A Generic Plug-and-Play Navigation Fusion Strategy for Land Vehicles in GNSS-Denied Environment

更新时间:2023-07-25 19:46:40 阅读: 评论:0

Apr.2019Vol.36No.2 Transactions of Nanjing University of Aeronautics and Astronautics
A Generic Plug‑and‑Play Navigation Fusion Strategy for
Land Vehicles in GNSS‑Denied Environment
LAI Jizhou*,BAI Shiyu,XU Xiaowei,LÜPin
Key Laboratory of Navigation,Control and Health‑Management Technologies of Advanced Aerocraft,Ministry of Industry and Information Technology,College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing
211106,P.R.China
(Received15March2019;revid3April2019;accepted4April2019)Abstract:Achieving accurate navigation information by integrating multiple nsors is key to the safe operation of land vehicles in global navigation satellite system(GNSS)‑denied environment.However,current multi‑nsor fusion methods are bad on stovepipe architecture,which is optimized with custom fusion strategy for specific nsors.Seeking to develop adaptable navigation that allows rapid integration of any combination of nsors to obtain robust and high‑precision navigation solutions in GNSS‑denied enviro
nment,we propo a generic plug‑and‑play fusion strategy to estimate land vehicle states.The propod strategy can handle different nsors in a plug‑and‑play manner as nsors are abstracted and reprented by generic models,which allows rapid reconfiguration whenever a nsor signal is additional or lost during operation.Relative estimations are fud with absolute nsors bad on improved factor graph,which includes nsors’error parameters in the non‑linear optimization process to conduct nsor online calibration.We evaluate the performance of our approach using a land vehicle equipped with a global positioning system(GPS)receiver as well as inertial measurement unit(IMU),camera,wireless nsor and odometer.GPS is not integrated into the system but treated as ground truth.Results are compared with the most common filtering‑bad fusion algorithm.It shows that our strategy can process low‑quality input sources in a plug‑and‑play and robust manner and its performance outperforms filtering‑bad method in GNSS‑denied environment.
Key words:GNSS‑denied;multi‑nsor fusion;plug‑and‑play;factor graph;land vehicles
多功能笔CLC number:V249.32Document code:A Article ID:1005‑1120(2019)02‑0197‑08
0Introduction
One of the esntial technologies that ensure re‑liable operation of land vehicles is navigation.Cur‑rent land vehicles heavily rely on global navigation satellite system(GNSS).However,when land vehi‑cles run in the den or even GNSS‑denied environ‑ment,GNSS signal degrades or even fails to locate land vehicles[1].
When GNSS signal is unavailable,accurate navigation solutions can be obtained through inte‑grating multiple nsors.Multi‑nsor fusion meth‑ods have been deeply studied and widely applied in the field of land vehicles[2‑4].However,the naviga‑tion systems are bad on stovepipe architecture[5],which is customized for specific nsors and mea‑surement sources.It brings about huge costs when‑ever the navigation system requires changes or up‑dates.To change existing fusion architectures,De‑fen Advanced Rearch Projects Agency(DAR‑PA),USA launched All Source Positioning and Navigation(ASPN)project in2010[6].ASPN proj‑ect aims to develop adaptable navigation that allows rapid integration of any combination of nsors to en‑able low cost,and amless navigation solutions for military urs on any operational platform and in any
*Corresponding author,E‑mail address:laijz@nuaa.edu.
How to cite this article:LAI Jizhou,BAI Shiyu,XU Xiaowei,et al.A Generic Plug‑and‑Play Navigation Fusion Strategy for Land Vehicles in GNSS‑Denied Environment[J].Transactions of Nanjing University of Aeronautics and Astronautics,2019,36(2):197‑204.
http://dx.doi/10.16356/j.1005‑1120.2019.02.002
Vol.36 Transactions of Nanjing University of Aeronautics and Astronautics
environment.Many rearchers have performed re‑
arch on ASPN.
For the software systems,Elsner and Juang de‑
signed the plug‑and‑play multinsory fusion
schemes bad on robot operating system (ROS)[7‑8].For the fusion architectures and algo‑rithms,filtering‑bad estimation methods are most‑
禁毁小说ly ud.Soloviev et al.propod reconfigurable inte‑
gration filtering Engine(RIFE).In RIFE,various
nsors are reprented by generic class.Each
class is defined by the type of nsor measurement
and the filter can be reconfigured by instantiating a
nsor object whenever a new nsor is connected to
system[9].Lynen et al.propod multi‑nsor‑fusion
extend kalman filter(MSF‑EKF)to process
time‑delayed,relative and absolute measurements
from a theoretically unlimited number of different
nsors.Its modular design allows amless han‑
dling of additional/lost nsor signals[10].Groves
propod nsor fusion modular integrated architec‑
ture,where different subsystems are constructed to
process and integrate different sources[11].Zhu et al.
prented a goal‑driven nsor configuration.CPU
time,power,and weight are combined to reconfig‑
ure nsor suite and all chon measurements are in‑
tegrated using EKF[12].Although above rearch
has achieved satisfactory results,the filtering‑bad
methods have in common that they restrict the state
vector to the most recent state and marginalize out
all old information,which brings out suboptimal per‑
formance[13‑14].In contrast to filtering‑bad meth‑
ods,a graphical model known as factor graph repre‑
nts information fusion problem as a graph‑bad
nonlinear least squares optimization.It encodes the
connectivity between the unknown variable nodes
and the received measurements.Multinsory fusion
methods via factor graph can handle delayed and
asynchronous sources in a flexible way becau past
states are kept during the global optimization pro‑
cess[15].And it outperforms EKF becau of the
re‑linearization process[16].Chiu et al.propod a
constrained optimal lection for nsors bad on
factor graph and the optimal subts of nsors are
lected with available resources,navigation accura‑cy and obrvability index[17].Considering the re‑al‑time application,Merfels et al.propod a slid‑ing‑window factor graph method for autonomous ve‑hicles[18].Watson et al.evaluated the effectiveness of robust optimization techniques using the factor graph framework.It shows that the factor graph al‑gorithm in conjunction with robust optimization can achieve reasonable performance in the GNSS‑de‑graded environment[19].However,above rearch is still optimized with custom fusion solutions,which is inadequate for the flexible and extensible needs of land vehicles navigation system.
Seeking to develop adaptable navigation that al‑lows rapid integration of any combination of nsors to enable amless,robust and accurate navigation solutions in GNSS‑denied environment,we pro‑po a generic plug‑and‑play fusion strategy bad on factor graph for land vehicles.The strategy is de‑signed using abstraction method.Various abstract nsor models are designed by the type of nsors,rather than for a specific nsor.When a nsor is connected into the navigation system,the specific nsor model is built from the abstract model and its error registration is imple
mented.The propod strategy allows rapid reconfiguration of any combina‑tion of nsors.Also,its modularity enables the fu‑sion architecture to be flexible and extensible to new nsors and new capabilities.In addition,time‑de‑layed nsor data,which prents low‑quality char‑acteristics,can be procesd in a natural way bad on the improved factor graph,in which error param‑eters of nsors are also added into the graph model to conduct nsor online calibration.We evaluate performance of the propod strategy using a land vehicle equipped with heterogeneous nsors.It shows that our strategy can process low‑quality data in a plug‑and‑play and robust manner and its perfor‑mance outperforms the most common filter‑bad method.
1Generic Sensor Fusion Strategy The propod strategy is shown in Fig.1,whi‑ch consists of three parts,preprocessing layer,ab‑stracting layer and fusing layer.
198
No.2LAI Jizhou,et al.A Generic Plug‑and‑Play Navigation Fusion Strategy for Land Vehicles …1.1Preprocessing layer
In the preprocessing layer ,raw measurement
sources are procesd into usable navigation infor‑mation.When a nsor is connected into the sys‑tem ,it is recognized and corresponding ID is at‑tached into this source.Then ,data conversion is conducted according to specific nsor type.For ex‑ample ,images of camera are converted into po es‑timates.Considering that nsors are placed in dif‑ferent locations of a vehicle ,spatial parameters among different nsors obtained from an offline cal‑ibration are offt in space‑time alignment.Also ,time stamping is implemented in this step.Relative and absolute measurements are also aligned by trans‑formation between different frames.1.2
Abstracting layer
In the abstracting layer ,various abstract nsor models are designed according to the type of n‑sors.This layer consists of four abstract models ,that is ,dead reckoning model ,position model ,ve‑locity model ,and attitude model.The specific mod‑el of a nsor can be instantiated using its templates by identifying information ’s ID.Also ,nsor error registrations are conducted.For example ,a nsor ’s specific noi and error parameters are added into the built model.
Dead reckoning model reprents recursive n‑sors ,such as inertial or other dead reckoning n‑
sors.Its abstract model can be conceptually de‑scribed by following continuous nonlinear differen‑
tial equation
x =f DR (x ,α,Δ)
(1)
where x is the navigation state ,reprenting the ve‑hicle ’s position ,attitude and velocity ;Δthe incre‑ment of the vehicle measured by nsors and αthe calculated model of errors in nsors.Other models reprent nsors that provide with other measure‑ment information ,that is ,position ,velocity and at‑titude.Their abstract models can be described in a unified way
z =h M (x )+n
(2)
where x is navigation state ,reprenting the vehi‑cle ’s position ,attitude and velocity ;z the informa‑tion measured by nsors and n a measurement noi ,which is assumed to be zero mean Gaussian noi.h M is the measurement function ,relating be‑tween the measurement and navigation state.1.3
Fusing layer
In the fusing layer ,non‑linear optimization methods bad on factor graph is formulated.A fac‑tor graph is a bipartite graph G =(F ,
X ,E )with two types of nodes :Factor nodes f i ∈F and vari‑able nodes x i ∈X .Edges e ij ∈E can exist only be‑tween factor nodes and variable nodes ,and are pres‑ent if and only if the factor f i involves a variable x i .The factor graph G defines one factorization of the function f (X )as
f (X )=f i (X i )
(3)
where X i is the t of all variables x i connected by
an edge to factor f i
[20].A factor describes an error between the predict‑ed and actual measurements.Assuming a Gaussian noi model ,a measurement factor can be written as
f i (X i )=d [h i (X i )-z i ]
(4)
蜀漆where h i (X i )is the measurement model as a func‑tion of the state variables X i ;z i the actual measure‑ment and d (⋅)a cost function ,which is the squared Mahalanobis distance ,defined as d (e )≜e T Σ-1e ,with Σbeing the measurement covariance.Process models can be reprented using factors in a similar manner.
Eq.(3)should be minimized by adjusting the estimates of the variables X .The optimal estimate
is
Fig.1
Generic multi‑nsor fusion strategy
199
Vol.36
Transactions of Nanjing University of Aeronautics and Astronautics the one that minimizes the error of the entire graph [21]
X =arg min X
(∏i
f i
(X i
))
(5)
Different nsor information is added into the factor graph as variable and factor nodes.The time‑delayed and asynchronous measurements can be incorporated into the factor graph in a natural way ,leading to better estimates for current states.
2An Improved Sensor Fusion
如何祛眼袋Method for Land Vehicles Bad on Factor Graph
The structure of the improved multinsory fu‑sion method is shown in Fig.2.Bad on factor gra‑ph
framework ,nsor errors are added into the graph model to implement global optimization.The optimized error parameters are utilized to calibrate nsor measurements.Owing to nsor error online calibration ,better estimates for the whole trajectory can be obtained.
Considering that the most common nsors in typical navigation applications of land vehicles ,im ‑proved factor graph for land vehicles is built in Fig.3.The considered nsors are IMU ,GPS ,od‑ometer ,visual nsors ,and wireless nsors.In this paper ,GPS factor is built in the graph model to be adaptive to various applications.However ,GPS signal is not fud with other nsors but ud as ground truth in the field tests to prove the perfor‑mance of the propod algorithm in GNSS‑denied environment.
Sensors ’error parameters are added into graph
to implement global optimization.Black hollow cir‑cles mean navigation states and f IMU means IMU factor.Jasper hollow circles mean IMU bias ,which is introduced at a lower frequency than navigation states as it changes slowly during operation.Blue solid circles mean odometer factor while grey hol‑low circles reprent scale factor error of odometer.Red ,yellow and purple solid circles mean visual odometry ,wireless nsor ,and GPS factor ,respec‑tively.Green hollow circles reprent scale error of camera.Navigation states of land vehicles and error parameters of nsors are optimized together to im ‑prove estimation accuracy.Error parameters are ud to modify corresponding measurements.Sen‑sor factors are built as follows 2.1
IMU factor
IMU factor is built to connect navigation states at two quential times.Considering time k and time k +1,IMU factor is derived as
f IMU (x k +1,x k ,αk )≜d (x k +1-h (x k ,αk ,z k ))(6)
where x k +1and x k are navigation states at time
k +1and k ,respectively ;z k =[αk
ωk ]is the giv‑
en IMU measurements ,that is ,acceleration and an‑gular rate ;αk the bias of inertial nsor ,which is es‑timated to modify the IMU nsor data.The Euler integration prediction function with a noi is adopt‑ed to reprent h (⋅).In the same way ,bias factor can be described as
f bias (αk +1,αk )≜d (αk +1-
g (αk ))
酒店管理知识
泥鳅鱼(7)
where αk +1and αk are the bias at time k +1and k ,respectively.Bias is modelled as constant error.
2.2Odometer factor
Odometer provides with velocity information
and its factor can be reprented as
大赦f ODO (x k ,βk )≜d (z ODO k
-h ODO (x k ,βk ))
(8)
Fig.2
Structure of the improved fusion
method
Fig.3
Improved factor graph for land vehicles
200
No.2LAI Jizhou,et al.A Generic Plug‑and‑Play Navigation Fusion Strategy for Land Vehicles…
where z ODO k and x k are the velocities of odometer and navigation state at time k;βk is the scale factor er‑ror,which is obtained to modify odometer data.In the same way,scale factor error can be derived as
f scale(βk+1,βk)≜d(βk+1-g(βk))(9) whereβk+1andβk are the scale factor errors at time k+1and k,respectively.Scale factor error is mod‑elled as constant error.
2.3GPS factor
GPS factor is built to provide with absolute po‑sition and its factor can be modelled as
f GPS(x k)≜d(z GPS k-h GPS(x k))(10) where z GPS k and x k are the positions of GPS and navi‑gation state at time k.
2.4Wireless nsor factor
Wireless nsor provides ranging information to ba stations.When wireless nsor can receive at least three ranging information to ba stations who positions are obtained in advance,it can pro‑vide with position in the given frames and its factor can be modelled as
f WS(x k)≜d(z WS k-h WS(x k))(11) where z WS k and x k are the positions of wireless nsor and navigation state at time k.
2.5Visual nsor factor
Visual nsor provides with relative position when visual odometry algorithm is ud.After the relative and absolute measurements are aligned,it provides with po information in the global frame. Its factor can be reprented as
f VOP(x k)≜d(z VOP k-h VOP(x k,λk))(12)
f VOH(x k)≜d(z VOH k-h VOH(x k))(13) where z VOP k and x k are the position of visual nsor and navigation state at time k;z VOH k is the yaw of vi‑sual nsor at time k;λk the scale error and it is mod‑elled as constant error.Its factor can be reprented as
f scale(λk+1,λk)≜d(λk+1-g(λk))(14) whereλk+1andλk are the scale errors at time k+1 and k,respecti
vely.
3Experiment
In the field tests,we u a land vehicle equipped with a GPS receiver as well as IMU,ste‑reo camera,UWB(a kind of wireless nsor)and odometer.The land vehicle is shown in Fig.4.GPS receiver provides with preci positioning of centi‑meter‑level solutions when it operates in real‑time kinematic(RTK)mode,which is treated as ground truth.GPS is not integrated into the navigation sys‑tem,which only to evaluate the performance of the propod strategy in GNSS‑denied environment. Data acquisition module is designed bad on ROS.
The trajectory of the field test is shown in Fig.5with Google map.The starting point is mar‑ked with a star and arrows show the driving direc‑tion.A certain color of the trajectory means the cor‑responding ction where a certain combination of nsors is integrated into the navigation system,be‑cau some nsors are available in specific circum‑stances.For example,red line is surrounded by ba stations,and the UWB is available only in this part.Also,the roadway in blue part is the area where feature is spar,which leaves the camera in an unusable state and not be integrated into the navi‑gation system.In the test,different information sources are integrated to the system whenever they are available.
挺身而出的近义词
When a nsor is connected into system,spe‑cific models are constructed and corresponding fac‑tors are added into the factor graph.And time‑de‑layed and asynchronous measurements can be fud in the factor graph in a truly plug‑and‑play manner since past states are kept to perform global optimiza‑tion.
We compare our results with the most common filtering‑bad method,EKF.The drawback of
a Fig.4Land vehicle ud in the field test
201

本文发布于:2023-07-25 19:46:40,感谢您对本站的认可!

本文链接:https://www.wtabcd.cn/fanwen/fan/82/1116917.html

版权声明:本站内容均来自互联网,仅供演示用,请勿用于商业和其他非法用途。如果侵犯了您的权益请与我们联系,我们将在24小时内删除。

标签:知识   禁毁   小说   酒店   管理   泥鳅
相关文章
留言与评论(共有 0 条评论)
   
验证码:
推荐文章
排行榜
Copyright ©2019-2022 Comsenz Inc.Powered by © 专利检索| 网站地图