Logistic Regression Model-based Analysis of Self-driving Car Disengagement Influencing Factors

Accelerated by breakthroughs in sensor fusion and edge computing, autonomous vehicle (AV) technologies now face stringent regulatory requirements for disengagement event logging. As a leading global center for AV testing, California's Department of Motor Vehicles (DMV) data highlights that disengagement incidents remain a critical measure of technological maturity. Utilizing the 2019 Autonomous Vehicle Disengagement Reports, this study constructs a binary logistic regression model to systematically examine the key determinants of disengagement events. This study highlights the key contributing elements through meticulous feature selection and model interpretation, providing practical solutions to mitigate disengagement risks. Python analysis identifies three primary factors: sudden halts caused by unintended manoeuvres, inconsistencies in perception between vehicle sensors and human assessment, and scenarios requiring manual intervention to avoid stationary vehicles in the lane.