Cycling Safety, Slow Mobility and Street Sensing

The rapid urbanization of cities has intensified demand for in- novative sensing solutions to monitor complex urban dynamics effi- ciently (Du et al., 2019). Drive-by sensing (DS), which leverages sensor- equipped vehicles for spatial–temporal data collection, has gained prominence due to its cost-effectiveness and extensive coverage po- tential (Birenboim et al., 2021; Ji et al., 2023a; Ma et al., 2014; Anjomshoaa et al., 2018). The selection of appropriate sensing plat- forms depends critically on specific monitoring requirements. For air quality (Hasenfratz et al., 2015; Nagendra Shiva et al., 2019) and urban heat island phenomena (Fekih et al., 2021), high-frequency sampling at high spatial resolutions (typically 1 km grids) is essential. Taxis (O’Keeffe et al., 2019; Chen et al., 2020; Hou et al., 2025b) and buses (Ji et al., 2023b; Ariss et al., 2024; Huang et al., 2024), with their expansive operational ranges and continuous service patterns, are well- suited for such grid-based monitoring. In contrast, dedicated sensing vehicles remain confined to targeted small-scale inspections due to their ∗ Correspondence to: Room 501, New Architecture Building, Tsinghua University, Haidian District, Beijing, 100084, China. E-mail address: ylong@tsinghua.edu.cn (Y. Long). 1 These authors contributed equally to this work. 2 We define ‘‘urban features’’ as physical, functional, and perceptual attributes observable at street-view level.

The surge in global electric bicycle ownership has exerted immense pressure on bicycle infrastructure. Theo- retically, there ’ s a need to reassess the risk factors associated with multiple bike lane users. Based on this, there ’ s a practical need to re-evaluate the safety and quality of outdated infrastructure. This paper aims to reconsider risk factors related to bicycle infrastructure safety in the context of electric bicycles sharing lanes with traditional bicycles. Moreover, many countries lack precise spatial data concerning bicycle infrastructure. This study in- troduces a mobile sensing method based on bicycles, aiming to acquire daytime and nighttime bike lane datasets in a cost-effective, efficient, and large-scale manner. A computer vision-based bicycle risk factor assessment model was established, and the distribution of bicycle safety risk factors was visually analyzed. Research data was collected from a representative 59.5-kilometer bicycle lane area in Beijing. The results confirm the signif- icant impact of the surge in electric bicycles, with electric bike users accounting for 72.1% of cyclists, 32.3% wearing helmets, and 8.4% riding against traffic. During the day, the highest-ranking risk factors include the type of bicycle lanes (half lacking dedicated lanes or being shared), roadside parking, and subpar road conditions. At night, insufficient street lighting are notable concerns. The research methodology is easily replicable and can be extended to new multi-user coexistence cycling environments or countries without bicycle spatial data, offering insights for bicycle safety policies and road design.

根据中国自行车协会的最新数据，截至 2022 年，中国电动自行车的保有量已攀升至 3.5 亿辆，逐渐成为众多市民的首选出行方式。然而，传统自行车与新型电动自行车的大规模并行使用，无疑加剧了骑行环境的安全风险。现有研究缺乏针对骑行环境安全评价的指标体系，无法对骑行环境中的潜在风险进行量化分析。在此背景下，本研究聚焦北京市四环内骑行环境的现状，以自采集的方式获取了北京四环内 2963.4 公里道路共 116107 张骑行图像作为数据基础，构建了一套包含 12 项风险因素的指标体系，并确定各风险因素的权重，以进行加权计算得到风险指数，最后构建计算机视觉模型识别各风险因素，并对识别结果进行聚类。本研究还在梳理北京市近年来慢行交通政策的基础上，结合风险因素识别结果，为未来北京市骑行环境建设提供了建议与参考。

This output forms part of the Beijing City Lab research direction on Cycling Safety, Slow Mobility and Street Sensing. Download the publication below.

Cycling records from emerging dockless bike-sharing services provide new opportunities to gain insight into the interactions between multiple fine- scale cycling characteristics and built environmental elements. Using Beijing as an example and the street as the analytic unit, this study examined the associations between three cycling characteristics and spatial visual elements while controlling for other built environmental features. The results showed that most visual elements were significantly associated with cycling characteristics, but their performance differs across models for trip distance, speed, and volume. The results also indicated that individuals riding long distances or at fast speeds preferred streets with more sky and greenery views. Likewise, wider streets with less spatial disorder, tended to have a higher riding volume. The findings can enhance the understanding of cycling behaviors and promote the implementation of urban design for more bikeable streets.

55BUILT ENVIRONMENT VOL 46 NO 1 WHAT MAKES A CITY BIKEABLE? theft locks, GPS tracking, and mobile appli- cation-based user interfaces, have enabled a wider deployment of dockless shared bikes (Si et al ., 2019). In the Chinese context, the bike-sharing market started to grow rapidly in 2016, with OFO, Mobike, Bluegogo and other similar start-ups entering the market. Driven by investments and growth, 20 million shared bikes were launched nationwide in just two years (Bi, 2018). These dockless bikes have several key merits: bike locations are searchable on mobile applications, users can return bikes anywhere anytime, and bike Bikeability Biking is beloved by a wide range of people for its environmental (Zhang and Mi, 2018) and physical health (Giles-Corti et al., 2010) benefi ts. Many cities worldwide started shared bike programmes years ago (DeMaio, 2009; Fishman et al., 2013; Shaheen et al., 2010; Si et al., 2019). However, it was only recently that shared bikes became ‘smart’ as a result of the development of information technology (Guo et al ., 2017). These technological advance- ments, such as the implementation of anti- What Makes a City Bikeable? A Study of Intercity and Intracity Patt erns of Bicycle Ridership using Mobike Big Data Records YING LONG and JIANTING ZHAO This paper examines how mass ridership data can help describe cities from the bikers’ perspective.