电子与信息学报
電子與信息學報
전자여신식학보
JOURNAL OF ELECTRONICS & INFORMATION TECHNOLOGY
2015年
2期
484-488
,共5页
谷阳%宋千%李杨寰%马明%周智敏
穀暘%宋韆%李楊寰%馬明%週智敏
곡양%송천%리양환%마명%주지민
人员自主定位%鞋载传感器%先验知识%粒子滤波%轨迹修正
人員自主定位%鞋載傳感器%先驗知識%粒子濾波%軌跡脩正
인원자주정위%혜재전감기%선험지식%입자려파%궤적수정
Pedestrian navigation%Foot-mounted sensors%Apriori knowledge%Particle filter%Trajectory calibration
为了解决卫星导航信号被“遮挡”条件下定位问题,该文提出一种基于惯性鞋载传感器的高精度人员自主定位方法。该方法通过经典的拓展卡尔曼滤波辅助的零速更新(ZUPT-aided EKF)算法解算鞋尖惯性测量数据得到人员的初步运动轨迹,并创新性地提出一种粒子滤波框架下利用建筑物结构先验知识对轨迹进行修正的方法。根据大多数建筑物的结构,将行走平面划分为8个方向,包含4个主方向(走廊朝向)和4个辅助方向。根据粒子偏离8方向的程度按照高斯函数对粒子的权值进行更新,并用剩余重采样的方法避免了粒子的退化。实测数据验证了该文提出的方法,结果表明:该方法比轨迹修正前和传统轨迹修正的方法有更好的精度,在861 m的复杂轨迹下定位误差仅为2.7 m,定位精度优于0.5%;同时该方法有较好的一致性,不同楼层间的行走定位误差保持在2 m内,可以进行稳定持续地定位。
為瞭解決衛星導航信號被“遮擋”條件下定位問題,該文提齣一種基于慣性鞋載傳感器的高精度人員自主定位方法。該方法通過經典的拓展卡爾曼濾波輔助的零速更新(ZUPT-aided EKF)算法解算鞋尖慣性測量數據得到人員的初步運動軌跡,併創新性地提齣一種粒子濾波框架下利用建築物結構先驗知識對軌跡進行脩正的方法。根據大多數建築物的結構,將行走平麵劃分為8箇方嚮,包含4箇主方嚮(走廊朝嚮)和4箇輔助方嚮。根據粒子偏離8方嚮的程度按照高斯函數對粒子的權值進行更新,併用剩餘重採樣的方法避免瞭粒子的退化。實測數據驗證瞭該文提齣的方法,結果錶明:該方法比軌跡脩正前和傳統軌跡脩正的方法有更好的精度,在861 m的複雜軌跡下定位誤差僅為2.7 m,定位精度優于0.5%;同時該方法有較好的一緻性,不同樓層間的行走定位誤差保持在2 m內,可以進行穩定持續地定位。
위료해결위성도항신호피“차당”조건하정위문제,해문제출일충기우관성혜재전감기적고정도인원자주정위방법。해방법통과경전적탁전잡이만려파보조적령속경신(ZUPT-aided EKF)산법해산혜첨관성측량수거득도인원적초보운동궤적,병창신성지제출일충입자려파광가하이용건축물결구선험지식대궤적진행수정적방법。근거대다수건축물적결구,장행주평면화분위8개방향,포함4개주방향(주랑조향)화4개보조방향。근거입자편리8방향적정도안조고사함수대입자적권치진행경신,병용잉여중채양적방법피면료입자적퇴화。실측수거험증료해문제출적방법,결과표명:해방법비궤적수정전화전통궤적수정적방법유경호적정도,재861 m적복잡궤적하정위오차부위2.7 m,정위정도우우0.5%;동시해방법유교호적일치성,불동루층간적행주정위오차보지재2 m내,가이진행은정지속지정위。
During GPS outages, the foot-mounted inertial-based sensors are common replacement in pedestrian navigation. The Zero velocity UPdaTe-aided Extended Kalman Filter (ZUPT-aided EKF) is often used to resolve the trajectory of a walking pedestrian with acceleration and angular rate measurements from foot-mounted sensors. However, the trajectory suffers from long-term drifts, which needs to be calibrated. This paper proposes a particle filter based approach for trajectory calibration, which exploits apriori knowledge of building structures to update particle weight. The buildings are supposed to have four “domain” directions, which is defined by the layout of corridors. The navigation frame is divided by eight directions, including four “domain” directions and four complementary directions, and the weight is assigned according to the eight directions using a Gaussian function. Finally, several real-scenario experiments are carried out, which can demonstrate that the proposed approach have better accuracy and consistency than the results without calibration or traditional methods, as the proposed approach can reach a location error of 2.7 m in a complex-trajectory walk of 861 m and the accuracy is better than 0.5%; the fact that the location error remains below 2 m in different floors also demonstrates the good consistency of the approach. As a result, the proposed approach can perform stable and continuous positioning.