AEDUS2: Adaptable Environments for Distributed Ubiquitous Systems
| /People /Proposal |
Adaptable Environments for Distributed Ubiquitous Systems is an EPSRC Funded project addressing issues in: adaptive software environments; policy and security; and requirements engineering and modelling.
The original EPSRC Platform Grant has been renewed for a further 5 years.
Local Contact: Morris Sloman
The overall objectives include:
- To investigate adaptive software environments and infrastructures to support autonomic ubiquitous systems. This will include dynamic programming environments, self-managing architectures, context-aware systems, middleware and data integration techniques.
- To produce mechanisms that allows the sharing of data in highly dynamic heterogeneous environments consisting of both low power and larger computing nodes.
- To investigate policy-based tools and techniques for supporting trust, security and privacy in ubiquitous and large-scale systems. This will include security mechanisms for low-powered sensor nodes.
- To develop tools and techniques for requirements specification, refinement and analysis as well as behavioural modelling of adaptive systems.
- To work in multi-disciplinary teams to apply the above tools and techniques to applications such as healthcare, building environment control, sensor networks, embedded systems and collaborative unmanned autonomous vehicles.
- To focus on issues identified in the Ubiquitous Computing Grand Challenge
In each of the following areas we anticipate continued mainstream project funding from various sources but we will also make use of the platform grant to fund short-term 3-12 month investigative experimentation and initial studies.
Autonomic Ubiquitous Computing is one of the key areas the section will be focusing on over the next few years. In order to cope with the scale of ubiquitous systems, and be useable by non-technical people, it is essential for them to be self-managing, self-configuring, self-protecting and self-healing as people become increasingly dependent on such systems. A related unsolved problem is still making systems aware of and adaptive to a user’s current context. In healthcare applications, determining current context can be of critical importance. We consider that policy-based specification of self-management and adaptive strategy will be an important issue for all autonomic systems, not just for ubiquitous systems. However being able to specify strategy in terms of goals which can be refined to policies in low-powered portable devices is not achievable at present. A possible approach which we hope to investigate is how ubiquitous systems which interact with people can learn policies and preferences, possibly based on inferring context and tracking activity. Other work will investigate reinforcement learning for autonomic web services.
We aim to package the Beasties and software developed in the ANS project as a self-build kit that can be used to teach ubiquitous computing in third-level education institutes as well as used by the home enthusiast (a la Sinclair zx81). This is because of the component-based nature of the beastie hardware (there is no need for soldering) and the Tesserae component based software. To build compilers for differing languages and make the kit complete is something that would not be funded by a research council but will benefit teaching and research in the UK.
Privacy, Trust and Security. Determining a measure of trust based on evidence, recommendations, risk is very difficult both for ubiquitous and internet applications. However Trust has been considered an important aspect for making decisions relating to interactions with ‘unknown’ entities – this may involve forms of negotiation to build up trust relationships as well as using risk evaluation and analysis of past interactions to derive trust metrics. We will investigate how to make use of automated trust based decisions in a variety of applications including healthcare, military and e-commerce. Determining simple techniques for authenticating devices which need to collaborate in ubiquitous applications and distributing encryptions keys is still a problem due to limited computational power and wireless communication. Many systems need security mechanisms which can adapt to context as well as attacks. Empowering people to manage the information related to current activity and location and thus have control over their privacy is another important issue. An interesting approach is to investigate how to encapsulate privacy enforcement constraints into, data using techniques analogous to Digital Rights Management, so that privacy is still enforced when the data is transferred to new users.
Data Exchange': A significant problem in ubiquitous computing is the need to exchange data between various entities without the intervention of an expert user that can program the mapping between the data formats of the entities. We will investigate how recent work in P2P database systems and model management may be deployed in tools and techniques which automate the matching of data schemas held at entities, and then allows those entities to exchange information. We will need to adapt existing approaches to cope with the highly variable availability of entities, and rapid changes in group membership, and also optimise the utilisation of scare computational and wireless network resources found in many environments.
Requirements Engineering: A challenging problem in this environment is the requirements engineering and modelling of adaptive systems. The approach is that adaptability of a system should be supported by interleaving phases of model analysis and refinement of its requirements. The requirements stage should provide the potential and dimension of change. Initial work on the representation of some aspects of potential change using modes at the architecture level is promising. However, current architecture modelling, analysis and animation tools deal largely with a static snapshot of the systems architecture. Adaptive systems have high availability requirements and there is a need to model and analyse these architectures during adaptation as well as during normal operation. The aim is to integrate the architectural work with that on partial models and probabilistic models to give more realistic models of adaptation. This poses challenges in both theory and practice. It is essential that a sound semantics is found for integrating the various models: architecture modes, behaviour, probabilities, and partiality. A compositional approach to modelling and analysis appears to be key in dealing with adaptive systems. Our experience with LTSA and associated tools confirms that, for practical use, it is essential that this is also supported by tools. The intention is to examine both embedded systems, such as autonomous vehicles, and large scale systems, such as publish-subscribe networks.
We have started to investigate incorporating goal-oriented approaches to requirements engineering (GORE) as an extension to our event-based behaviour modelling approach. We have revisited synthesis from scenarios incorporating assumptions on controllability and monitorability, and studied the relation between the synchronous temporal logics used in GORE and the asynchronous semantics and logics we have used previously . We are currently studying synthesis of LTS models from GORE specifications to support verification of requirements and also support moving from towards design models. A new line of research has been started in extending our approach to behaviour modelling and analysis to probabilistic systems. We aim to apply our results to the domain of publish/subscribe systems.
Healthcare Applications: We intend to continue using Healthcare as an application focus, particularly for ubiquitous systems and trust and security research. It is a very challenging application area and we have built up good relations with Prof. Yang’s group in the Department working in this area and also with other Departments e.g. in the BiosensorNet project
Autonomous Unmanned Vehicles: this is a new military application area for us which arose out of our involvement in the SEAS DTC. Many of the issues relating to dynamic ad-hoc collaborations, need for self-management, wireless communication and adaptive systems are similar to other ubiquitous computing applications. The environment can be more hostile from a security perspective but power and processing capabilities are less of an issue.
Environmental Monitoring and Collabortion Support: The creative industries amount to 8% of the UK annual income, which is worth over £112.5 billion to the economy accounting for over a million jobs. There is a lack of research into how to measure the performance of the creative workspace and how a given environment affects this function. Conventional methods used in other industries do not work for the more adhoc nature of the creative industries. However live sensor technology has been identified as a potential solution. The technical challenge here is to produce a performance measurement tool-kit that measures quantitative environmental data (such as ambient temperature) while eliciting non-quantitative data (such as sense of wellbeing) and maps these to creative outputs. The sensor network technology must be non-intrusive where necessary or, where required, creatives can harness the ubiquitous technology to help with actual creativity process by sharing ideas, communicating, socialising etc. The number of nodes needed necessitates battery power, wireless communication and autonomic management.