With rapid growth of megacities in developing countries comes an ever-increasing need for adequate urban infrastructure. In particular, public transport systems are key to enabling sustainable mobility in the face of unprecedented urbanisation. In recent decades, many cities have turned to centralized bus rapid transit systems (BRTs) as a flexible and affordable means to transport millions across an ever-changing urban landscape.
How should such systems optimally be designed? Working with Professor Rema Hanna, Professor Gabriel Kreindler, and collaborators, I explored ways to quantify and optimize the design of the TransJakarta BRT system in Jakarta, a massive bus network serving around 1 million people each day.
To answer this question, we started by setting up an economic geography model of commuting behavior that allowed us to make predictions of how bus ridership would change under different network designs. Identifying optimal networks in such systems is notoriously challenging as their inherent complexity and the high dimensionality of the problem can make even sophisticated computers quickly run out of computing power. To manage this challenge, the team employed an algorithm based on simulated annealing, an optimization technique used to solve high-dimensional vehicle routing problems. This algorithm constantly proposes thousands of new network designs (Fig. 1) and models the current network after the most promising proposals. After carefully making simplifying assumptions and amassing thousands of lines of code, we were thus able to assess the impacts of changes to the BRT on ridership, wait times, and commuting volumes.
Ultimately, the goal of this work is to characterize and better understand an efficient design of the TransJakarta bus route network, that is, finding a bus network that gets most Jakartans to where they need to be quickly, directly and with low wait times.
Figure 1: Example of simulated bus route network in the study area. The light blue area is the Special Capital Region of Jakarta (Jakarta DKI).
This project is being developed by Tilman Graff, supervised by Prof. Rema Hanna and Prof. Gabriel Kreindler, and sponsored by HMUI.
Researcher: Rema Hanna