*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Applying Distributed Online Learning in Vehicular Network
Committee:
Dr. Blough, Advisor
Dr. Y. Chang, Chair
Dr. Sivakumar
Abstract: The objective of the proposed research is to realize distributed online active learning in vehicular networks. With the advancement in Machine Learning and Artificial Intelligence (AI), Autonomous Vehicles (AVs) began to migrate from laboratory development and testing conditions to driving on public roads. Their deployment in our environmental landscape offers a potential for decreases in road accidents and traffic congestion, as well as improved mobility in overcrowded cities. Although common driving scenarios can be relatively simply solved with classical perception, path planning and motion control methods, the remaining unsolved scenarios are corner cases in which traditional methods fail. These unsolved cases are the keys to deploying AVs on road, requiring an enormous amount of data collection and high quality human annotation. However, this is very cost-ineffective considering the ever-changing real world scenarios and various road/weather conditions. To address the above challenges, we propose a new collaborative distributed online active leaning framework supporting connected vehicles and future AVs. Two pieces of preliminary research have been conducted aiming to 1) secure critical information that is disseminated among nearby vehicles through distributed consensus; and 2) perform task-oriented group formation (ToG), where groups of cooperative vehicles are tailored for specific computation tasks to be executed. Building upon preliminary research, we will further address the following sub-problems: first, develop a new learning method that allows a group of vehicles to cooperatively and distributedly learn with local data to improve model accuracy; second, extend the learning method to work with ToG to realize online learning in vehicular networks. Finally, an integrated simulation environment, synthesizing existing tools and additional distributed learning modules will be implemented and a comprehensive simulation study will be conducted to evaluate performance and demonstrate the online active learning framework.
