Ming Chen,Xuesong Wang
Abstract:This study aims at proposing a new forward collision warning (FCW) algorithm and verifying the effectiveness of warning through both theoretical computation and driving simulation experiment. One key characteristic of FCW is that the warnings must compatible with drivers’ natural behavior and risk perception. In a previous classical kinematic-based warning algorithm, the warning onset range was calculated as the minimum safety distance to stop based on the kinematic principle. In this algorithm, the expected response deceleration, which measures how hard the driver of subject vehicle (SV) would brake under the pre-crash scenario, was assumed as a linear function of lead vehicle (LV) deceleration rate and relative speed. However, this led to an “incompatible issue” between the output of warning rage with drivers’ risk perception. Specifically, when a LV brakes harder (in a larger deceleration rate), the warning is presented later (with a shorter warning onset range), which is opposite to drivers’ normal perception. In order to solve this problem, the predicting model for expected response deceleration was improved with a nonlinear function of LV deceleration, relative speed and their interaction term. Then a new warning algorithm was developed based on this predicting model. The modeling efforts was based on a total of 30 drivers’ braking data collected under different risk level of rear-end scenarios in the Tongji University Driving Simulator. The database includes a total of 173 rear-end events, but only 111 of them in which drivers only take the braking maneuver and no collision occurred were used to model the expected response deceleration. Domain of validity was examined through theoretical computations for a wide range of initial conditions of subject vehicle speed, relative speed, and LV deceleration rate. The proposed algorithm was then implemented and tested in the driving simulator. The general effectiveness of warnings was significant for reducing both collisions (collision rates decreased by 90%) and severity (impacting speed decreased by 93%). Drivers also show good acceptance with the warning timing of the algorithm. The results of this study can be helpful in improving the design of forward collision warning algorithm to be more effective and robust.