![](https://csdnimg.cn/release/download_crawler_static/15177631/bg3.jpg)
A distributed middleware for self-configurable wireless sensor networks 3
services are designed to simplify and facilitate the
application developments and to help the network adapt to
the dynamics with the consistency to existing network
programming paradigms.
As proposed in Intanagonwiwat et al. (2000) and
Intanagonwiwat et al. (2003), the directed diffusion is a
widely used distributed routing protocol for WSN. It makes
routing decisions according to the data requested. One of the
most significant issues of the directed diffusion protocol is
the packet flooding. The region filter developed in Ivester
and Lim (2006) is an improvement to the directed diffusion,
since it maintains message re-transmission mechanisms
within a specific region and filters out some traffic in the
networks.
The other important issue of directed diffusion is node
failure and mobility treatment. There are some mechanisms
responding to topology changes that are built in the directed
diffusion. When data stop arriving at one node due to its
upstream node going down or moving away, directed
diffusion will trigger the event of flooding similar interest
packets in the entire network. This basic form of
adaptability is adequate for simple query-response systems,
but not sufficient for complex applications due to long delay
required to find a replacement, network flooding, and the
undetermined physical location of the substitution.
Additionally, the task of reconfiguring the upstream nodes
will introduce extra overhead when adapt to directed
diffusion changes at the upper layer. This can not be
effectively achieved.
This research involves two previous researches. The
distributed services middleware layer (Lim, 2001, 2002),
which proposed only the main idea of distributed services,
but not implementation details of each service. The directed
diffusion routing protocol (Intanagonwiwat et al., 2000,
2003) is the other work involved. The following goals are to
be achieved in the research:
• to design and implement application programming
interface (APIs) for the distributed composition service,
which is the most important component of the
distributed services
• to maintain the normal networking functionalities under
abnormal networking conditions with the help of
distributed services
• to achieve the design principles of WSN middleware
(Radhika and Malarvizhi, 2012), data-centric, dynamic
adaptive, user-transparent, scalability, and energy-
efficient.
In general, the major contributions of this research work can
be summarised as follows: firstly, abstracting lower-level
functionalities, providing a coordination interface to the
applications, and making the heterogeneous WSN
development simple and efficient; secondly, designing and
implementing the API based on the unique features of
WSN, such as resource constraints, complex link quality,
dynamic topology and scalability, tackling these special
issues of WSN. These solutions are evaluated by the ISEE,
a self-developed testing environment and they are
outperformed compared with directed diffusion. Finally, this
research is the preparation for the target tracking application
under the real-world scenarios.
The remaining of the paper is organised as follows:
Section 2 presents some technologies related to the
distributed composition service. Section 3 discusses the
architecture of the composition service with its APIs.
The real implementation is demonstrated in Section 4.
The performance is evaluated in Section 5. Section 6 draws
the conclusion by summarising the research work.
2 Related work
2.1 Directed diffusion
As proposed in Intanagonwiwat et al. (2000, 2003), the
directed diffusion is a distributed routing protocol with
routing decisions made according to the data requested. The
directed diffusion has some creative features including
data-centric dissemination, reinforcement-based adaptation
to the empirically best path, in-network data aggregation
and caching.
Directed diffusion uses both positive
reinforcement and negative reinforcement, which sends data
to the lower delay neighbouring node with a higher data rate
and to the higher delay neighbouring nodes with lower data
rate. Through this method, directed diffusion inhibits nodes
on the long delay path from packet forwarding. The directed
diffusion offers accessibility to the diffusion core
behaviours through publish and subscribe APIs. Though
some mechanisms reacting to the dynamic topology changes
are built in the directed diffusion, there is still a long
recovery delay and network traffic generated at the time.
Thus, there is much work to be done to improve its
performance regarding to the time-sensitive and resource-
constrained WSN applications.
2.2 Middleware for WSN
Software development for WSN requires novel
programming paradigms and technologies. Conventional
principles of communication are mostly inapplicable due to
dynamic topology changes and the need for cooperative
task processing in WSN (Hadim and Mohamed, 2006a,
2006b; Radhika and Malarvizhi, 2012). As addressed
by Laukkarinen et al. (2011, 2012) and Mohamed and
Al-Jaroodi (2011), introducing distributed middleware to
the WSN application development has become a
breakthrough in the WSN development, which brings
adaptability, simplicity, and efficiency to this field. In the
WSN development, the distributed middleware refers to two
aspects: WSN application on one hand and the distributed
application which interacts over the network on the other
hand. Primary objective of the middleware layer is to hide
the complexity of the network environment by isolating the
application from protocol handling, memory management,
network functionality, and parallelism. The benefits of
adopting middleware to the WSN implementations can be
summarised as follows: