SWARMS: Swarm Intelligence for Autonomous Cars

SWARMS：自动驾驶汽车的群体智能

• 批准号: 273356841
• 负责人: Professor Dr. Raúl Rojas
• 金额: --
• 依托单位: Dahlem Center for Machine Learning and Robotics (DCMLR)
• 依托单位国家: 德国
• 项目类别: Priority Programmes
• 财政年份: 2015
• 资助国家: 德国
• 起止时间: 2014-12-31 至 2020-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/273356841?language=en
• 关键词: SWARMS; Swarm; Intelligence; Autonomous; Cars

In this project we want to advance the state of the art towards achieving peaceful coexistence between autonomous and non-autonomous vehicles. Our vision for the future is that autonomous and non-autonomous vehicles will recognize the implicit or explicit intentions of traffic participants, will communicate and will cooperate. During a transition phase comprising several decades, autonomous vehicles must adapt to the traffic produced by human drivers. We propose to recognize the intentions of other drivers using the trajectories, speeds, and state of their vehicles, matched to the map of the city using graph fusion. We distinguish between "implicit" and "explicit" intentions. Explicit intentions can be communicated using a V2V and V2X communication unit. Implicit intentions must be extracted from the relationship between the current vehicle behavior, the constraints of traffic, and the map of the city. While autonomous vehicles can communicate their intended trajectory to surrounding autonomous vehicles, so that they adapt their behavior, it is not obvious how to present this information to human drivers. We propose to test two kinds of HMI: in one form of interaction the information is just passed in a graphic or haptic manner to the driver, in the second form of interaction an autonomous car, or a central computer, can set constraints to the human driver. We propose to implement swarm behavior for autonomous vehicles. Our extensive tests in three countries (in Berlin, Germany, in San Antonio, USA, and in Mexico City) have shown that while human drivers adapt immediately to changes in the environment, or in the map of the city, even in the absence of traffic markings (for example for lanes), this is still extremely difficult for robotic cars. Thus, we want to extend our navigation software so that in the absence of lane markings, or even in the absence of a map, an autonomous car can become part of a "driving swarm" and can steer with the flow. Other than safer autonomous navigation, immediate further applications of the technology to be developed consist of: the creation of city maps on-the-fly (that is, with one single manual drive), autonomous path-planning without marked lanes and in unstructured environments, as well as driving in car pools.

在这个项目中，我们希望推进最先进的技术，实现自动驾驶和非自动驾驶汽车之间的和平共处。我们对未来的愿景是，自动驾驶和非自动驾驶汽车将识别交通参与者的隐含或明确意图，并进行沟通和合作。在几十年的过渡阶段，自动驾驶汽车必须适应人类驾驶员产生的交通。我们建议使用轨迹，速度和车辆状态来识别其他驾驶员的意图，并使用图形融合与城市地图相匹配。我们区分“隐含”和“明确”的意图。可以使用V2V和V2X通信单元来传达明确的意图。隐含意图必须从当前车辆行为、交通约束和城市地图之间的关系中提取。虽然自动驾驶汽车可以将其预期轨迹传达给周围的自动驾驶汽车，以便它们调整自己的行为，但如何将这些信息呈现给人类驾驶员并不明显。我们建议测试两种HMI：在一种交互形式中，信息只是以图形或触觉方式传递给驾驶员，在第二种交互形式中，自动驾驶汽车或中央计算机可以为人类驾驶员设置约束。我们建议实现自动驾驶汽车的群体行为。我们在三个国家（德国的柏林、美国的圣安东尼奥和墨西哥城）进行的广泛测试表明，尽管人类驾驶员能够立即适应环境或城市地图的变化，即使没有交通标记（例如车道），但这对于机器人汽车来说仍然非常困难。因此，我们希望扩展我们的导航软件，以便在没有车道标记的情况下，甚至在没有地图的情况下，自动驾驶汽车可以成为“驾驶群”的一部分，并可以随流转向。除了更安全的自主导航之外，该技术即将开发的进一步应用包括：即时创建城市地图（即，仅需一次手动驾驶），无标记车道和非结构化环境中的自主路径规划，以及在拼车中驾驶。