Back

Speaker "Joe Xing" Details Back

 

Topic

A Data-driven Human Behavior Learning System to Coach Self-driving Cars

Abstract

While both the heterogeneity and complexity of the hardware and software system consisting of autonomous cars is increasing significantly, a robot car is presented withtremendous amount of combinations of possibilities in decision-making processeswhen it comes to the problem of how to drive in a world where human drivers are coexisting with robots. The process oftesting and validation of such huge amount of possibilities, thus enabling naturalistic driving behaviors of robot cars,can be cumbersome and alleviated by utilizing simulated environment. Both drivercentricmicroscopic and system-oriented macroscopic simulation, with human correlations taken into account, are essential when it comes down to the fidelity level of thesimulation process therefore the trained prediction power. AI technologies can be leveraged toaddress fundamental issues of system modeling of driver’s intention and behaviors, trafficadaptive capabilities, etc.
 
The speaker shares a methodology that has been be used to build a training and learning system for robot cars. Data-driven approaches are exploited to learn real human roadbehaviors, thus model and produce simulated traffic scenes, trajectories and intentions ofroad users, etc. System level architectures are also described on how to take into accountthe correlations among the road users in macroscopic traffic flow simulations,etc. Scientific experimental results on how the individual road user’sbehavior could affect the traffic adaptive capabilities will be presented at the conference.
 
Keywords:
Driving Behavior, Traffic Simulation, Artificial Intelligence
 

Profile

Dr. Xing is the CEO of ROMAN ROADS, INC., which is a startup company based at Silicon Valley. The company is focusing on building disruptive technology of human behavior learning system dedicated for self-driving cars. Dr. Xing joined Stanford in 2013 as Engineering Physicist / Staff Scientist, where he worked at the national laboratory SLAC with job responsibility being developing and managing an in-house software system of data analysis, machine learning, sensor / detector read-out, big data framework etc. Dr. Xing joined NIO, USA in 2016 as Staff Data Scientist where he was focusing on developing state-of-the-art deep learning, reinforcement learning algorithms for autonomous driving vehicles. Dr. Xing is an innovative entrepreneur who founded his own startup company Athena Robotics in 2017, where he was the CTO in charge of full stack of self-driving car software technologies including perception, prediction, simulation and planning. Athena Robotics, Inc was selected by Y Combinator for the summer batch of 2018. Dr. Xing joined Borgward, BAIC Group in 2018 as the Director of Artificial Intelligence for Autonomous Driving, who is in charge of the R&D on artificial intelligence for self-driving car technologies. Dr. Xing is an exceptional scientist and AI researcher with outstanding publication records of more than 15,000 Google Scholar citations on top-notch science, AI journals and conference. Dr. Xing was also invited speaker on applications of deep neural network. Moreover, Dr. Xing is an experimentalist who always believe in data and experimental results.