University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Guest Talks

title

Cooperative diversity for wireless ad hoc networks

speaker

Dr. Vladimir Stankovic, University of Lancaster
Lancaster, UK

date & place

Thursday, 20.04.2006, 16:15 h
Room C252

abstract

Distributed source coding (DSC) refers to separate compression and joint decompression of distinct correlated sources. Recently, there has been a flurry of activities in both theory and practice of DSC, because of potential applications to wireless sensor networks, Wyner-Ziv video coding, and hyper-spectral imagery. However, evidently, there is still a huge gap between research achievements and the employment of DSC to real-world systems and services. In this talk, two promising applications of DSC will be presented, where DSC is exploited in an unconventional manner, as means of communication rather than compression.

First, it will be shown how to exchange efficiently information in a wireless ad hoc network based on DSC principles. The tackled problem is the deployment of a team of intelligent agents, i.e., robots, in the environments where human presence is impossible; it is ignited by many current and emerging applications ranging from collecting data in inaccessible areas, exploring unknown zones, and robot rescuing missions in disasters, to those closely related to national and homeland security, such as robot surveillance patrols. Deployed in the environment never experienced before, each agent builds in real time its own knowledge model of the surrounding (based on, e.g., captured images and video), and exploits it later to perform required action. Cooperation via exchanging acquired experiences among agents is necessary to enhance and speedup learning, increase survivability, and enable more experienced robots to act as supervisors and coordinators. There are two major challenges of this cooperative learning. The first one is the selection of features that will be extracted from the knowledge model and shared with others. These features must be in such an aggregate form so that a receiving agent can exploit them in real time to augment and adjust its knowledge model. The second challenge lies in the actual information exchange among agents over severely corrupted wireless links, where practical limitations (tiny, low-complexity, low-power transmitters/receivers) impose addition communication difficulties. Another issue has to do with the ad hoc nature of the underlying communication network, where each node (agent) unpredictably moves, joins or leaves the network. The talk will show how DSC can be used to develop decentralized communication techniques that mitigate effects of noise and interference while having in mind practical limitations.

The talk will continue with a seemingly unrelated application of real-time multimedia streaming over heterogeneous (wireless-wireline) networks. The main challenge of wireless multimedia communications lies in the stringent bandwidth and time-delay requirements of real-time multimedia and severe impairments of the wireless channels. A novel system, that will be presented, leverages the knowledge on network information theory, multimedia processing, error control, and networking. Due to its spatial diversity gain realized via DSC, the system promises a significant performance improvement over conventional solutions.