Field Study of Minnesota’s Automated Shuttle Revealed That the Rate of Yielding to Pedestrians was 12.1 Percentage Points Higher Compared to Human-Driven Vehicles.

METHODOLOGY

The study assessed the deployed shuttle technology through multiple methods:

• Operator and Manufacturer Interviews. The researchers conducted six interviews with representatives of AV manufacturers and MCM operators.
• Field Observations. Researchers observed and recorded driver behavior and driver-pedestrian interactions along the MCM route at three signalized and three unsignalized sites between May and August 2022. The team collected all data through a mixed approach of paper notation and digital notation via a survey platform. Key measurements of interest included yielding rates, frequencies of queueing behind the yielding vehicles, and overtaking rates and multiple threat pass/attempt rates.
• Minnesota State Fair Study. Eighty-five participants completed four simulated driving scenarios in which they watched a short video and were told to imagine themselves as the driver and a post-study questionnaire.
• Crowdsourcing Study. The research team conducted a preliminary study regarding MCM signal interpretation via a short online survey completed by 242 participants.
• Driving Simulation. Forty-six participants completed two simulated drives encountered the MCM with one of the three different possible signaling systems, with and without a pedestrian present.

FINDINGS

The analysis found that the direct risk between the MCM itself and pedestrians is minimal. However, the indirect risk to pedestrians due to changes in human driver activity around the MCM appeared to be significantly greater.

• Field Observations
  • Compared to regular vehicles, the MCM had a statistically significantly greater yielding percentage to pedestrians at intersections (12.1 percent).
  • Compared to regular vehicles, the MCM was associated with a nearly 30-fold increased risk of being overtaken while turning right at a signalized intersection (an adjusted risk ratio of 30.8).
  • At unsignalized intersections, regular vehicles yielded less to pedestrians (rate of .739 failure to yield per crossing) compared to the MCM (rate of 0.164 failure to yields per crossing).
  • At unsignalized intersections, the MCM had nearly four times the risk of a queue forming while yielding, 8 times the risk of a multiple threat pass attempt, and 36 times the risk of a multiple threat pass occurring.
• Minnesota State Fair Study
  • Results from the simulated driving scenario indicated primary factors that affected participants’ decision to pass were speed (52 percent), the rear lights (33 percent), or other factors (15 percent).
  • There was an observable difference in participant responses across the two signaling systems with a larger majority of participants indicating they would wait behind the shuttle under the turn signal (92.7 percent) compared to the hazard signal (78.3 percent).
• Crowdsourcing Study
  • Most participants (77.6 to 85.0 percent), indicated that they would wait behind the shuttle in the turn signal and LED message board conditions. This rate of reported waiting dropped to 58.7 percent for those who experienced the hazard condition.
• Driving Simulation
  • Participants in the text/icon condition were approximately 1.94 times more likely to stop and wait behind the MCM compared to the hazard signal condition.
  • Participants in the text/icon condition were less likely (0.38 times) to make a multiple threat pass compared to the hazard signal condition.