Predictive modeling of dosimetric quality assurance in radiation therapy

Abstract

As the Canadian population ages, there is an increasing demand for senior transportation services that cannot be met by regular public transportation systems. This thesis aims to bring attention to the combinatorial optimization problem of senior transportation. By extending current vehicle routing problem knowledge and computational technologies, we construct novel problem definitions for the Senior Transportation Problem (STP) and the Senior Transportation Problem with Overbooking (STPOB) based on data provided by a partnering non-profit organization specializing in senior transportation services. Solution approaches including mixed integer programming, constraint programming, two logic-based Benders decompositions and a construction heuristic are developed and empirical analyses on both randomly generated datasets and real-life datasets are performed. Constraint programming demonstrates best results being able to solve to optimality large real-life instances of up to 270 vehicles with 385 requests for the STP and up to 210 vehicles with 320 requests for the STPOB under 600 seconds.

Type

Related