PIS users may be opportunistic in the sense that they may use the system only sporadically, implying that they may not be subject to training prior to system use.
• PIS introduce the property of context awareness as a result of the pervasive artefacts capability to collect, process, and manage environmental or user-related information on a realtime basis.

The goal of the Trust Analysis Methodology cite{Presti2005} is to help in the design of the pervasive system by highlighting the trust issues inherent to the system. It is a guide rather than a model as it does not define rigorously exact terms but rather provides a means to discover trust issues.
The methodology involves iteration over four steps followed by a final fifth step.
In particular:

• Scenario (step 1): is a short, fictional narrative, set in the near future that describes people’s daily lives, concentrating on their use of pervasive computing under examination. 
• Trust Analysis (step 2):   involves Trust Analysis Grid, where the rows of the grid correspond to vignettes in the scenario, while the columns of the grid correspond to categories of trust issues that will be checked against the vignettes. A vignette corresponds to one or several pieces of one or several sentences of the scenario and constitutes a cohesive group with regard to the trust issues.
• Peer Review (step 3): supports the extraction of trust issues from the perspective of another potential user, who may have a different view on trust issues. This peer review may be the occasion to discover some missing trust issues and complement the reviewer’s point of view.
• Scenario Refinement (step 4): in this step the scenario is refined by adding new text and vignettes, or removing existing ones.
• Guiding the Design of the Pervasive System (step 5): consists in using the Trust Analysis Grid to draw some guidelines in order to help in the design of the pervasive systems under consideration.

The MetaSelf  is a  development process for engineering dependable and controllable self-organising (SO) systems  consisting of four phases. The Requirement and Analysis phase identifies the functionality of the system along with self-* requirements
specifying where and when self-organisation is needed or desired.
The Design phase consists of two sub-phases:

• D1: the designer chooses architectural patterns and self-* mechanisms; 
• D2: the individual autonomous components are designed. The necessary metadata and policies are selected and described, and  the self-* mechanisms are simulated and possibly adapted or improved.
  The Implementation phase produces the metadata and executable policies. In the Verification phase, the designer makes sure that agents, the environment, artefacts and mechanisms work as desired. Potential faults and their consequences are identified, similar to the way failure modes and effects analysis works. Corrective measures (redesign or dependability policies) to tolerate or remove the identified faults are taken accordingly.

This work summarises five relevant methods for developing self-organising multi-agent systems and presents several method fragments coming from these methods. In particular, the paper presents Adelfe, Customised Unified Process, MetaSelf, General Methodology and the Simulation Driven Approach.

(Morandini2009) proposes to extend a goal-oriented engineering methodology to deal with the modelling of organisations that are able to self-organise in order to reach their goals in a changing environment. To deliver on this aim, the authors combine Tropos4AS (Morandini2008), an extension of TROPOS cite{Tropos2005} for adaptive systems, with concepts, guidelines and modelling steps from the ADELFE methodology, which provides a bottom-up approach for engineering collaborative multi-agent societies with an emergent behaviour.
Authors have, first of all, extended the Tropos4AS meta-model with concepts  such as skills, aptitudes, and actions  coming from Adelfe meta-model. Second, authors have introduced in the combined meta-model also the Agents&Artifact meta-model in order to model the environment.
After that, the authors  adapt the TROPOS4AS modelling process to modelling of the newly introduced concepts.
The resulting MAS has self-adaptation properties, having agents that are able to change their behaviour according to changes in the environment, and having
organisations that adapt themselves to changing needs.
Tropos4AS extends TROPOS goal models to enable the description of systems that shall be able to adapt to environmental changes. Tropos4AS introduces modelling of an actor’s perceived environment and of possible failures that can be identified and prevented by recovery activities. Moreover, goals can be annotated to define the runtime goal achievement behaviour, with various goal types and conditions related to the
environment, for goal creation, achievement and failure. The resulting design model is mapped into implementation language constructs to derive a code skeleton.