Statistical Model-Checking for Autonomous Vehicle Safety Validation - Archive ouverte HAL Access content directly
Conference Papers Year : 2017

Statistical Model-Checking for Autonomous Vehicle Safety Validation

Model-checking statistique pour la vérification de véhicules autonomes


We present an application of statistical model-checking to the verification of an autonomous vehicle controller. Our goal is to check safety properties in various traffic situations. More specifically, we focus on a traffic jam situation. The controller is specified by a C++ program. Using sensors, it registers positions and velocities of nearby vehicles and modifies the position and velocity of the controlled vehicle to avoid collisions. We model the environment using a stochastic high level Petri net, where random behaviors of other vehicles can be described. We use HASL, a quantitative variant of linear temporal logic, to express the desired properties. A large family of performance indicators can be specified in HASL and we target in particular the expectation of travelled distance or the collision probability. We evaluate the properties of this model using COSMOS 1. This simulation tool implements numerous statistical techniques such as sequential hypothesis testing and most confidence range computation methods. Its efficiency allowed us to conduct several experiments with success.
Fichier principal
Vignette du fichier
SIA_SimNum.pdf (332.31 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-01491064 , version 1 (16-03-2017)


  • HAL Id : hal-01491064 , version 1


Benoît Barbot, Béatrice Bérard, Yann Duplouy, Serge Haddad. Statistical Model-Checking for Autonomous Vehicle Safety Validation. Conference SIA Simulation Numérique, Société des Ingénieurs de l'Automobile, Mar 2017, Montigny-le-Bretonneux, France. ⟨hal-01491064⟩
828 View
510 Download


Gmail Facebook X LinkedIn More