�鶹��Ѱ�����

Sentao Miao: Have you ever wondered if the true secret to walkable neighborhoods is not just about how close things are—but how the experience of walking is designed to invite you out of your car and onto the street?

For decades, urban planning has focused on car-centric development, leading to congestion, pollution and less healthy lifestyles. The “15-minute city” concept—where daily needs are within a short walk—has gained traction. But traditional planning tools often focus only on accessibility: how close are shops, services, and amenities? Our research, however, asks a deeper question: what makes people actually choose to walk? To answer this, we developed a data-driven framework that goes beyond simple distance metrics.

We partnered with a major map service provider in China, analyzing millions of anonymized travel records from Beijing. This allowed us to observe real-world walking and driving patterns across the city. We used a latent class logit model—a type of discrete choice model—to estimate how different facility layouts influence people’s decisions to walk, accounting for the fact that not all trips are the same. Some trips are utilitarian, like commuting or errands; others are hedonic, like leisure or shopping. Our model captures this heterogeneity by classifying trips into these categories based on land use and temporal patterns. Our model predicts the probability of a resident choosing to walk for a given trip as a function of both objective factors—like distance and public transit—and subjective factors—such as the density and type of facilities along the route. We found, for example, that shopping facilities strongly increase walk appeal for leisure trips, while dining facilities can sometimes have a negative effect due to sidewalk crowding.

We formulated a choice-based facility layout optimization problem. This is a complex, nondeterministic polynomial time problem, so we developed an efficient greedy algorithm with provable performance guarantees, leveraging the concept of weak submodularity. Our approach allows us to determine the optimal placement of different facility types to maximize the aggregate walking probability of all residents. We tested our framework in the Wudaokou neighborhood of Beijing. Compared to traditional distance-based methods, our approach significantly improved walk appeal, with minimal impact on accessibility.

Interestingly, we found that in already walkable environments, dispersing facilities leads to higher walk appeal. But in less walkable areas, clustering amenities along central streets is more effective. Our results also revealed a “wisdom of crowds” effect in existing layouts and emphasized the importance of neighborhood context in shaping optimal design patterns. Our research demonstrates that designing for walkability requires more than just proximity. By integrating large-scale mobility data, advanced choice modeling, and optimization algorithms, we can create neighborhoods that not only function well, but also feel inviting and vibrant for pedestrians.

So, the next time you walk through your neighborhood, consider: is it designed for cars, or for people? With data-driven tools, we can help cities make walking the easy—and enjoyable—choice.