Behavior Decision and Path Planning for Cognitive Vehicles using Behavior Networks

Abstract

Behavior decision in human brains is a process which involves different regions of the brain, each one considering specific driving aspects. These regions interact with each other and are able to stimulate or suppress other areas. The executed behavior is a result of a fusion process and depends on the structure of the network and the motivations of individual behaviors. At the Institute of Interactive Diagnosis-and Servicesystems in Karlsruhe, a behavior based architecture, called Behavior Network, was developed and is used since several years to control walking machines. The used approach of coupling and modularization of behaviors on different layers with reactive or deliberative character can be transferred to cognitive vehicles, since the complexity of behavior interaction is also given here. This biologically-oriented method seems adequate to derive behaviors for driving and perception, which are often combinations of several sub-behaviors with different motivations (road-following, lane-keeping, speed control, collision avoidance). The attempt to control the vehicle with human-like behaviors has the advantage of good traceability of the executed manoeuvres, as well as the use of humans as teachers to parameterize behaviors. This paper describes how to implement a behavior network for road traffic with basic behaviors to execute safe driving manoeuvres, and how to ensure safety in certain situations by using fusion nodes.