Conferences, workshop & seminars, 2005

Conferences, workshop & seminars, 2005

Seminars
Current year, previous year, next year

2005 Conferences

AINTEC, Asian Internet Engineering Conference, Dec. 13 – 15, 2005, Bangkok, Thailand.
AINTEC is to bring together researchers and engineers interested in the practical and theoretical problems in Internet technologies. It aims at addressing issues pertinent to the Asian region with vast diversities of socio-economic and networking conditions while inviting high quality and recent research results from the global international research community to be presented.
The technical sessions will include invited talks by leading experts, presentations of papers, demos, posters and a pre-conference Asian School on Computer Science (11-12 December, 2005).
In response to the recent natural disaster in Asia, AINTEC'05 main topics concentrate on the survival of the Internet in order to provide alternative means of communications in emergency and chaotic situations, as follows:
- Mobile IP and Network Mobility
- Mobile Ad Hoc and Emergency Networks
- Multimedia or Multi-Services IP-based Networks
- Peer-to-Peer
- Measurement and Performance Analysis
- Internet over Satellite Communication
- IP End-/Site-multihoming
Proceedings will be published by Springer Verlag LNCS.
All are welcome to attend the conference (Dec. 13 – 15) and the School (Dec. 11 – 12).

2005 Seminars

Simultaneous localization and mapping with visual landmarks, by: Dr. Matthew N. Dailey, Sirindhorn International Institute of Technology, Thailand.
Tuesday, Oct. 4, 2005, 11:00, CSIM#209.
SLAM is the problem of a mobile robot building a map of its environment while simultaneously having to determine its location within that map. Although a great deal of success has been achieved with laser range finders as sensors and a planar world assumption, low-cost commercial robots will benefit from robust SLAM using cameras only. This problem, called Visual SLAM or VSLAM, is one of the most fundamental challenges at the intersection of robotics and computer vision. There has been some recent progress in building maps composed of 3D point features, but extracting higher level geometric descriptions of the environment while exploring it is an open problem.
In this talk, I will describe work in progress towards solving the VSLAM problem by building visual landmark maps from lines in 3D space. Tracking observed lines in the environment through time not only aids in localization of the moving robot, but also provides strong clues about the geometric structure of the environment. My approach generalizes successful algorithms from the literature on 2D planar SLAM to the case of 3D worlds and a robot with 6 degrees of freedom in motion. Observations of lines in 3D space are collected with a trinocular stereo camera and fused through time to create stochastic maps describing the estimated structure of the environment and the robot's current uncertainty about that structure. Severe odometric noise is overcome by maintaining multiple stochastic maps
corresponding to multiple hypotheses about the path the robot has taken up to the current point in time. The algorithm shows promise in a realistic virtual reality simulation, and I am currently working to apply it to real-world imagery collected from a real trinocular stereo camera.
Accelerating the relevance vector machine via data partitioning, by Dr. Armin Shmilovici, Ben-Gurion University of the Negev, Israel.
Tuesday, Aug. 16, 2005, 14:00, TC#201.
The Relevance Vector Machine (RVM) is a method for training sparse generalized linear models, and its accuracy is comparably to other machine learning techniques. For a dataset of size N the runtime complexity of the RVM is O(N3) and its space complexity is O(N2) which makes it too expensive for moderately sized problems. We suggest three different algorithms which partition the dataset into manageable chunks. Our experiments on benchmark datasets indicate that the partition algorithms can significantly reduce the complexity of the RVM while retaining the attractive attributes of the original solution.
Algorithms for rectilinear polygon partitioning for workspace division amongst k UAVs, by Mr. Amit Agarwal, PhD Candidate, Nanyang Technological University, Singapore.
Monday, Jun. 20, 2005, 14:00, CSIM#210.
Our goal is to decompose an arbitrary pathwise continuous rectilinear polygon with holes into a fixed number of nonoverlapping pathwise continuous rectilinear open subsets. This goal involves two additional hard constraints - a. The area of each subset must equal an a priori value. b. The union of the closure of the subsets must be equal to the polygon itself. We give a decomposition algorithm that satisfies these constraints. Our goal is motivated by a problem involving parallel utilization of multiple unmanned reconnaissance aerial vehicles (URAVs) with heterogeneous endurance characteristics and equipped with similar sensors with heterogeneous capabilities to conduct a photographic scan of regions with urban-like structures. We discuss our motivation in greater details.
Amit Agarwal obtained a B. Tech. in Industrial Engineering from National Institute of Technology, Jalandhar, India in 1998. He received the prestigious 'The J.N. Tata Foundation for the Higher Education of Indians' Scholarship Award for further studies in 1998. He received an MS in Electrical Engineering from University of Missouri, Rolla, USA, in 2001. During his MS, he worked on problems in quantum logic as they relate to modeling for efficient risk representation. Currently, he is a PhD candidate at Nanyang Technological University, Singapore. His work is partly funded by Singapore Technologies (ST) Engineering and is in collaboration with ST Aerospace. It focuses on algorithmic issues in usage and cooperation planning for unmanned aerial vehicles for efficient region coverage. It frequently draws on knowledge of applied graph theory, applied computational geometry and evolutionary optimization techniques. He is on the program committee of Military & Security Applications of Evolutionary Computation Workshop, GECCO 2005, Washington D.C., USA. He has chaired technical sessions at IEEE/RSJ IROS 2004, Sendai, Japan and CIRAS 2003, Singapore. He has regularly reviewed submissions for IJCNN since 2002.
Localisation and linguistic anomalies, by: Dr George R S Weir, University of Strathclyde, Wales,
Monday, Feb. 14, 2005, 13:00, CSIM#209.
Some interactive software applications aim to accommodate users whose first language is not English. Usually, this entails a focus on translation and related features of localisation. While this motivation is worthy, the results are often less than ideal. This presentation aims to raise awareness of the shortcomings of localisation, in the hope of improving the prospects for successful second-language support. Three varieties of linguistic irregularity in localised systems are described and we suggest that these anomalies are direct results of localisation. This underlines the need for better end-user guidance in managing local language resources and supports our view that complementary local resources may offer a better means of second language user support.