Bikeability and bike-way network design problem

Abstract

To achieve the car-lite goal, cycling is increasingly gaining its popularity in recent years as a sustainable and attractive alternative mode of transportation to automobiles. Although bicycle is energy efficient and environmentally friendly, many trips within cycling distance of a commuter’s home are accessed via automobiles. Bikeability refers to the extent to which it is conducive and comfort for bicycling. To promote cycling as an alternative mode of transportation to automobiles and have a better understanding of cycling, this thesis aims to investigate the bikeability and bike-way network design problem systematically. In this regard, this thesis contains three sets of models or methods to address the research on bicycle. For bikeability analysis, to address the cycling comfort level of bicyclists, Chapter 2 derives a Cycling Comfort Index (CCI) as a measure for various types of cycling facilities. For data collection, an Instrumented Probe Bicycle (IPB), equipped with a video camera and a set of sensors is ridden by bicyclist in Singapore. Convolutional neural network (CNN) algorithm is employed for automatic video processing. The extreme gradient boosting (XGBoost) algorithm is applied to calculate CCI based on cycle track characteristics and survey participants’ comfort perception ratings. Understanding bikeability can provide valuable information to bike-way network design planning in real-world. For the bike-way network design problem, the first set of methods in Chapter 3 formulate a multi-objective integer linear programming model to determine the spatial layout of bike-way networks and bike-way link types under space-time accessibility constraint. Four objectives include to maximise the accessibility, minimise the number of intersections, maximise bicycle level of service (BLOS), and minimise total construction cost, with monetary budget and space-time constraint are considered. Augmented epsilon-constraint method (AUGMECON) is applied as the solution algorithm, and two numerical examples are used to generate non-dominated bike-way network design plans. Then, Chapter 4 addresses the road diet network design problem in consideration of a multi-modal transportation network with bike-and-ride, automobile and bicycle, under stochastic demand scenarios. A mixed integer non-linear bi-level optimisation model is formulated. A path-flow penalty-based branch-and-bound algorithm is applied to derive the optimal road diet plan for upper level, and the method of successive averages is utilised for multi-modal traffic assignment for lower level. Numerical test is utilised to demonstrate the feasibility of model and effectiveness of solution algorithm.