A model-based design of an onboard stereo vision system: obstacle motion estimation for cooperative automated vehicles

Abstract

Cooperative automated vehicle software needs to use advanced software engineering design methodologies to manage the complexity of distributed software. This article presents a design model of a stereo vision system for cooperative automated vehicles based on object-oriented analysis and design methodologies and unified modeling language concepts. We use a rational unified process and the 4+1 architectural view model to design the stereo vision system. The designed model is a part of the perception system of the ego vehicle. The stereo vision system continuously perceives the traffic environment using an on-board front-facing stereo camera and sends information to the vehicle control system. The stereo vision system is developed on Ubuntu 16.04 LTS platform with the Nvidia DriveWorks framework using the C++ programming language, and it is deployed on Nvidia Drive PX2 automotive-grade embedded hardware platform. The designed and developed system is evaluated on the Carla simulation environment and the Renault Twizy cooperative automated vehicle research platform. The experimental results show that the approach is applicable, and the system is capable of running in real-time.