Designing and evaluating bicycle networks from a sustainable urban mobility perspective

The promotion of sustainable transportation modes, such as cycling, has become a crucial aspect of urban planning and design. Designing an efficient bicycle network is a complex optimization problem that requires the analysis of various factors, including infrastructure layout, connectivity, safety, and accessibility. Moreover, the fastest routes are not always the preferred options for cyclists, who may consider multiple criteria when choosing routes from their origins to their destinations. Therefore, this research aims to propose an optimization strategy for evaluating and designing bicycle networks, increasing, consequently, bicycle use in the cities. The proposed strategy is validated through experiments using real-world data collected from the bicycle network in the city of Parma, Italy. The problem addressed in this research can be divided into the following two parts. First, we conduct an analysis to identify cyclists' preferences regarding a set of road characteristics, namely route length, safety, and practicability. Based on these preferences, cyclists can be categorized into different profiles, each associated with specific weights they assign to these characteristics when selecting routes. To this end, we propose two mathematical formulations and corresponding algorithms, depending on the type of data used in the identification process. One formulation is applied when cyclist flow data are known. These flows can be collected, for instance, from cameras across the city. The other formulation is used when cyclists' paths for a set of origin and destination pairs are available. This type of information can be obtained, for example, through bike-sharing services. The proposed formulations and algorithms were validated using both randomly generated data and real-world data from Parma. Next, given a set of possible interventions in the bicycle network (e.g., constructing new bicycle lanes, improving the quality of existing ones, etc.) and the cyclist profiles identified in the first part of the problem, we select the best combination of interventions that minimizes the overall cost perceived by the cyclists traveling between different origins and destinations. The selection of interventions is subject to demand fulfillment and budget constraints. To address this problem, we propose a mixed-integer linear programming formulation and a set of optimization methods, including a branch-and-bound algorithm, two heuristics based on solving knapsack problems using dynamic programming, and an enumeration mechanism. Computational experiments were conducted using both randomly generated and real-world instances, the latter based on data collected from Parma. The city was represented as a graph comprising more than 40,000 nodes and 95,000 arcs. An intervention set of 10 intervention types was defined, with each type including multiple interventions affecting different regions of the city. The results demonstrated that the proposed approaches efficiently solve the studied problem and provide valuable managerial insights. Finally, as an ongoing work, we aim to develop a strategy to analyze the impact of the proposed interventions on motorized vehicle users. Given the previously identified cyclist profiles in Parma and the set of possible interventions to the city’s cycling network, the goal is to determine the most effective interventions to implement, considering not only the overall cost perceived by cyclists but also the effects on motorized traffic.