On Detecting Drunk Drivers in Mixed Autonomous Platoons Using Vehicles Velocity Measurements

Drunk drivers, who critically and continuously threaten road safety, are usually detected using biological sensors within the same vehicle or after drunk driving behavior is observed by police patrols. Observing an obvious drunk driving behavior indicates a relatively high blood alcohol concentration, and thus, detecting drunk drivers with a relatively low blood alcohol concentration is difficult. This study uses output-only measurements available from motion sensors in the longitudinal drive of mixed autonomous and human-driven platoons to detect drunk drivers’ behavior in human-driven vehicles within the same platoon. The human-driven vehicle is assumed to be between two autonomous vehicles that are able to share information with each other. The proposed approach relates a set of motion sensor measurements with another within the platoon and does not require the knowledge of the excitation signal or the dynamics of the platoon. Numerical simulations are first implemented to test the proposed approach, and then, VISSIM software is implemented to simulate realistic road topography. Further validations on laboratory mobile robots are presented in this article, where a class of abnormal driving conditions that includes human-driven vehicles is simulated. Experiments are carried out by measuring abnormal human-driven vehicles within connected autonomous robots. This method is shown to deal with various system uncertainties, and the approach can deal with the drunk driver case. The drunken driving conditions are modeled mathematically with different blood alcohol concentration levels, which correspond to different effects and different drunk driving behaviors. The proposed transmissibility-based drunk driver detection algorithm is shown to detect drivers with risky relatively low alcohol concentrations.