Accellerating PRISM Validation of Vehicle Platooning Through Machine Learning

Abstract

The paper deals with machine learning to accelerate model-checking of vehicle platooning. The idea is to simulate collision events and provide a set of data for using machine learning and obtain intelligible rules. From the set of rules, new simulations are made in order to trigger formal verification of border ranges between collision and safety. Using the PRISM tool, it is possible to obtain probability information on border region (hardly obtainable under clear box machine learning) and correct rules definition according to formal verification.