Improving Throughput by Planning in Autonomous Intersection Management
- LecturerDr. Tsz-Chiu Au (Department of Computer Sciences, University of Texas at Austin)
Host: Dr. Tsan-sheng Hsu - Time2010-10-26 (Tue.) 10:30 – 11:30
- LocationAuditorium 106 at new IIS Building
Abstract
Abstract:
Fully autonomous vehicles are technologically feasible with the
current generation of hardware, as demonstrated by recent robot car
competitions such as 2007 DARPA Urban Challenge. This milestone
creates an opportunity to reconsider modern transportation
infrastructure, investigating more efficient systems that leverage
a mostly autonomous vehicle population. Dresner and Stone proposed
a new intersection control mechanism called Autonomous Intersection
Management (AIM) and showed that intersection control can be made more
efficient than traditional control mechanisms, including traffic
signals and stop signs. In this talk, I will present a study of the
relationship between the precision of cars' motion controllers and
the efficiency of the intersection controller. I will then describe
a planning-based motion controller that can increase the efficiency
of the autonomous intersection control mechanism by reducing the
chance that autonomous vehicles stop before intersections. Finally,
I will present some more recent developments in my research group,
Focusing especially on our mixed reality experiment platform on which
a physical vehicle can interact with many simulated vehicles.