Lan, Li and Ti – INTEGRATING GENERAL PRINCIPLES INTO MIXED INTEGER PROGRAMMING TO OPTIMIZE SCHEMATIC NETWORK MAPS 2019

Description

This study presents a new automated method for generating schematic network maps by integrating four general principles into a mixed-integer programming (MIP) framework. The principles—preserve topological relationships and the network’s main structure (global), and retain relative position and relative length of features (local)—are modeled as explicit constraints and combined with established constraints (e.g., few line bends, limited orientations, small total length). Important strokes are identified and straightened or split into sub-strokes, and an extended coordinate/direction system supports octolinear directions. The implementation uses C# to build constraints and IBM CPLEX to solve the MIP. Experiments on a Chinese high-speed rail network and the Beijing metro compare the new method with an existing MIP approach and official maps using fractal dimension, usability (clarity, recognition of major lines, visual simplicity, satisfaction), and optimality gap. Results show the new method yields lower fractal dimensions, higher usability scores close to official maps, and substantially better computational optimality gaps. Limitations include lack of automated labeling and potential infeasibility when too many strokes are fixed for preservation.