4-D dynamic vision for intelligent motion control

Abstract Real-time machine vision is discussed in the framework of intelligence for motion control. Remote sensing and internal representation of the spatio-temporal environment in the interpretation process including objects and agents (subjects) are considered to be vital ingredients to intelligence. A prediction-error-feedback scheme has been developed for the realization of real-time vision with moderate computing power needed, utilizing high level spatio-temporal world models. Applications investigated are: road vehicle guidence including reactions to obstacles (stopping, convoy driving and passing), vehicle docking (satellite model plant) and aircraft landing (hardware-in-the-loop simulation). The first one is discussed to some extent.