Development of neural and symbolic AI control architecture for overtaking

Abstract

A hybrid model approach which combines neural and symbolic AI methodologies has been introduced as a control system framework for the autonomous guidance of vehicles of the future. The subsumption architecture, originally established by R. Brooks of MIT and later expanded by P. Maes of VUB Brussels is used as the philosophical foundation of the architecture. A series of experiments using the developed architecture for the overtaking by an autonomous vehicle of other vehicles on a structured roadway is being attempted. There, the flow of traffic on a two-lane roadway (through-traffic and passing lanes) in which the control system is being tested, has been modeled by a simulator. The proposed control system is tested on this simulator. The paper describes the over-all architecture of the autonomous control system, including the simulation environment, a description of the experiments being conducted, as well as preliminary results from the experiments.<>