Informing infrastructure interventions via novel near-miss data collection

Bicyclist fatalities have increased 77% since 2010. Using infrastructure to eliminate conflict with motor vehicles, meanwhile, is the most effective means of protecting bicyclists. Crash data indicates where interventions are needed, but near misses provide a richer dataset and can predict future incidents. To date, there is no single source of near-miss data for infrastructure managers. Velo proposes a fleet of bicycle-mounted sensors, leveraging autonomous vehicle technologies, to collect near-miss data—in particular vehicle overtakes, passing distances, and prevailing traffic speeds—to identify risk patterns in the Pittsburgh region. We leverage an existing, commercially-deployed sensor fleet and build an equitable fleet via Bike Share Pittsburgh. We focus on creating observational models of risk, using near-miss data to identify hotspots. In collaboration with Carnegie Mellon University’s Mobility Data Analytics Center, we further extrapolate to unseen regions by augmenting our dataset using satellite imagery, traffic flow, and road geometry. Velo’s capabilities include hardware and software development, focusing on autonomous vehicle algorithms, and is supported by the Robotics Factory and Innovation Works facilities. Our anticipated results include validated predictive models capable of scaling via bike share and consumers to various regions, with potential future applications for pedestrian and transit scenarios.