![](bg2.png)
2 DOMAIN DESCRIPTION AND
RELATED RESEARCH
2.1 Domain Description: A Case Study
of U-hospital
Figure 1 shows opportunistic location-based services
under dynamic changes of service area in a hospital
environment. In this environment, the following
assumptions have been made. First, the whole
environment can be divided into small spaces like
rooms or floors. Such a space is called a unit space
and is the basic unit for location awareness. Second,
communication devices are divided into stationary
nodes and mobile nodes according to their
functionality and features. Mobile nodes can be
attached to people or medical devices in the form of
small tags. A stationary node can be installed in the
ceiling or walls in every unit space to provide
location reference functionality and a
communication access point for mobile nodes.
In this environment, a location-based distributed
architecture is more suitable than a traditional
centralized one. A centralized architecture has
difficulty in supporting services that require real-
time response with a large number of sensor nodes.
The reason for this is that messages generated in the
services need to be delivered to the server using
multi-hop communication, and therefore bottlenecks
arise on the paths to the server.
In this environment, requests of particular
services should be processed by the agent which is
in the same location with the service customers
(mobile nodes). In figure 1, the tracking service for
the wheelchair (mobile node m9) is processed, not
by the central server, but by the agents in the S9 and
S3 stationary nodes. Due to the location-based
distributed agent architecture (based on the location
of each mobile node), fast response can be provided
despite the large number of mobile nodes. To
implement a location-based distributed agent
architecture, one possible solution is to have all
service agents statically running in all stationary
nodes, but this approach is too inefficient to
implement. To solve this problem, a location-based
reconfiguration architecture was used for the service
agents. As shown in figure 1, if necessary, any
stationary node can activate any service agent and
expand its service area by creating a clone agent
(meaning a new instance of the same service agent)
to handle location-based service request from mobile
nodes. By this, all mobile nodes can freely move
anywhere without restriction. Of course, if the
mobile nodes’ locations are reduced, then the service
area also shrinks.
Figure 1: Dynamic changes of service area in a U-hospital.
2.2 Related Research
Recently, opportunistic computing or opportunistic
networking has become an important concept in the
service computing area, driven by the rapid growth
of mobile computing and ad hoc networks. The main
concept of opportunistic computing, that “when two
devices come into contact, it provides a great
opportunity to match services to resources, exchange
information, cyberforage, execute tasks remotely,
and forward messages” (Conti and Kumar, 2010), is
highly suitable for location-based services and social
network service applications because
communication and computing are processed by
means of social relationship and collaboration
among communication nodes. In this paper, based
on the concepts just described, a middleware
architecture is proposed which can create
communication opportunities and provide exchange
services between stationary nodes and mobile nodes
which approach each other.
There are many researches related agent based
middleware in WSN environment. TeenyLIME
(Costa et al., 2006) is a tuplespace-based application
middleware which is designed for a WSN
environment without base station. TeenyLIME is an
extension of LIME. In this paper, a new tuplespace
implementation which can provide event-driven
asynchronous read/write on top of TinyOS has been
provided. Agilla (Fok et al., 2009) is a mobile-agent-
based service middleware for a WSN environment.
In this middleware, an agent can be moved or copied
among mobile nodes without losing its internal
contexts. Communications among agents is
accomplished by the abstract functions of a
tuplespace and an internal neighbor list. However,
these studies are also based on multi-hop
communications and need to deal with complex
calculations on the mobile-node side, and therefore
it is difficult to adapt their proposals to provide the
services needed for real-time response with a large
PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems
250