摘要
提出了一种用于车辆自主导航的多传感器数据融合方法。首先分析了车辆自主导航过程的态势与威胁,建立了车辆行驶某时刻场景模型。然后综合现有各数据融合模型的优势,运用混合、分层的融合结构改进自主行驶的有效性和准确性。最后将该方法集成到湖南大学自主研制的无人驾驶汽车测试平台上。在城市道路的实际场景中,初步实现了基于路线和交通锥形标的自主导航和行驶任务,车辆自主行驶的融合决策准确率可达到97.5%,表明该方法具有可接受的性能。
A multi-sensor data fusion scheme for vehicle autonomous navigation is proposed in this paper. Firstly, the situation and threat in the course of vehicle autonomous navigation are analyzed and a vehicle driving scene model at certain moment is established. Then the advantages of existing data fusion models are synthesized and a kind of hybrid layered fusion structure is used to improve the accuracy and effectiveness of autonomous driving. Finally, the scheme is integrated into the test platform for the unmanned vehicle developed by Hunan University. In the actual scene on urban road, the tasks of lane and traffic cone-based autonomous navigation and driving are preliminarily accomplished with the accuracy rate of fusion decision making for vehicle autonomous 97.5 %, indicating that the scheme proposed has an acceptable performance. driving reaching
出处
《汽车工程》
EI
CSCD
北大核心
2009年第7期640-645,共6页
Automotive Engineering
基金
教育部科技创新工程重大项目培育项目(708067)
教育部长江学者与创新团队发展计划项目(531105050037)资助