Adaptive Decentralized Traffic Forecasting for Intelligent Transportation

点击翻译按钮获取中文摘要

Autonomous vehicles will be a dominating technology in the future. Driver assistive technologies have evolved substantially over recent years. It will not be long before automated technology in cars replace human drivers. Self-driving vehicles will provide tremendous independence to people with disabilities, non drivers and senior citizens. Statistics Canada's 2016 census reported that for the first time since confederation, seniors outnumbered children in Canada. As the adult population increases, the number of vehicles will also tend to increase. In the future, both self driving and non-self driving vehicles with co-exist. Traffic congestion increases air pollution and travel driving time causing driver frustration, stress and in some cases road rage. ******My research program envisions that traffic congestion will be one of the major challenges for the transportation industry in urban areas and my program will take proactive steps to address this and other transportation issues through Intelligent Transportation Systems (ITS). ITS will exploit state-of-the art technology including motion sensors, increased available computation power and innovative communication protocols to inform drivers on the move about traffic and road conditions. ITS will include road network (static), vehicle-to-vehicle technology in which vehicles communicate with nearby vehicles using wireless connectivity and vehicle-to-infrastructure technology where vehicles communicate with networked road-side devices. This type of communication creates a highly decentralized, dynamic, ad hoc, mobile network called a vehicular ad hoc network (VANET). *** ***A major objective will be to develop a traffic aware routing algorithm that will consider both the road network and dynamic VANET data. Assume you are driving on a street and there is an accident somewhere towards the end of the street. The speed of the vehicles on the street will decrease. This, of course, affects other vehicles on the street and creates a traffic jam. The roads surrounding the congested street will obviously also be affected. What is interesting and not yet considered in predicting traffic is that over a period of time, other roads not necessarily close in proximity to the congested road may also get affected. I proposed to consider all the roads that are influenced by the congested road and cluster them together? We can train these clusters using machine learning techniques to find the patterns and relationship between them using real time data provided by road-side devices. The trained clusters can then be used to predict traffic any time evolving to reflect current conditions. The clusters will be made available to VANET vehicles to predict traffic and re-route if necessary. ******My research program will train HQP in cores areas of computer science and engineering that will prepare them for in-demand careers in science. In this cycle of my Discovery Grant, I expect to train 2 UG, 3 MSc and 3 PhD. ***********