System Support for Ubiquitous Computing Workshop
At the 8th Annual Conference on
Ubiquitous Computing (Ubicomp
2006)
In Orange
UbiSys 2006 Workshop Report
Ubisys was extremely well attended this year with 47 registered participants from Europe, the USA and Asia. There was a good mix of academia and industry ensuring balanced discussion.
The day was split up into 2 paper sessions and 3 panel sessions. For the paper sessions 6 papers had been chosen from 27 submissions and focused around 4 key themes.
Introduction and One Minute Madness
The day kicked off with a 1 minute madness session to allow all participants who wanted to the opportunity to briefly introduce themselves and their research interests. From there we moved straight into the first panel session exploring barriers to adoption and evaluation techniques.
Barriers to Adoption and Evaluation: Quantifying Magic in Ubicomp
Led by Roy Campbell of UIUC the barriers and evaluation panel was designed to explore issues that were creating significant barriers to deployment of UbiComp systems infrastructure and the related issue of how to evaluate ubicomp deployments. John Barton of IBM led with a though provoking position presentation on the gap between the assumptions made by Ubisys researchers and the reality of the current technology and business environment into which they hope to deploy. He argued that we need to do a better job of quantifying the parts of the systems that use assumptions about technology, business or social trends so that other researchers can evaluate the research. He described these assumptions as the 'magic' in the system.
Doreen Chen of Samsung followed with a discussion of the barriers to deployment of ubicomp technologies into the home. Her focus was on the difficulty of developing systems that were capable of dealing with the failures inevitable in the modern networked home.
The final panelist was Jonathon Engelsma of Motorola who talked about the challenges faced by systems developers when developing systems for mobile devices. He highlighted the tension between the four domains of device portability, network challenges, channel challenges and usability.
In the ensuing discussion the idea of developing a canonical set of use cases or scenarios was discussed as a way to allow better comparison of ubicomp systems. Equally there was resonance for the idea that the community has to develop more resilient systems infrastructure and perhaps a common or shared open platform (or open source platform).
Paper Session 1: Use of common abstractions
After the break, the first paper session (http://www.magic.ubc.ca/ubisys/program.html) focused on common abstractions for ubicomp application development.
Stefan Egglestone of Nottingham University discussed how the Equip Component Toolkit developed within the UK funded Equator project was used to foster collaboration between artists, designers and developers during the lifecycle of a ubicomp installation. A key theme in his talk was the need for system researchers to engage and support end users if UbiComp systems are to be adopted. Mike Blackstock of U of British Columbia then presented a proposed common model for ubicomp environments based on an extensive survey of existing UbiComp systems. His proposed set of abstracts and an initial supporting middleware could be used to support the development of a stand alone Ubicomp application. However its main goal was to provide a meta-middleware that allowed application developers to bridge between existing ubicomp middleware platforms. Finally Mattias Kranz presented how the Player/Stage middleware, originally designed to support robotic applications can be used for ubicomp applications due to its rich support for sensors.
Lunch and Poster Session
Lunch was combined with a poster session ensuring that participants could continue the morning discussions as well as engage over the 10 or so posters on display. Posters are available at: http://www.magic.ubc.ca/ubisys/position.html
Panel 2: Middleware directions and Challenges
After lunch, the second panel was chaired by Nitya Narsimhan of Motorola and addressed middleware and platform issues. Nitya's panelists were
During this panel there was discussion of the required support for transitions between home, car, office and mobile (city). Platforms for the home, car, office and city have commonalities but significant differences such as the devices, networks used and associated constraints. Guenter highlighted some of the issues for service composition within ad-hoc and constrained networks and then Marcus explored spatial context information for car driver assistance. Trevor focused on an abstraction layer that supported multiple mobile devices and allowed users to easily enter a new environment and connect and interact with the environment. With a particular focus on the home, the abstraction layer was designed to hide the complexities of device connection and service discovery. Finally Michael wrapped up with a thought provoking presentation on the possibilities of using Ad-hoc communications to support cached content for super distribution on mobile phones.
In the discussion it was highlighted that there are significant issues which are not technical, eg DRM, business models, content delivery channels are affecting our ability to deploy. Of particular note is that the main domains, home, car, work and outside all have different constraints and so hamper common approaches to systems support for ubicomp applications.
Paper Session 2: Emerging Software Engineering Approaches
The 2nd paper session addressed emerging software engineering approaches to support mobility, programming networks of sensors, and situated display based interactions. Aline Senart presented how context reasoning can be used to address the unreliable communications between components on ad hoc networks. Urs Bischoff presented a high level language can be used to specify the global behaviour of a sensor network which is then compiled and downloaded to individual sensors for execution. Finally Rui Jose presented a proposed system for supporting applications hosted by situated displays interacting with mobile devices and sensors nearby.
Panel 3 - Challenges
The final panel of the day looked at the main challenges faced by the community and what could be done to address some of these challenges. Led by Rodger Lea of UBC the panelists included:
This panel worked to bring to light some key challenges for the ubisys community and summarize the significant issues and themes raised throughout the workshop. Specific challenges for ubicomp systems such as reliability, power issues, privacy and security were highlighted. The multi-disciplinary nature of the community and the difficulties involved in working across those disciplines was noted. Challenging and useful ubicomp applications were presented such as environment monitoring, energy conservation and "social computing". Adrian focused on the need to open up about the hard lessons learned in deployment of systems, even our mistakes, and focus on making it possible to build trivial applications in a trivial way (ubicomp in a box). This can be done by “eating your own dog food”. Roy and Max argued that as a community we don't agree on what a ubicomp system is, what it should do (common set of scenarios), and on some common approaches including high level abstractions and design patterns for building systems. We need to work on agreement around these issues. Alan focused on the challenges the community faced and argued that the most practical way to solve these was to agree on specific problems and develop real solution, for example solutions that could be shown and used at UbiComp 2007.
Wrap up
While there were a number of interesting and significant issues raised, the following themes appeared in a number of presentations and during discussions throughout the day.
Evaluation - it is clear that a key problem for the UbiSys community is the difficulty in evaluating systems and in particular in comparing systems. Several aspects were discussed including a proposal to better quantify the hidden assumptions in a system, the possibility of developing a common set of scenarios for easier comparison as well as a more formal method of classifying ubicomp systems.
Common abstraction, toolkits or platform. Significant discussion explored whether the community would benefit from a common approach to a platform or toolkit. Some felt that agreeing on a common set of abstractions would help the community by ensuring systems could interact. Toolkits were discussed in several places and highlighted as a lightweight way to provide a common platform, equally there was discussion of a common platform - possibly open source - that the community could benefit from. While there was no consensus reached on this issue there was a strong interest in rallying the community to make steps towards a common set of approaches, technologies and platform(s).
Resiliance/reliance and testing. A clear problem for ubiComp deployment is the fragility of the systems we deploy. In discussions the workshop touched on resiliant software techniques, distributed debugging and simulation. Participants felt that was a key area for further study.
Eat your own lunch. A key theme that appeared in a number of presentations and at the final panel was the need for the community to develop systems that were then used both for ongoing research and for other groups's research. It was identified that the community is missing an opportunity to progress more rapidly because of the amount of rework and reinvention that is occurring. Obviously this point is related to the issue of common toolkits and platforms.
Several participants argued strongly for the need to make trivial applications trivial to develop and deploy: Their call for ‘Ubicomp in a box’ resonated within the workshop although the issues of how to develop an agreed solution are still complex.
Grand challenges. A reoccurring theme during the day was the issue of what was driving our UbiComp systems development - whether it was a set of common scenarios, or a set of well defined grand challenges, the systems community needs a well defined set of targets. Areas discussed were health, environment, safety and security as key grand challenges.
Next steps
Significant interest has been generated into the possibilities of the ubisys community developing a consensus on high level definitions, requirements, abstractions, and components needed in a ubicomp platform. This would be the first step towards agreeing on particular technology implementations and/or an agreed set of standard approaches or APIs for UbiSys researchers. It is hoped that initial steps towards a better definition of these core components will begin within the community in the lead up to 2007 UbiSys.
Preparation for UbiSys 2007 is already underway, if you want to become involved contact us at ubisys@magic.ubc.ca
Last updated: November 15, 2006