城市环境下车辆高容错多模型融合定位技术研究

Accurate and reliable vehicle positioning is one of the most important foundations that can influence the function effectiveness of Intelligent Transportation Systems (ITS). However, with the increasing complexity of urban traffic environment, it often causes the several-minute failure of GPS used widely in current vehicle positioning. In this context, the integrated positioning technology based on multi-sensor fusion has become the research hotspot in recent years. Unfortunately, in the existing research, there are still obvious deficiencies in the positioning accuracy and reliability when GPS fails for several minutes in the complex urban environment. The main reason is that in the existing research, there are still problems in the accuracy of the vehicle state model, the validity of the auxiliary observation information, and the fault tolerance of fusion algorithm, etc. To solve such problems effectively, the highly-maneuvering state model of a vehicle running in the urban environment will firstly be established in this project. Then, the two-stage estimation strategy will be designed to obtain more accurate auxiliary observation information about the vehicle-body motion before it is fused. Thirdly, based on the interacting multiple-model fusion architecture, a system-level sensor fault diagnosis mechanism combining both the analytical redundant idea and the time-series redundant idea is proposed and introduced, in order to form the multiple-model fusion algorithm with high fault tolerance for vehicle positioning in the urban environment. Besides, the effectiveness of the algorithm should be demonstrated through carrying out field tests. The main innovation of this project is that the proposed fusion algorithm for vehicle positioning considers the key factors affecting the positioning comprehensively and systematically. Therefore, accurate and reliable vehicle positioning can be realized even when GPS fails continuously for 2-3 minutes in the complex urban environment, which is one of the major challenges in the ITS field. It can be concluded that this project will have important theoretical and practical value.

车辆准确可靠的定位是智能交通系统ITS功能有效发挥的一个重要基础。但日益复杂的城市环境常会造成目前广泛应用的GPS定位持续几分钟的较长时间失效。基于多传感器融合的组合定位成为近年来的研究热点，然而，现有研究对于城市环境下GPS的较长时间失效，在定位准确性和可靠性上仍存在明显的不足，主要是由于在车辆运行模型准确性、辅助观测信息有效性以及融合算法容错性等方面还存在问题。针对这些问题，本项目首先研究建立城市环境下车辆的高机动运行模型；然后，设计车体运动辅助观测信息的两级估计策略，以便在融合前获取更为有效的辅助观测信息；最后，在交互式多模型融合算法中研究建立联合解析冗余和时序冗余的系统级传感器故障诊断机制，以提出高容错的车辆多模型融合定位算法，并通过实车试验验证。本项目较为全面、系统地考虑了影响定位的关键因素，将有效解决复杂城市环境下车辆难以准确可靠定位这一ITS难题，具有重要的理论和应用价值。

车辆准确可靠的定位是智能交通系统ITS功能有效发挥的一个重要基础。但日益复杂的城市环境常会造成目前广泛应用的GPS定位持续几分钟的较长时间失效。基于多传感器融合的组合定位成为近年来的研究热点，然而，现有研究对于城市环境下GPS的较长时间失效，在定位准确性和可靠性上仍存在明显的不足，本项目针对车辆运动模型的准确性、辅助观测信息的有效性以及融合算法的容错性等方面开展了系统、深入的研究，取得的主要创新成果如下：. 1）通过深入比较车辆纵向运动和横向运动的各种常用运动学模型，从动力学角度进行改进，建立了包含多个动态模型的车辆高机动运行模型；. 2）依据车载传感器的工作特点，并结合所建立的高机动运动模型，研究了包含车辆速度、姿态角以及相对加速度等在内的大量辅助观测信息的准确可靠估计方法，提高后续融合定位的准确性和可靠性；. 3）针对城市复杂环境，提出了车辆高容错多模型融合定位算法，建立联合解析冗余与时序冗余的系统级传感器故障诊断隔离机制，并采用基于智能算法的定位误差训练补偿机制，显著提升GPS失效时车辆定位的可靠性和精度。. 目前，本项目已经实现了预期研究目标，较好地满足了车辆辅助驾驶领域中的定位精度需求。相关研究成果发表在IEEE Transactions on Industrial Electronics、IEEE Transactions on Vehicular Technology、IEEE Transactions on Intelligent Transportation Systems、Information Fusion等国际车辆、交通以及导航领域权威期刊。项目执行期间共计发表SCI收录论文14篇（其中中科院JCR一区及二区的高水平论文4篇），EI收录论文10篇，申请国家发明专利27项（已授权16项），参加学术会议7次，培养博士后1名，博士、硕士研究生13名。