基于WiFi的定位技术中,对接收信号强度(received signal strength indication,RSSI)的平稳性要求较高,本文在分析室内WiFi信号强度统计特征的前提下,以Friis传输方程和运动学方程为基础,利用抗差卡尔曼滤波方法估计信号强度,达到了信号...基于WiFi的定位技术中,对接收信号强度(received signal strength indication,RSSI)的平稳性要求较高,本文在分析室内WiFi信号强度统计特征的前提下,以Friis传输方程和运动学方程为基础,利用抗差卡尔曼滤波方法估计信号强度,达到了信号平滑的目的,从源头上为WiFi定位精度提供保障,定位结果表明采用本文所提方法可以明显提高定位精度。展开更多
In order to meet the requirements of high-precision vehicle positioning in complex scenes,an observation noise adaptive robust GNSS/MIMU tight fusion model based on the gain matrix is proposed considering static zero ...In order to meet the requirements of high-precision vehicle positioning in complex scenes,an observation noise adaptive robust GNSS/MIMU tight fusion model based on the gain matrix is proposed considering static zero speed,non-integrity,attitude,and odometer constraint models.In this model,the robust equivalent gain matrix is constructed by the IGG-Ⅲmethod to weaken the influence of gross error,and the on-line adaptive update of observation noise matrix is carried out according to the change of actual observation environment,so as to improve the solution performance of filtering system and realize high-precision position,attitude and velocity measurement when GNSS signal is unlocked.A real test on a road over 600 km demonstrates that,in about 100 km shaded environment,the fixed rate of GNSS ambiguity resolution in the shaded road is 10%higher than that of GNSS only ambiguity resolution.For all the test,the positioning accuracy can reach the centimeter level in an open environment,better than 0.6 m in the tree shaded environment,better than 1.5 m in the three-dimensional traffic environment,and can still maintain a positioning accuracy of 0.1 m within 10 s when the satellite is unlocked in the tunnel scene.The proposal and verification of the algorithm model show that low-cost MIMU equipment can still achieve high-precision positioning when there are scene feature constraints,which can meet the problem of high-precision vehicle navigation and location in the urban complex environment.展开更多
针对传统视觉即时定位与建图(simultaneous localization and mapping,SLAM)算法的环境静态假设在高动态场景下不成立,导致无法实现准确定位的问题,通过在视觉SLAM前端引入语义模块、优化动态特征点剔除策略,构建动态鲁棒的相机自定位...针对传统视觉即时定位与建图(simultaneous localization and mapping,SLAM)算法的环境静态假设在高动态场景下不成立,导致无法实现准确定位的问题,通过在视觉SLAM前端引入语义模块、优化动态特征点剔除策略,构建动态鲁棒的相机自定位系统。引入YOLOv4识别动态和静态目标,根据特征点与动、静态目标框的位置关系及动态点占比将所有特征点划分为动态和静态,将动态点从定位算法中剔除。为准确评估算法有效性,构建复杂城市道路场景数据集,实验结果表明,该方法能有效抑制动态目标给相机自定位带来的不利影响,在多段图像序列中实现更低的定位误差,提升相机的定位精度和运动轨迹准确性。展开更多
文摘基于WiFi的定位技术中,对接收信号强度(received signal strength indication,RSSI)的平稳性要求较高,本文在分析室内WiFi信号强度统计特征的前提下,以Friis传输方程和运动学方程为基础,利用抗差卡尔曼滤波方法估计信号强度,达到了信号平滑的目的,从源头上为WiFi定位精度提供保障,定位结果表明采用本文所提方法可以明显提高定位精度。
基金Youth Program of National Natural Science Foundation of China (No. 41904029)Scientific Research Project of Beijing Educational Committee (No. KM202010016009)。
文摘In order to meet the requirements of high-precision vehicle positioning in complex scenes,an observation noise adaptive robust GNSS/MIMU tight fusion model based on the gain matrix is proposed considering static zero speed,non-integrity,attitude,and odometer constraint models.In this model,the robust equivalent gain matrix is constructed by the IGG-Ⅲmethod to weaken the influence of gross error,and the on-line adaptive update of observation noise matrix is carried out according to the change of actual observation environment,so as to improve the solution performance of filtering system and realize high-precision position,attitude and velocity measurement when GNSS signal is unlocked.A real test on a road over 600 km demonstrates that,in about 100 km shaded environment,the fixed rate of GNSS ambiguity resolution in the shaded road is 10%higher than that of GNSS only ambiguity resolution.For all the test,the positioning accuracy can reach the centimeter level in an open environment,better than 0.6 m in the tree shaded environment,better than 1.5 m in the three-dimensional traffic environment,and can still maintain a positioning accuracy of 0.1 m within 10 s when the satellite is unlocked in the tunnel scene.The proposal and verification of the algorithm model show that low-cost MIMU equipment can still achieve high-precision positioning when there are scene feature constraints,which can meet the problem of high-precision vehicle navigation and location in the urban complex environment.
文摘针对传统视觉即时定位与建图(simultaneous localization and mapping,SLAM)算法的环境静态假设在高动态场景下不成立,导致无法实现准确定位的问题,通过在视觉SLAM前端引入语义模块、优化动态特征点剔除策略,构建动态鲁棒的相机自定位系统。引入YOLOv4识别动态和静态目标,根据特征点与动、静态目标框的位置关系及动态点占比将所有特征点划分为动态和静态,将动态点从定位算法中剔除。为准确评估算法有效性,构建复杂城市道路场景数据集,实验结果表明,该方法能有效抑制动态目标给相机自定位带来的不利影响,在多段图像序列中实现更低的定位误差,提升相机的定位精度和运动轨迹准确性。
