Title: Flow Optimization for Infrastructure-guided Self-driving Vehicles: Coordinated Traffic Management for Intersections, Street Networks, and Highway Ramps
Date: Thursday, August 20th
Time: 1:00pm EDT
Location: George Tower Scheller Tower (TS3) Room 248 and Microsoft Teams
Teams link: https://teams.microsoft.com/meet/239340721314313?p=DAC4WyeP4HfIMRi7Z9
Meeting ID: 239 340 721 314 313
Passcode: 2wy7JZ7L
Candidate:
Xiaochen Shi
Ph.D. Candidate in Operations Research
H. Milton Stewart School of Industrial and Systems Engineering
Georgia Institute of Technology
Committee members:
Dr. Anton Kleywegt, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. Santanu Dey, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. Srinivas Peeta, H. Milton Stewart School of Industrial and Systems Engineering, School of Civil and Environmental Engineering, Georgia Institute of Technology
Dr. Devansh Jalota, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology
Dr. Angshuman Guin, School of Civil and Environmental Engineering, Georgia Institute of Technology
Abstract:
Traffic congestion has long been a major issue in metropolitan cities around the world. The development of connected and autonomous vehicle technologies provides new opportunities to improve traffic efficiency through more precise vehicle-level coordination. With vehicle-to-infrastructure communication, infrastructure can collect detailed information from individual vehicles and coordinate their movements at origins, intersections, and freeway on-ramps. This dissertation studies three traffic control problems under an Infrastructure-Guided Self-Driving (IGSD) environment: autonomous intersection management, flow optimization in urban street networks, and coordinated ramp metering for freeway systems.
In chapter 2, we develop an autonomous intersection management framework for IGSD vehicles at an isolated intersection. An intersection controller fully controls the vehicles approaching the intersection, groups them into non-conflicted batches so that total delay of this re-scheduled crossing sequence is minimized. A dynamic programming-based batching algorithm is developed to solve the scheduling problem with optimality guarantee and polynomial time complexity. Numerical tests demonstrate that the proposed batching algorithm achieves near-optimal solutions with substantially lower computational time than the benchmark MIP model.
Chapter 3 extends the intersection-level batching framework to urban street networks traffic control. We create a layered traffic management framework, including intersection-level batching, origin access control, and simulation-based path determination. The access control method re-schedule vehicles departures at origins, creating a balance between at-departure delay and after-departure delay. The simulation-based path determination algorithm provides alternative route choices for complex networks. Simulation results show that the proposed framework improves network throughput, reduces congestion, and prevents gridlock under high-demand conditions.
In chapter 4, we propose a microscopic coordinated ramp metering framework for multi-ramp freeway systems. The multi-ramp control problem is first formulated as a Markov Decision Process that explicitly tracks every individual vehicle. To overcome the curse of dimensionality, we develop a family of fluid approximation models to relax discrete vehicle movements into continuous flows while preserving key microscopic features. Numerical experiments show that the proposed fluid models perform close to the optimal MDP solution in small networks and outperform benchmark ramp metering strategies by 27% in larger freeway networks.