一种单目相机/三轴陀螺仪/里程计紧组合导航算法
|
摘要
针对车载基于MEMS-IMU/里程计/GPS组合导航系统在无GPS信号情况下,定位误差发散较快问题,提出一种基于单目相机/三轴陀螺仪/里程计紧组合导航方法,实现了较高精度定位。该方法分为初始化阶段和导航阶段两部分,在初始化阶段,通过三轴陀螺仪和里程计在线估算单目相机的尺度因子;在导航阶段,先通过基于三轴陀螺仪/里程计的航位推算实时解算载体位姿,再将单目相机观测到的带有尺度因子的特征点作为观测量,通过非线性优化方法,估计出载体位姿误差并进行修正。利用四轮小车开展算法验证实验表明,室外非闭环情况下累积误差约为里程的1.9%,在室外闭环情况下累积误差为里程的0.5%。
Aiming at the problem of fast divergence of positioning errors in vehicle integrated navigation system based on MEMS-IMU/odometer/GPS when without GPS signal, a method of tightly coupled navigation based on monocular camera/three-axis gyroscope/odometer is proposed, which achieves high precision positioning. The method is divided into two parts: initialization stage and navigation stage. In the initial stage, the scale factor of monocular camera is estimated on-line by three-axis gyroscope and odometer;in the navigation stage, the carrier position and attitude are calculated in real time by three-axis gyroscope/odometer based dead reckoning, and then the feature points with scale factor observed by monocular camera are taken as observation variables, and the carrier position and attitude errors are estimated and corrected by non-linear optimization method. Experiments using four-wheeled car show that the cumulative error is about 1.9% of the mileage in the outdoor non-closed-loop case and 0.5% of the mileage in the outdoor closed-loop case.
Aiming at the problem of fast divergence of positioning errors in vehicle integrated navigation system based on MEMS-IMU/odometer/GPS when without GPS signal, a method of tightly coupled navigation based on monocular camera/three-axis gyroscope/odometer is proposed, which achieves high precision positioning. The method is divided into two parts: initialization stage and navigation stage. In the initial stage, the scale factor of monocular camera is estimated on-line by three-axis gyroscope and odometer;in the navigation stage, the carrier position and attitude are calculated in real time by three-axis gyroscope/odometer based dead reckoning, and then the feature points with scale factor observed by monocular camera are taken as observation variables, and the carrier position and attitude errors are estimated and corrected by non-linear optimization method. Experiments using four-wheeled car show that the cumulative error is about 1.9% of the mileage in the outdoor non-closed-loop case and 0.5% of the mileage in the outdoor closed-loop case.
引文
[1]齐广峰,吕军军锋.MEMS惯惯性技术的发展展及应用[J].电子设计工程,2015(1):87-899.Qi G,Lv J.Evolution and application of MEMS inertial l technology[J]].Electronic Design Engineee ring,2015(1)::87-89.
[2]杨涛,赵子阳阳,李醒飞,等.多星座GNSS/INS紧耦合合方法[J].中国国惯性技术学报报,2015,23(1)::38-42.Yang T,Zha o Z,Li X,et a l.Tightly-coup led integration n method for m ulti-constellati on GNSS/INS[J].Journal of f Chinese Inertii al Technology,2015,23(1):388-42.
[3]Halder S,Ghh osal A.A surv vey on mobile anchor assisted d localization techniques in wireless sensoo r networks[J]..Wireless Netww orks,2016,222(7):1-20.
[4]Morales J,Roysdon P,Kas ssas Z.Signals of opportunity y aided inertiall navigation[C]]//Proceedings of ION GNSSSConference.22016:1492-15001.
[5]严恭敏,秦永永元,杨波.车车载航位推算系系统误差补偿技技术研究[J].西西北工业大学学学报,2006,24(1):26-30.Yan G,Qin Y,Bo Y.On Error compenn sation techno--logy for vehii cular Dead Reec ckoning(DR)s ystem[J].Jour--nal of Nortt hwestern Poly ytechnical Unii versity,2006,,24(1):26-30.
[6]Mur-Artal R,Tardos J D.Fast relocalisaa tion and loop p closing in kee yframe-based SLAM[C]//20114 IEEE Inter--national Coo nference on Robotics andd Automation n(ICRA).20144:846-853.
[7]Mur-Artal R,Montiel J M MM,Tardos J D.ORB-SLAM:a a versatile and accurate mono ocular SLAM syy stem[J].IEEEEtransactions oo n Robotics,20015,1:16.
[8]周绍磊,吴修振,刘刚,等.一种单目视觉ORB-SLAM//INS组合导航方法[J].中国惯性技术学报,2016,24(5)::633-637.Zhou S,Wu XX,Liu G,et al.Integrated navv igation method d of monoculaa r ORB-SLAM/INS[J].Journn al of Chinese e Inertial Technn ology,2016,224(5):633-637.
[9]Forster C,Carlone L,Della aert F,et al.IMU preintegra--tion on mani fold for efficien nt visual-inertiaa l maximum-a--posteriori estt imation[C]//Ro obotics:Sciencee and Systems..Rome,2015.
[10]]Forster C,Pizzoli M,Scar ramuzza D.SVO:Fast semi--direct monocc ular visual odo ometry[C]//IEEEE International l Conference oo n Robotics and d Automation.22014:15-22.
[2]杨涛,赵子阳阳,李醒飞,等.多星座GNSS/INS紧耦合合方法[J].中国国惯性技术学报报,2015,23(1)::38-42.Yang T,Zha o Z,Li X,et a l.Tightly-coup led integration n method for m ulti-constellati on GNSS/INS[J].Journal of f Chinese Inertii al Technology,2015,23(1):388-42.
[3]Halder S,Ghh osal A.A surv vey on mobile anchor assisted d localization techniques in wireless sensoo r networks[J]..Wireless Netww orks,2016,222(7):1-20.
[4]Morales J,Roysdon P,Kas ssas Z.Signals of opportunity y aided inertiall navigation[C]]//Proceedings of ION GNSSSConference.22016:1492-15001.
[5]严恭敏,秦永永元,杨波.车车载航位推算系系统误差补偿技技术研究[J].西西北工业大学学学报,2006,24(1):26-30.Yan G,Qin Y,Bo Y.On Error compenn sation techno--logy for vehii cular Dead Reec ckoning(DR)s ystem[J].Jour--nal of Nortt hwestern Poly ytechnical Unii versity,2006,,24(1):26-30.
[6]Mur-Artal R,Tardos J D.Fast relocalisaa tion and loop p closing in kee yframe-based SLAM[C]//20114 IEEE Inter--national Coo nference on Robotics andd Automation n(ICRA).20144:846-853.
[7]Mur-Artal R,Montiel J M MM,Tardos J D.ORB-SLAM:a a versatile and accurate mono ocular SLAM syy stem[J].IEEEEtransactions oo n Robotics,20015,1:16.
[8]周绍磊,吴修振,刘刚,等.一种单目视觉ORB-SLAM//INS组合导航方法[J].中国惯性技术学报,2016,24(5)::633-637.Zhou S,Wu XX,Liu G,et al.Integrated navv igation method d of monoculaa r ORB-SLAM/INS[J].Journn al of Chinese e Inertial Technn ology,2016,224(5):633-637.
[9]Forster C,Carlone L,Della aert F,et al.IMU preintegra--tion on mani fold for efficien nt visual-inertiaa l maximum-a--posteriori estt imation[C]//Ro obotics:Sciencee and Systems..Rome,2015.
[10]]Forster C,Pizzoli M,Scar ramuzza D.SVO:Fast semi--direct monocc ular visual odo ometry[C]//IEEEE International l Conference oo n Robotics and d Automation.22014:15-22.