HiQuadLoc: An RSS-Based Indoor Localization System for High-Speed Quadrotors 1 Tuo Yu*, Yang Zhang*, Siyang Liu*, Xiaohua Tian*, Xinbing Wang*, Songwu.

Slides:



Advertisements
Similar presentations
Application a hybrid controller to a mobile robot J.-S Chiou, K. -Y. Wang,Simulation Modelling Pratice and Theory Vol. 16 pp (2008) Professor:
Advertisements

Fast Algorithms For Hierarchical Range Histogram Constructions
(Includes references to Brian Clipp
A KLT-Based Approach for Occlusion Handling in Human Tracking Chenyuan Zhang, Jiu Xu, Axel Beaugendre and Satoshi Goto 2012 Picture Coding Symposium.
A Platform for the Evaluation of Fingerprint Positioning Algorithms on Android Smartphones C. Laoudias, G.Constantinou, M. Constantinides, S. Nicolaou,
Accuracy Characterization for Metropolitan-scale Wi-Fi Localization Ying Wang, Xia Li Ying Wang, Xia Li.
A Novel Cluster-based Routing Protocol with Extending Lifetime for Wireless Sensor Networks Slides by Alex Papadimitriou.
Simple Linear Regression
1 ENHANCED RSSI-BASED HIGH ACCURACY REAL-TIME USER LOCATION TRACKING SYSTEM FOR INDOOR AND OUTDOOR ENVIRONMENTS Department of Computer Science and Information.
Robust Lane Detection and Tracking
Visual Odometry for Ground Vehicle Applications David Nister, Oleg Naroditsky, James Bergen Sarnoff Corporation, CN5300 Princeton, NJ CPSC 643, Presentation.
Novel Self-Configurable Positioning Technique for Multihop Wireless Networks Authors : Hongyi Wu Chong Wang Nian-Feng Tzeng IEEE/ACM TRANSACTIONS ON NETWORKING,
Department of Computer Engineering Koc University, Istanbul, Turkey
Shanshan Chen, Christopher L. Cunningham, John Lach UVA Center for Wireless Health University of Virginia BSN, 2011 Extracting Spatio-Temporal Information.
Simple Linear Regression. Introduction In Chapters 17 to 19, we examine the relationship between interval variables via a mathematical equation. The motivation.
Jason Li Jeremy Fowers Ground Target Following for Unmanned Aerial Vehicles.
EE392J Final Project, March 20, Multiple Camera Object Tracking Helmy Eltoukhy and Khaled Salama.
Rutgers: Gayathri Chandrasekaran, Tam Vu, Marco Gruteser, Rich Martin,
Kalman filter and SLAM problem
Applied Transportation Analysis ITS Application SCATS.
Introduction Due to the recent advances in smart grid as well as the increasing dissemination of smart meters, the electricity usage of every moment in.
Algorithm Taxonomy Thus far we have focused on:
Indoor Localization using Wireless LAN infrastructure Location Based Services Supervised by Prof. Dr. Amal Elnahas Presented by Ahmed Ali Sabbour.
Cut-And-Stitch: Efficient Parallel Learning of Linear Dynamical Systems on SMPs Lei Li Computer Science Department School of Computer Science Carnegie.
POWER CONTROL IN COGNITIVE RADIO SYSTEMS BASED ON SPECTRUM SENSING SIDE INFORMATION Karama Hamdi, Wei Zhang, and Khaled Ben Letaief The Hong Kong University.
Accuracy Evaluation of Stereo Vision Aided Inertial Navigation for Indoor Environments D. Grießbach, D. Baumbach, A. Börner, S. Zuev German Aerospace Center.
Segmental Hidden Markov Models with Random Effects for Waveform Modeling Author: Seyoung Kim & Padhraic Smyth Presentor: Lu Ren.
3D SLAM for Omni-directional Camera
Sérgio Ronaldo Barros dos Santos (ITA-Brazil)
The Scientific Method Honors Biology Laboratory Skills.
The Collocation of Measurement Points in Large Open Indoor Environment Kaikai Sheng, Zhicheng Gu, Xueyu Mao Xiaohua Tian, Weijie Wu, Xiaoying Gan Department.
视觉的三维运动理解 刘允才 上海交通大学 2002 年 11 月 16 日 Understanding 3D Motion from Images Yuncai Liu Shanghai Jiao Tong University November 16, 2002.
Complete Pose Determination for Low Altitude Unmanned Aerial Vehicle Using Stereo Vision Luke K. Wang, Shan-Chih Hsieh, Eden C.-W. Hsueh 1 Fei-Bin Hsaio.
Parameter/State Estimation and Trajectory Planning of the Skysails flying kite system Jesus Lago, Adrian Bürger, Florian Messerer, Michael Erhard Systems.
Maximum Network Lifetime in Wireless Sensor Networks with Adjustable Sensing Ranges Cardei, M.; Jie Wu; Mingming Lu; Pervaiz, M.O.; Wireless And Mobile.
On Distinguishing the Multiple Radio Paths in RSS-based Ranging Dian Zhang, Yunhuai Liu, Xiaonan Guo, Min Gao and Lionel M. Ni College of Software, Shenzhen.
Scaling Laws for Cognitive Radio Network with Heterogeneous Mobile Secondary Users Yingzhe Li, Xinbing Wang, Xiaohua Tian Department of Electronic Engineering.
Forward-Scan Sonar Tomographic Reconstruction PHD Filter Multiple Target Tracking Bayesian Multiple Target Tracking in Forward Scan Sonar.
The Application of The Improved Hybrid Ant Colony Algorithm in Vehicle Routing Optimization Problem International Conference on Future Computer and Communication,
Optimization of Wavelength Assignment for QoS Multicast in WDM Networks Xiao-Hua Jia, Ding-Zhu Du, Xiao-Dong Hu, Man-Kei Lee, and Jun Gu, IEEE TRANSACTIONS.
1 Virtual Patrol : A New Power Conservation Design for Surveillance Using Sensor Networks Prasant Mohapatra, Chao Gui Computer Science Dept. Univ. California,
December 9, 2014Computer Vision Lecture 23: Motion Analysis 1 Now we will talk about… Motion Analysis.
The Hardware Design of the Humanoid Robot RO-PE and the Self-localization Algorithm in RoboCup Tian Bo Control and Mechatronics Lab Mechanical Engineering.
A Passive Approach to Sensor Network Localization Rahul Biswas and Sebastian Thrun International Conference on Intelligent Robots and Systems 2004 Presented.
1 Research Question  Can a vision-based mobile robot  with limited computation and memory,  and rapidly varying camera positions,  operate autonomously.
CHAPTER 5 SIGNAL SPACE ANALYSIS
Real-Time Simultaneous Localization and Mapping with a Single Camera (Mono SLAM) Young Ki Baik Computer Vision Lab. Seoul National University.
1 Utilizing Shared Vehicle Trajectories for Data Forwarding in Vehicular Networks IEEE INFOCOM MINI-CONFERENCE Fulong Xu, Shuo Gu, Jaehoon Jeong, Yu Gu,
Spatiotemporal Saliency Map of a Video Sequence in FPGA hardware David Boland Acknowledgements: Professor Peter Cheung Mr Yang Liu.
Jin Yan Embedded and Pervasive Computing Center
Course14 Dynamic Vision. Biological vision can cope with changing world Moving and changing objects Change illumination Change View-point.
I Am the Antenna Accurate Outdoor AP Location Using Smartphones Zengbin Zhang†, Xia Zhou†, Weile Zhang†§, Yuanyang Zhang†, Gang Wang†, Ben Y. Zhao† and.
Visual Odometry for Ground Vehicle Applications David Nistér, Oleg Naroditsky, and James Bergen Sarnoff Corporation CN5300 Princeton, New Jersey
Web: ~ laoudias/pages/platform.htmlhttp://www2.ucy.ac.cy/ ~ laoudias/pages/platform.html
1 What happens to the location estimator if we minimize with a power other that 2? Robert J. Blodgett Statistic Seminar - March 13, 2008.
Global Clock Synchronization in Sensor Networks Qun Li, Member, IEEE, and Daniela Rus, Member, IEEE IEEE Transactions on Computers 2006 Chien-Ku Lai.
2010 IEEE Fifth International Conference on networking, Architecture and Storage (NAS), pp , 2010 作者: Filip Cuckov and Min Song 指導教授:許子衡 教授 報告學生:馬敏修.
Optic Flow QuadCopter Control
An Algorithm to Follow Arbitrarily Curved Paths Steven Kapturowski.
On Mobile Sink Node for Target Tracking in Wireless Sensor Networks Thanh Hai Trinh and Hee Yong Youn Pervasive Computing and Communications Workshops(PerComW'07)
Optimal Relay Placement for Indoor Sensor Networks Cuiyao Xue †, Yanmin Zhu †, Lei Ni †, Minglu Li †, Bo Li ‡ † Shanghai Jiao Tong University ‡ HK University.
Hybrid Indoor Positioning with Wi-Fi and Bluetooth: Architecture and Performance IEEE Mobile Data Management 2013 Artur Baniukevic†, Christian S. Jensen‡,
Smartphone-based Wi-Fi Pedestrian-Tracking System Tolerating the RSS Variance Problem Yungeun Kim, Hyojeong Shin, and Hojung Cha Yonsei University Bing.
Distributed Systems Lecture 5 Time and synchronization 1.
Sensor-Assisted Wi-Fi Indoor Location System for Adapting to Environmental Dynamics Yi-Chao Chen, Ji-Rung Chiang, Hao-hua Chu, Polly Huang, and Arvin Wen.
Vision Based Motion Estimation for UAV Landing
Subway Station Real-time Indoor Positioning System for Cell Phones
AirPlace Indoor Positioning Platform for Android Smartphones
Scalability of Wireless Fingerprinting based
Mole: Motion Leaks through Smartwatch Sensors
Presentation transcript:

HiQuadLoc: An RSS-Based Indoor Localization System for High-Speed Quadrotors 1 Tuo Yu*, Yang Zhang*, Siyang Liu*, Xiaohua Tian*, Xinbing Wang*, Songwu Lu** *Shanghai Jiao Tong University **University of California at Los Angeles

2 Outline Motivation System Architecture Working Process Localization Algorithm Measurement and Evaluation Conclusion 2

Motivation   In the field of indoor localization and navigation, since GPS is not available, how to locate fast-moving UAVs (Unmanned Aerial Vehicles) such as quadrotors is still a challenging topic.   Most of the previous works are based on vision or ultrasound detection. 3

Motivation   J. Eckert, R. German and F. Dressler, “On autonomous indoor flights: High-quality real-time localization using low- cost sensors,” in IEEE ICC, pp , Jun

Motivation   F. Kendoul, I. Fantoni and K. Nonami, “Optic flow-based vision system for autonomous 3D localization and control of small aerial vehicles,” in Robotics and Autonomous Systems, vol. 57(6), pp ,

Motivation   In the field of indoor localization and navigation, since GPS is not available, how to locate fast-moving UAVs such as quadrotors is still a challenging topic.   Most of the previous works are based on vision or ultrasound detection.   Additional infrastructures such as off-board sensors and cameras are still needed, which leads to extra cost and energy consumption.   We aim to apply the Wi-Fi fingerprint-based method, one of the most widely used technologies in indoor localization. 6

Motivation   The existing Wi-Fi RSS-based indoor localization systems cannot be directly applied to locate high-speed quadrotors for the following reasons: ① ① Flight speed impacts localization accuracy severely.   An RSS measurement will take at least 0.1s to 1s, during which the quadrotors (35km/h) would have moved for 1m to 10m. 7 Localization result MeasurementPosition

Motivation ② ② The workload of measuring all the RSS data in 3-D space is much higher than that in 2-D case.   Some technologies such as surface-based interpolation only record the average value of RSS at each calibration point, which loses the information included in the statistical features of RSS caused by the complex channel environment. 8

Motivation   We need to:   Estimate the real flight path of a quadrotor with the limited number of times for RSS measurement;   Consider the reduction in accuracy caused by the communication delay between the quadrotor and the server;   Estimate the probability distributions of RSS values at most cubes, instead of estimating the average values of RSS at these cubes only. 9

10 Outline Motivation System Architecture Working Process Localization Algorithm Measurement and Evaluation Conclusion 10

System Architecture   The system consists of quadrotor (smartphone) and localization server. 11

12 Outline Motivation System Architecture Working Process Off-line Stage On-line Stage Localization Algorithm Measurement and Evaluation Conclusion 12

Working Process: Off-line Stage ① ① Divide the localization region into cubes with constant size. ② ② Measure RSS at 1 of each 8 cubes only. ③ ③ Upload the data to the localization server. ④ ④ Process the data according to the 4-D RSS Interpolation Algorithm. 13

Working Process: Off-line Stage   4-D RSS Interpolation Algorithm   : a cube with coordinate.     We collect training data for 1 of each 8 cubes:   s = 1,…,M and M is the number of APs; N is the number of training data at each cube. 14

Working Process: Off-line Stage   The probability for to appear at is   : a cube in the fingerprint map generated by AP s.   Since p( ) is constant, let. It is continuous in the 4-D space   We have gotten the values when 15

Working Process: Off-line Stage   Thus we can use the cubic spline interpolation in the 4-D space to estimate when. 16 Before InterpolationAfter InterpolationReal Fingerprints

Working Process: On-line Stage ① ① The quadrotor sends a message to the server including the length of time slot T.   Note that T must be longer than the minimal RSS measurement time according to the quadrotor’s hardware performance. 17

Working Process: On-line Stage ① ① The quadrotor sends a message to the server including the length of time slot T. ② ② In each time slot, the quadrotor measures RSS, and sends a message including the data to the server. The message also contains:   The value of communication delay measured in the last localization process.   Whether the quadrotor is turning. 18

Working Process: On-line Stage ① ① The quadrotor sends a message to the server including the length of time slot T. ② ② In each time slot, the quadrotor measures RSS, and sends a message including the data to the server. The message also contains:   The value of communication delay measured in the last localization process.   Whether the quadrotor is turning. ③ ③ The localization server estimates the position of the quadrotor, and sends the result back. 19

Working Process: On-line Stage ① ① The quadrotor sends a message to the server including the length of time slot T. ② ② In each time slot, the quadrotor measures RSS, and sends a message including the data to the server. The message also contains:   The value of communication delay measured in the last localization process.   Whether the quadrotor is turning. ③ ③ The localization server estimates the position of the quadrotor, and sends the result back. ④ ④ The quadrotor calculates the time interval between the sending out of the message and the return of the result, and sets it as the new. 20

Working Process: On-line Stage   Turning Detection Using Direction Sensor   It is hard for the client to detect the flight direction of quadrotors because a quadrotor can make lateral movements without changing its head direction.   We notice that when a quadrotor is moving in a specific direction, its normal vector will have a drift angle for the same direction. 21

Working Process: On-line Stage   During a flight, the Turning Detector periodically measures XXXX processes them by low-pass filter, and calculates 22

Working Process: On-line Stage   If for continuous duration, the quadrotor is hovering, during which it may change its direction (turning starts).   Once for another, the turning ends. 23

Working Process: On-line Stage   Once and for more than, the quadrotor is turning. is the mean value for the recent values of α.   When and for more than, the turning is ended. 24

25 Outline Motivation System Architecture Working Process Localization Algorithm Preliminary Localization Algorithm Path Estimation Path Fitting Location Prediction Measurement and Evaluation Conclusion 25

Localization Algorithm 2626 preliminary localization algorithm path estimation path fitting location prediction

Preliminary Localization Algorithm   The algorithm is based on a frequently-used probabilistic model.   Find the largest in. Its corresponding denotes the estimated location for the quadrotor D RSS Interpolation Algorithm The result of each RSS measurement operated by the quadrotor

28 Outline Motivation System Architecture Working Process Localization Algorithm Preliminary Localization Algorithm Path Estimation Path Fitting Location Prediction Measurement and Evaluation Conclusion 28

Path Estimation   The path estimation method is based on Kalman Filter.   The state of the quadrotor in the time slot:   The motion model:   The preliminary localization model: 29

Path Estimation   Kalman Filter (process) 3030 Preliminary localization result Filtered Location

Path Estimation   Parameter Readjustment During Turning   One disadvantage of Kalman filter is that the localization accuracy at corners decreases obviously. 3131

Path Estimation   Parameter Readjustment During Turning   Once the server receives a turning-start signal, the following steps are executed:   Reducing leads to a higher weighting factor for the newest result. 3232

33 Outline Motivation System Architecture Working Process Localization Algorithm Preliminary Localization Algorithm Path Estimation Path Fitting Location Prediction Measurement and Evaluation Conclusion 33

Path Fitting   Kalman filter works well only when its parameters match with the real case, which can hardly be satisfied since the indoor environment varies.   The method of curve fitting is based on the assumption that a high-speed quadrotor tends to move in nearly straight lines. (The case of turning has been considered.)   The computation complexity of 3-D curve fitting is large and is severely impacted by the initial parameters of the curvilinear function.   We focus on the projected curve of the flight path on the 2- D ground. 34

Path Fitting   To avoid the case that the flight path is perpendicular to the X-axis or the Y -axis, we exchange the two axes and choose the fitting function with the maximum correlation coefficient. 35

Path Fitting   To avoid the case that the flight path is perpendicular to the X-axis or the Y-axis, we exchange the two axes and choose the fitting function with the maximum correlation coefficient. 36 Quadratic polynomial fitting

37 Outline Motivation System Architecture Working Process Localization Algorithm Preliminary Localization Algorithm Path Estimation Path Fitting Location Prediction Measurement and Evaluation Conclusion 37

Location Prediction   Due to the communication delay between the client and the server, there still exists significant localization error.   The delay has been uploaded.   contains the estimated velocity vector of the quadrotor.   Extend the motion curve of the quadrotor by   The server replies the client with the final localization result. 38

Location Prediction   Due to the communication delay between the client and the server, there still exists significant localization error.   The delay has been uploaded.   contains the estimated velocity vector of the quadrotor.   Extend the motion curve of the quadrotor by   The server replies the client with the final localization result.   Note that the computation complexity of the whole algorithm is only proportional to the times of RSS measurements in a path section, which are usually O(1). Thus the response time of the server is bounded. 39

40 Outline Motivation System Architecture Working Process Localization Algorithm Measurement and Evaluation Conclusion 40

Measurement and Evaluation   Evaluation of 4-D RSS Interpolation Algorithm   Space: ; AP: 4 E586Bs-2 at the four corners.   Full fingerprints: uses all the data collected from all the 405 cubes for localization.   Interpolated fingerprints: uses the data collected from 75 cubes for localization. 41   The interpolation reduces the accuracy of localization by 0.10m to 0.17m   The workload of fingerprint collection can be reduced by 81%.

Measurement and Evaluation 42

  Evaluation of Localization Algorithms   Compared with normal RSS-based systems, HiQuadLoc has reduced the location error by 62.8%. 43

Measurement and Evaluation   Evaluation of Parameter Readjustment During Turning   We change the values of respectively.   We analyze alone to rule out the additional gain of other methods.   We focus on the ±5 localization results around each corner. 44   When the error is minimum when XXXXXX respectively.   It is shown that changing the parameters of Kalman filter is necessary.

Measurement and Evaluation   Evaluation of HiQuadLoc for Different Flight Speeds   We control the quadrotor to fly in a straight line for different speeds: 3m/s, 2m/s and 1m/s. 45   3m/s: average error: 2.19m, reduced by 53.0%   2m/s: average error: 1.76m, reduced by 51.0%.   1m/s, average error: 0.89m, reduced by 66.4%.   Since the location error caused by delay is severer in the higher- speed case, the contribution of the location prediction method is more obvious than that in the lower-speed case.

46 Outline Motivation System Architecture Working Process Localization Algorithm Measurement and Evaluation Conclusion 46

Conclusion   Our contributions:   An RSS-based indoor localization system which can be applied on quadrotors moving at high speed.   The methods of path estimation, path fitting and location prediction to improve accuracy.   A 4-D RSS interpolation algorithm to reduce the workload by more than 80% during the offline data training phase.   The results of experiments show that HiQuadLoc reduces the average location error by more than 50% compared with normal RSS-based systems. 4747

Thank You