Evaluating Location Compliance Approaches for Automated Road Vehicles

摘要

This work presents techniques for efficiently checking location compliance of automated road vehicles. We refer to location compliance as an allowed translational and rotational positioning of a vehicle on a road network, i.e., the vehicle does not enter forbidden lanes or regions reserved for other traffic participants, such as bike lanes or dedicated bus routes. Previous work has mostly focused on efficient collision detection between traffic participants and static obstacles represented as bounded sets. We formulate location compliance as a set enclosure problem, which cannot be solved directly with collision detection; thus, different algorithms from computational geometry have to be applied. We present polygon enclosure and boundary mesh generation approaches and evaluate them using existing road geometries from the CommonRoad database. For a fair comparison, we generate thousands of random instances which are evaluated statistically.