Optimal designs of heterogeneous grid transit networks

Abstract

A general Continuum Approximation (CA) model is proposed for optimizing transit network designs (TND) in grid cities under spatially heterogeneous demand. While conventional studies often assume rigid geometric line configurations (e.g., unbranched orthogonal grids), our framework allows the grid bus lines to route more flexibly by making lateral movements and to form network configurations with line detouring, merging, and diverging. The resulting line and stop densities, as well as service headways, vary continuously across both directions of the city, constrained solely by vehicle flow conservation. By respecting non-uniform demand distributions, our heterogeneous networks substantially enlarge the class of heterogeneous network designs that can be represented and optimized within a tractable CA framework. To efficiently solve the optimization problem, we develop a sequential geometric programming framework that transforms the model into a sequence of standard geometric programming problems. Numerical experiments validate the accuracy of the proposed model and the solution method by comparing system metrics estimated by the CA models against the actual values computed from the discretized network designs. Under representative spatially heterogeneous demand scenarios, comparisons demonstrate that our model effectively reduces generalized costs by over 7% against existing homogeneous and restricted heterogeneous TND models. Key findings indicate that: (i) the proposed framework consistently outperforms these conventional counterparts across all tested scenarios; (ii) the fully heterogeneous structure becomes particularly advantageous when patron demand exhibits strong spatial heterogeneity; and (iii) these flexible designs yield the greatest benefits in high-demand, low-wage, and large-area cities.