Perceived walkability using machine learning

My goal is to improve urban mobility based on the concept of walkability and with the assistance of big data and machine learning to address the challenges. My solution focuses on the first two challenges (road safety and encouraging more walking), while the last two I only propose complementary measures based on the concept of walkability. This is because enhancing road safety and encouraging walking, if successfully addressed, can effectively reduce traffic congestion, and potentially encourage the use of public transport.

  • Road safety from walkability point of view

Safety on the road does not only refer to statistical indicators. It is often deeply rooted in psychological perception. At neighborhoods level, areas with a diverse mix of land use or with small blocks tend to be safer. At the street level, enhancing safety involves measures such as limiting car speeds, introducing greenery to separate carriageways and footpaths. At the same time, the transparency of the building facades can also greatly influence pedestrians’ perceptions of street safety.

We can use street view images or city models (e.g., digital twin model) to build a survey application where citizens can rank the safety of an image or location based on the model. Once sufficient information is available, computer vision techniques can be applied to process the information in the images or in the model to understand the correlation between perceived safety and environmental variables. This information can be combined with other safety indicators (e.g., noise level, density of road networks) to train classification models that can facilitate citywide safety predictions.

The effectiveness of the model can be evaluated by experts. Once the locations with low levels of safety are identified by the classification models, we can improve the environment according to the correlation results which are originally collected from citizens. This iterative process, grounded in both data and public perception, forms a comprehensive approach to enhancing road safety in the city.

  • Encouraging more walking

Many studies shows that street level walkability has more impact on walking behaviors than walkability on a macro scale. Previous research has shown that many aspects (e.g., perception of safety, pavement quality, vegetation, transparency, visual complexity, aesthetic quality) have an impact on walking. Our goal is to use machine learning to understand these metrics through data (e.g., streetscape, noise, weather, surveys) and build classification models to predict the quality of walking environment in cities. The solution can be integrated with previous road safety solutions using a similar process where environmental and perceptual data is collected and then used for model training and prediction of walkability in the city. The walking experience is then verified to be improved by improving the environment of the walkable row difference.

  • Traffic congestion from walkability point of view

There is a concept called induced demand in traffic issues. It suggests that the wider the road is, the more traffic it tends to attract, because the widened road attract more latent demands. Addressing traffic congestion primarily involves promoting public transport and active commuting. If roads are wide and parking is convenient, more people will inevitably be encouraged to use private cars. Similarly, if the efficiency of public transport does not counteract the time spent commuting via private vehicles and if the infrastructure is not conducive for pedestrians or cyclists, convincing people to alter their habits becomes very challenging. Therefore, the most important issue in solving traffic congestion is not to widen roads, but to enhancing the efficiency of public transport system and at the same time creating pedestrian and cyclist-friendly environment.

  • Encourage public transport from walkability point of view

To encourage more people to use public transport, there are several things need to consider.

First, the public transport system should cover people’s origin (e.g., home) and destination (e.g., workplace). Usually, the destination is the downtown area, which generally has better public transport. The problem is that if the origin is a low-density suburb with low coverage of the public transport system, then no matter how good the public transport is in the downtown, people will still choose to use the private car. The continuity of the transport network and whether there can be a seamless transition from one mode of transport to another is essential. Second, the public transport needs to make ensure that people do not have to wait too long. If waiting times need to exceed 10 minutes, then public transport loses its advantages.  Third, it is important to improve the facilities of public transport, such as the design of bus stops, and whether people feel that waiting for a bus is too noisy from the roadside or they feel visually unsafe, will also fail to encourage people to use public transport.