A drive recorder provides useful information to analyze factors that caused accidents. The information includes the driving video, car velocity, acceleration data, brake/turning indicator data, and GPS data. In this research, 237 videos of accidents involving taxi-to-cyclist collisions at intersections were collected from drive recorders. The collision data were compared to near-miss incident data to identify the factors that made the difference between collisions and near-miss incidents. It was shown that collisions occurred when cyclists entered an area where a car deceleration over 0.55 G was necessary to stop the car from colliding with the cyclist. In left and right turn collisions, the car velocity was low and many drivers had already braked when entering the intersection. Therefore, the cause of the accident was likely related to the driver’s perception. To investigate the effectiveness of autonomous emergency braking (AEB)systems, car-to-cyclist collisions were reconstructed using PC-Crash simulations that were based on the videos from the drive recorders. Extending the field of view (FOV) of AEB system from 50° to 90° was effective in reducing the number of collisions in perpendicular and left- and right-turn collisions. With an ideal AEB with a 360° FOV and no sensor delay time, 52 of the 63 perpendicular collisions and 49 of 51 left and right turn collisions were avoidable.
Keywords:
Crash avoidance; Autonomous emergency braking; Car-cyclist collision; Accident reconstruction