This paper presents the methodology and initial results of an effectiveness estimation effort applied to lane change crash avoidance systems. The lane change maneuver was considered to be composed of a decision phase and an execution phase. The decision phase begins when the driver desires to perform a lane change. It continues until the driver turns the handwheel to move the vehicle laterally into the new lane or until the driver decides to postpone the lane change. During the decision phase, the driver gathers information about the road scene ahead and either present or upcoming traffic or obstacles in the destination lane. The execution phase begins when the driver starts the move into the new lane and continues until the vehicle has been laterally stabilized in the destination lane. If the driver aborts the lane change once started, the maneuver execution phase concludes when the vehicle has been laterally stabilized in the original lane. CAS technology may help both phases of a lane change but the focus of this paper was on modeling the potential effectiveness of a lane change CAS to assist the driver during the decision phase.
Since the decision phase, by definition, involves no lateral motion yet occurring, kinematic or dynamical modeling approaches were not applicable. Therefore, principles from reliability theory were applied to estimate potential lane change CAS effectiveness. This method included identification of several prototypical driver-CAS interaction modes (driver-only, CAS-only, driver-&-CAS in series, driver-&-CAS in parallel), characterization of their impact on driver reliability with the CAS, and derivations of estimates for each of the different model parameters. Finally, the model and parameter estimates were applied to develop preliminary estimates of lane change CAS effectiveness in terms of the percentage of crashes avoided.
The results obtained are based on lane change crash statistics, preliminary functional specifications on CAS reliability, lane change or side-CAS usage patterns from an on-road study with truck drivers, and numerous assumptions were data are lacking. The potential effectiveness estimates span a range of from 8% to 68% of all lane change crashes avoided with an average of 37 % of all lane change crashes potentially avoided. The paper concludes with a discussion of the many caveats and research issues uncovered by this modeling exercise.