喻园管理论坛2024年第98期(总第1030期)

演讲主题:Autonomous driving: Queueing game, information design and human-AI interactions

主讲人:何翘楚 南方科技大学副教授

主持人:许爱玲 供应链管理与系统工程系讲师

活动时间:2024年11月22日（周五）10：00-11：30

活动地点:管院大楼119室

主讲人简介：

南方科技大学商学院何翘楚副教授，国家特聘专家（青年），在管理科学与工程及其相关领域顶级期刊和会议有50多篇论文已发表或在返修，主持国自然（NSFC）面上、深圳市科创委、香港基础研究基金（GRF）等项目。曾任教于HKUST和UNCC，现为INFORMS会刊Service Science 副编辑。

活动简介：

I am presenting recent research on service operations and business models within a mixed autonomy paradigm, focusing on navigation algorithms influenced by differentiated information and driver-algorithm interactions. Leveraging queueing game models, Bayesian persuasion-based information design, and reinforcement learning for experimental validation, this research examines complex traffic flows, platform operation strategies, the effectiveness of intelligent navigation models, and behavioral patterns in human-machine interaction. These efforts aim to enhance traffic system efficiency and social welfare, thereby advancing the development and application of autonomous driving technologies.

Specifically, I will present findings from five (working) papers. At the micro level, we had a POM paper explores the interaction between autonomous vehicles (AVs) and human-driven vehicles (HVs) using queueing game frameworks. In another POM paper, we proposed smart navigation algorithms through information design methods to mitigate traffic congestion in routing games. A third working paper extends this framework to incorporate targeted algorithms for further theoretical development. Additionally, two ongoing projects employ experimental economics and artificial intelligence (AI) techniques to investigate behavioral interactions between navigation algorithms and human drivers, thereby validating theoretical models of human-AI interaction. These experiments also assess whether AI algorithms can optimize traffic flow via implicit cooperation, even in the absence of centralized control. Furthermore, the studies attempt to explain algorithm aversion behaviors through a dynamic information design framework.