Simulation, Learning, and Control for Urban Traffic Systems

Despite automobiles' contribution to socio-economic development, ever-increasing traffic congestion and accidents have resulted in over a trillion dollar costs annually worldwide. With the introduction of autonomous and connected vehicles, we face opportunities to improve the urban traffic systems. However, studying traffic in its mixture form is challenging due to not only limited data but also its multi-scale nature. In this talk, I will discuss my efforts at various traffic scales: macroscopic--how to efficiently reconstruct city-scale traffic; mesoscopic--how to control a mixture of autonomous and human-driven vehicles to achieve high throughput, and microscopic--how to use principled simulations to navigate an autonomous vehicle through environments with obstacles.