Esprit "IT for Mobility" 26900
|
Esprit "IT for Mobility" 26900 |
|
|
|
|
|
|
|
|
||||
| TEA's
technical objective is to develop an awareness-enabling add-on component
for mainstream mobile computing and communication devices, such as PDAs,
laptops, and GSMs. The word 'awareness' here refers to the permanent availability
of information about the minute-to-minute context in which a device is
being used.
Awareness data can cover where and when, but also - and more interestingly - information on place (e.g. office, home, park), its environmental features (e.g. darkness, crowdedness), and on the owner's activity (e.g. writing, sleeping, walking). The TEA add-on is responsible for the continuous dynamic profiling of the 'user-place-activity' (or context). Such information can be used for controlling incoming streams, annotating outgoing streams, or for setting device controls. |
For
example, a GSM may know whether or not to ring urgently or buzz subtly,
a PDA may know when to initiate a network connection, or to wait until
a connection is cheaper and more reliable, and a laptop can know when to
switch to low-power mode because its user is engaged in a phone conversation
across the room. The approach toward the technical objective of TEA, the
dynamic profiling of user-place-activity, is not feasible by computationally
intensive image analysis, nor is the use of high-precision positioning
technology, for example, compatible with our business objectives. Instead,
the TEA component relies on multiple primitive cues, extracted from an
open collection of low-cost sub-semantic sensors.
Relevant sensors that we will explore give visual, auditory, environmental, temporal, motion, and location cues. The idea is much similar to a person saying 'it's getting late', a statement about time, while it is actually based on, for example, it getting dark outside (a feature of the environment) and herself getting tired (a feature of the person's metabolism). For what it is worth, the analogy illustrates how combinations of low-level clues are indicative of semantic context characterizations. A second highlight is to take an adaptive approach for mapping (in fact, in two steps) from a high-dimensional sensor space to a small set of symbolically described context-profiles. Adaptive techniques, in casu Kohonen-type mappings and k-nearest neighbor algorithms, allow dealing with new situations and sensor data, adaptive to the habits and surroundings of the owner. Finally, the project will pursue a dedicated hardware implementation- possibly in so-called 'evolvable hardware', allowing for on-line adaptation and high efficiency- allowing for easy integration with portable computing devices. |
|||
| Business Objectives | ||||
| TEA's
business objective is to exploit a market opportunity for low-cost context-awareness
devices. The TEA consortium, on the other hand, aims at a low-cost component
- say less than 100 ECU retail price - that can be used on a variety of
mainstream devices. It will not require specific server-side functions
or changes to telecom infrastructure. We do envisage, however, dedicated
services exploiting the added value of an awareness-enabled device. For
instance, a city tourist service can, at each moment, serve the most relevant
information in the most adapted form to an awareness-enabled portable city
guide. A second business objective is thus to start such value-added services
for awareness-enabled devices.
In approach to the business objectives, existing user communities and applications will be used that are accessible either through one of the partners, any of the other IT for Mobility projects, or the various EC's concertation and support actions (e.g. I3-Net). The project is highly product-driven and drives to prepare for a short time-to-market, 5 years at most, but shorter for specific user communities. Exploitation issues such as market awareness, infrastructure constraints, usability and acceptance, will be taken into account from the start. |
The
project will use technical evaluation criteria that reflect the context
awareness of the TEA component: its sensitivity (wanted and unwanted),
accuracy, and latency characteristics will be established and 'life-long'
convergence properties will be evaluated.
From the user's point of view TEA will concentrate on evaluating in concrete scenarios the adequacy of 'information control' that is achieved, both in terms of relevance and form. |
|||
|
|
||||
| To summarize, the objectives of the TEA project are to: | ||||
|
|
|||
| last update: 28/06/2000 by Kristof Van Laerhoven | the previous sites can still be found here and here |
