Comprehensive Performance Assessment of Passive Crowdsourcing for Counting Pedestrians and Bikes

The literature review clearly demonstrates that emerging technologies have the main advantage of providing much broader and more diverse samples of the active transportation population with fewer resource constraints, while traditional monitoring methods have heavily relied on the massive efforts from data collectors, which often lead to limited data collection locations and under-sampled data. Nonetheless, the conventional counting methods have the strength with widespread application and thus more historical data available for various assessment purposes. Most transportation agencies have standard practices or guidance for collecting non-motorized data using the typical methods. In addition, they have developed “rules-of-thumb” to adjust the data collected to enhance the accuracy based on a large multitude of  evaluation/validation studies which cover various counting technologies , expansion of short-term counts to longer-period ones, estimate of average daily volumes, development of weather and other adjustment factors, calibration of nonmotorized volume modeling, etc. In contrast, there is little research dedicated to the evaluation of emerging counting methods even though there are growing applications of such technologies. As of now, most of these evaluation studies have been centered on the crowdsourced active data. As for passively crowdsourced data, little research has been conducted based on small-scale GPS and MPP data. To the best knowledge of authors, there is no evaluation performed based on the LBS for the walking and cycling count data accuracy. To fill the research gap, the proposed study is dedicated to the comprehensive assessment of LBS data accuracy for non-motorized metrics recently made available by StreetLight. Moreover, development of system, temporal, weather, land use and facility type adjustment factors will be carried out using the deep learning technology. 

The proposed study dedicated to the evaluation the counting performance of emerging technology is expected to benefit Californians in many ways:

  1. Shed extensive light on the understanding of the non-motorized counting accuracy associated with the emerging passive crowdsourcing technology, which is very important information for various California jurisdictions to make the proper choice of the available counting technologies.
     
  2. The better pedestrian and bike counts resulting from this project could aid the Californians in the following areas:
    1. Accurate modeling of transportation networks and estimating annual volumes
    2. Enhanced prioritization of pedestrian and bicycle projects
    3. Precise non-motorist exposure for relative risk analyses.
    4. Reliable tracking of changes in pedestrian and bicycle activity over time
    5. Better evaluation of the effects of new infrastructure on pedestrian and bicycle activity

Principal Investigator: 

Wen Cheng

PI Contact Information: 

wcheng@cpp.edu
Cal Poly Pomona

Project Number: 

2025