Distributed Network Resource Discovery in Context-aware Smart Environment

Project Proposal by Yitao Duan and Ling Huang
duan@eecs.berkeley.edu 
hlion@newton.berkeley.edu

A Context-aware Smart Environment enables user to interact with computers, typically embedded in 
the environment, in the way they do with other people, by speech, gesture, movement, affect, and 
context, thus greatly enhancing work efficiency and/or improving our experience with the 
interaction. Research into this area has been gainning popularity in recent years[1]. There are 
many issues involved in building infrastructure Support for such environment that are worth 
investigating from a system point of view such as resource/service discovery, data model, network
protocols, programming language, etc.. 

Many algorithms have been developed for network resource discovery[3], [4]. However, they didn't 
address the particular characteristics and constraints in the smart environment context. We 
identify the following to be unique features in a smart environment system:

1. Network could be hybrid and dynamic

Ad hoc networking has to be supported since users are free to move around, carrying mobile 
devices(PDAs, laptops, cameras, etc.) with them. Spontaneous addition, discovery, and usage of 
services in a network must be supported. At the same time, the system can also leverage stationary 
network facilities such as traditional LAN that is built into its infrastructure.

2. Resource is tight on each node.

Devices in Smart Environment are typically mobile or embedded such as PDAs, wireless sensors, and 
they are limited in terms of memory, persistent storage, processing power, communication bandwidth 
and/or battery life. Therefore efficiency of the alogrithm is extremely important.

3. Many collaborative tasks can be performed with each participant(node) knowing only partial 
information.

Resource discovery often can be abstracted to be each node finding the existance of all other nodes 
in the network[3], [4]. This may not be necessary in our context. In a smart environment where 
interacting entities, be it sensors or computers, are modeled as distributed agents[2], each node 
needs to know only the provider(s) of the services it requests. Reducing the amount and accuracy
of information for efficiency may be a good tradeoff (will pursue later).

In the project, we propose to develope algorithm(s), deterministic or stochastic, suitable for 
resource discovery in networks employed in smart environment systems. We plan to begin by analyzing 
existing algorithms and finding what are relevant and not relevant to our particular needs. Then we 
will try to improve the algorithm by taking into consideration the features and constraints of our 
problem at hand. Implementation or simulation will be done if time permits.


References

[1] MIT Intelligent Room Project, http://www.ai.mit.edu/projects/iroom/index.shtml
[2] M. Coen, “Building Brains for Rooms: Designing Distributed Software Agents,” Proc. Ninth 
    Innovative Applications of AI Conference, AAAI Press, 1997, pp. 971-988
[3] M. Harchol-Balter et al., Resource Discovery in Distributed Networks
[4] Shay Kutten et al., Deterministic Resource Discovery in Distributed Networks