Spatially-aware predictors for car subnetwork optimality region in bi-modal grids with dedicated bus lanes

Understanding the relationship between network topology and performance is essential for effective traffic management and the efficient allocation of road space among different transport modes. This study focuses on bi-modal grid networks composed of car lanes and Dedicated Bus Lanes (DBLs). We construct a large dataset by simulating a wide range of these networks by varying their number of nodes, aspect ratios, block lengths, and DBL layouts. Our analysis reveals that DBL placement creates spatial heterogeneity within the car subnetwork, even for otherwise homogeneous grid layouts. This heterogeneity arises from reduced lane capacity at intersections with DBLs, which we label as critical nodes. The reduced-capacity intersections effectively subdivide the car network into two interacting subnetworks: one with DBLs and one without. We introduce spatially-aware metrics based on the mean and maximum betweenness centrality of these critical nodes and evaluate their usefulness in predicting the optimality parameters of both subnetworks, where optimality parameters include capacity and the lower and upper critical densities at which capacity is achieved. Using ordinary least squares regression, we find that including these metrics significantly improves predictive performance compared to models using only conventional aggregate measures. From a transportation planning perspective, the results highlight the importance of strategic DBL placement, suggesting that positioning DBLs on moderately important—but not the most central—nodes can better balance the needs of public transit and private vehicles.