A margin–based approach to vehicle threat assessment. (9th March 2015)

• Record Type: Journal Article
• Authors: Constantin, Alexandre; Park, Junghee; Iagnemma, Karl
• Is Part Of: International journal of vehicle autonomous systems. Volume 12:Number 4(2014)
• Page Start: 384
• Page End: 411
• Publication Date: 2015-03-09
• Subjects: vehicle threat assessment -- driving assistance systems -- homotopy class -- lattice motion planning -- driver assistance -- autonomous navigation -- control margin estimation -- vehicle safety -- hazardous events; Autonomous vehicles -- Periodicals; Motor vehicles -- Technological innovations -- Periodicals; Motor vehicles -- Electronic equipment -- Periodicals; Robots -- Control systems -- Periodicals; Intelligent transportation systems -- Periodicals; 629.20285
• Languages: English
• ISSNs: 1471-0226

Abstract

In this paper we propose a novel approach to the threat assessment problem for Advanced Driver Assistance System (ADAS) and autonomous navigation decision making support. This threat assessment is based on estimation of the control margin afforded to a vehicle and is performed in a multi–threat framework. Given sensor information available about the surrounding environment, an algorithm first identifies corridors of travel through which the vehicle can safely navigate. The second stage then assesses the threat posed to the vehicle in each identified corridor via a metric associated with available control margin. For this purpose, the corridors are approximated by sets of trajectories generated from a lattice sampled in the vehicle's input space. The level of threat can then serve to influence autonomous navigation as an input to a decision–making layer. It also potentially allows a semi–autonomous control system to honour driver intent while ensuring safe and robust navigation in hazardous events. The benefit of such an approach is compared to common threat metrics in canonical scenarios. The method is also applied to the multi–lane road environment of highway navigation by post processing human driving data gathered from a simulator.