Analysis of the Field Effectiveness of General Motors Model Year 2018-2022 Advanced Driver Assistance System Features

Abstract

Executive Summary: This effort is the sixth in a series of studies examining the field effectiveness of various GM Advanced Driver Assistance Systems (ADAS) features aimed at addressing a wide range of system-relevant crash types. The current updated GM MY 18-22 study employed VIN-linked feature ADAS content data from 13,240,512 vehicles across all GM brands (i.e., Buick, Cadillac, Chevrolet, and GMC). These data were matched to police report crash data from 15 states, which resulted in 654,129 matched crash cases. ADAS feature effectiveness (i.e., percent reductions in system-relevant crashes) was estimated using “quasi-induced exposure” logistic regression. This method compares system-relevant and system-irrelevant (referred to as “control”) crash counts for equipped and unequipped vehicles. This controls for the lack of traditional exposure data (e.g., miles traveled) by selecting control crashes that should be unaffected by the feature examined (i.e., control crashes should occur at a similar rate in both ADAS equipped and unequipped vehicle populations). The logistic regression estimates were made adjusting for various covariates, including driver demographics (age and gender), speed limit, driver behavior (alcohol, fatigue, and distraction presence), driving context (weather, road, and road surface conditions), crash year, model year, and vehicle type/model. For the forward collision and lane departure features examined, sample sizes were large enough to support additional analyses of feature effectiveness for a more restricted set of crashes coded by the police to have “suspected injury” or higher injury severity for anyone involved in the crash (defined as “K”, “A” or “B” on the KABCO injury scale), which will be referred to in the summary below simply as the “injury” analysis. This injury-focused analysis can be contrasted with the “all crashes” analysis, which did not consider the police-reported injury level.