21 March 2018
Time: 3:00 - 4:00pm
The Antennas and Electromagentics Group is pleased to announce the following guest speaker seminar:
Speaker: Dr. Wei Dai, Imperial College London
Title: Origin-Destination Flow Estimation Without Traffic Assignment Prior Information
Time: Wednesday 21st March, 3-4pm
Venue: Engineering 3.24
Abstract: This work focuses on the inverse problem of estimating origin-destination (OD) flows from link flows. OD flow estimation is essential to many network analysis tasks as OD information specifies the travel demands in the network. Approaches in the literature for OD estimation typically require some prior information of the OD flows as well as the mapping from OD flows to link flows. In particular, a linear model is often used to map OD flows to link flows and the corresponding linear operator is commonly referred to as traffic assignment. In practice, prior information of the traffic assignment can be inaccurate and incomplete. Furthermore, even with full information of the traffic assignment, the OD flow estimation problem is still ill-posed and therefore further prior information of OD flows is required. The main contribution of this work is the new capability of estimating OD flows with no or little prior information of traffic assignment and OD flows. Instead of directly working on the linear system involving the OD flows, a new linear model is introduced of which the inquired flows are specified with their origins but not their destinations. The new model preserves all the OD flow information and at the same time reduces the dimension of the inquired flows in orders of magnitude. Due to this magnificent dimension reduction, the OD flow estimation problem becomes well-posed when the corresponding traffic assignment information is available. More importantly it is possible to estimate the OD flows and the traffic assignment simultaneously with no prior information of them. Detailed discussions are provided regarding the uniqueness of solutions for the joint estimation. Simulations are presented to demonstrate the new capability.
Speaker biography: Dr. Wei Dai is currently a Senior Lecturer in Electrical and Electronic Engineering at Imperial College London. He received his Ph.D. degree from the University of Colorado at Boulder in 2007, and was a Postdoctoral Research Associate at the University of Illinois at Urbana-Champaign from 2007 to 2011. The main theme of his interdisciplinary research interests is sparse signal processing including compressed sensing, super-resolution, and bilinear inverse problems. His other research interests include smart city, biomedical imaging, wireless communications, and information theory. On theoretical front, his subspace pursuit algorithm for compressed sensing has been cited more than 1500 times. On the more practical side, he was involved in the development of the first compressed sensing DNA microarray prototype in the world, and led the first hardware implementation of compressed sampling in the UK.