Interaction Between Digital Road Map Systems And Trinocular Autonomous Driving

Abstract

In addition to vehicles that are guided automatically along roads of preselected types, a project is described that guides a road vehicle to execute a complete driving mission. To increase the robustness of the current implementation a-priori knowledge about the vehicle's environment is extensively used. A digital road map can provide long range knowledge. The navigation system Travelpilot, commercially offered by Robert Bosch GmbH, includes a digital map that is used as a base to locate the vehicle and for route planning. Short range information about the vehicle's current state is derived from a sequence of images taken by three cameras behind the windscreen. This arrangement offers the necessary field of view for manoeuvres like turning at sharp corners. Initiated by machine vision surveillance, road junctions are tracked using a dynamic model of the vehicle-road-status. Using the knowledge about the route and the next intersection type, as contained in the road map, the most suited camera among the three available onces can be selected. This leads to a robust performance of manoeuvres even on more complex road networks