SE PhD Final Defense of Salomón Wollenstein-Betech

TITLE: Operations Management of Intelligent Transportation Systems

ABSTRACT: Congestion is a central problem in today’s transportation systems, particularly in densely populated areas. While congestion has been mitigated by public transport in the past, in recent years mobility solutions, e.g., bike sharing, e-scooters, Autonomous Mobility-on-Demand (AMoD), have developed rapidly to improve human mobility. This prospectus develops methods to design, plan, and evaluate transportation networks, as well as to enhance the operations management of multimodal AMoD systems.

A starting point of the analysis is the Traffic Assignment Problem (TAP), which serves as a core framework. First, a non-parametric method to jointly estimate the TAP inputs, i.e., the Origin-Destination demand and travel latency cost function, from traffic flow data is presented. Then, a novel reformulation of the so-called TAP with side constraints (TAP-SC), which may impose arbitrary constraints on the flows, is developed by introducing affine approximations of the travel latency function. It is shown that these reformulations are asymptotically optimal in the number of affine segments used. Numerical experiments show drastic reduction in computational times with accurate solutions.

The TAP-SC formulation is applied to optimizing the operations of AMoDs by jointly tackling the routing and load-balancing problem while considering reactive exogenous flow from private vehicles. In addition, this work jointly optimizes pricing and load-balancing policies of AMoDs in order to maximize a utility objective for the AMoD platform. This joint optimization, which employs user destination information, could improve current approaches by 7% to 40% based on case studies using New York City and Chicago taxi records data.

Ongoing work explores decomposition techniques for the TAP-SC, as well as modifications to solve the system-optimal (SO) and user-equilibrium (UE) problems. Moreover, the use of more realistic demand functions and bid-prices while designing real-time dynamic pricing schemes for AMoDs is a direction to be explored.

COMMITTEE: Co-Advisors:Christos Cassandras and Yannis Paschalidis, SE, ECE; David Castañón, SE, ECE; Mauro Salazar, ME, Eindhoven University of Technology

When 12:00 pm on Monday, November 30, 2020
Location Zoom