无人车的场景理解与自主运动

Scene Understanding and Autonomous Motion of Unmanned Vehicles

无人车无疑将深刻改变我们未来的交通出行和运输方式,正日益受到学术界、产业界和政府部门越来越多的关注。无人驾驶的感知-运动环路中存在两个紧密耦合的核心问题,即场景理解与自主运动。场景理解通过融合处理车载多传感器的感知数据,给出车辆运动决策的几何、物理层面的约束。而自主运动在场景理解给出的约束条件下实现安全、平稳的运动。主要综述无人车场景理解与自主运动的研究进展,并结合团队十余年来在无人驾驶领域的探索与实践,探讨复杂动态交通环境的无人车前沿技术。

Driverless cars are well known disruptive technologies,and gain more and more attention ranging from academia,enterprise,to governments. There exists two important and coupled problems in the perception-motion loop of an unmanned vehicle. They are scene understanding and autonomous motion. Scene understanding outputs motion constraints via the fusion of sensing data from multiple sensors,and autonomous motion adaptively drive the vehicle in accordance with the dynamic constraints posed by traffic scenes. In this paper,we present a comprehensive survey of the state of the art on the aspects of basic concepts,methodologies,and technologies of scene understanding and autonomous motion,based on our past 10 years of research on intelligent vehicles. We also discuss future challenging problems in the frontiers of autonomous driving.