An Adaptive Filter for a Small Attitude and Heading Reference System Using Low Cost Sensors

Similar documents
Automated Tuning of the Nonlinear Complementary Filter for an Attitude Heading Reference Observer

Adaptive Estimation of Measurement Bias in Six Degree of Freedom Inertial Measurement Units: Theory and Preliminary Simulation Evaluation

Attitude Estimation Version 1.0

RESEARCH ON AEROCRAFT ATTITUDE TESTING TECHNOLOGY BASED ON THE BP ANN

Application of state observers in attitude estimation using low-cost sensors

A Complementary Filter for Attitude Estimation of a Fixed-Wing UAV

Estimation and Control of a Quadrotor Attitude

EE565:Mobile Robotics Lecture 6

Inertial Odometry using AR Drone s IMU and calculating measurement s covariance

with Application to Autonomous Vehicles

Adaptive Unscented Kalman Filter with Multiple Fading Factors for Pico Satellite Attitude Estimation

Research Article Error Modeling, Calibration, and Nonlinear Interpolation Compensation Method of Ring Laser Gyroscope Inertial Navigation System

Tremor Detection for Accuracy Enhancement in Microsurgeries Using Inertial Sensor

Attitude Determination System of Small Satellite

CS491/691: Introduction to Aerial Robotics

Two dimensional rate gyro bias estimation for precise pitch and roll attitude determination utilizing a dual arc accelerometer array

Attitude determination method using single-antenna GPS, Gyro and Magnetometer

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 53, NO. 5, JUNE

IMU-Camera Calibration: Observability Analysis

Locating and supervising relief forces in buildings without the use of infrastructure

Quaternion based Extended Kalman Filter

UAV Navigation: Airborne Inertial SLAM

Autonomous Mobile Robot Design

Evaluation of different wind estimation methods in flight tests with a fixed-wing UAV

NONLINEAR ATTITUDE AND GYROSCOPE S BIAS ESTIMATION FOR A VTOL UAV

Fundamentals of attitude Estimation

VN-100 Velocity Compensation

Adaptive Kalman Filter for MEMS-IMU based Attitude Estimation under External Acceleration and Parsimonious use of Gyroscopes

Quadrotor Modeling and Control

The Research of Tight MINS/GPS Integrated navigation System Based Upon Date Fusion

Presenter: Siu Ho (4 th year, Doctor of Engineering) Other authors: Dr Andy Kerr, Dr Avril Thomson

System identification and sensor fusion in dynamical systems. Thomas Schön Division of Systems and Control, Uppsala University, Sweden.

Multi-layer Flight Control Synthesis and Analysis of a Small-scale UAV Helicopter

State Estimation for Autopilot Control of Small Unmanned Aerial Vehicles in Windy Conditions

1128 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 19, NO. 5, SEPTEMBER 2011

Attitude and Earth Velocity Estimation - Part I: Globally Exponentially Stable Observer

Chapter 4 State Estimation

Attitude Estimation for Augmented Reality with Smartphones

Discrete Time-Varying Attitude Complementary Filter

Adaptive Kalman Filter for Orientation Estimation in Micro-sensor Motion Capture

Discussions on multi-sensor Hidden Markov Model for human motion identification

EE 570: Location and Navigation

Smartphone sensor based orientation determination for indoor navigation

Unit quaternion observer based attitude stabilization of a rigid spacecraft without velocity measurement

Non-Drifting Limb Angle Measurement Relative to the Gravitational Vector During Dynamic Motions Using Accelerometers and Rate Gyros

Design of Adaptive Filtering Algorithm for Relative Navigation

Measurement Observers for Pose Estimation on SE(3)

HW VS SW SENSOR REDUNDANCY: FAULT DETECTION AND ISOLATION OBSERVER BASED APPROACHES FOR INERTIAL MEASUREMENT UNITS

A Close Examination of Multiple Model Adaptive Estimation Vs Extended Kalman Filter for Precision Attitude Determination

Integration of a strapdown gravimeter system in an Autonomous Underwater Vehicle

Attitude Determination for NPS Three-Axis Spacecraft Simulator

Adaptive Kalman Filter for MEMS-IMU based Attitude Estimation under External Acceleration and Parsimonious use of Gyroscopes

Tracking for VR and AR

Active Nonlinear Observers for Mobile Systems

TTK4190 Guidance and Control Exam Suggested Solution Spring 2011

Particle Filtering based Gyroscope Fault and Attitude Estimation with Uncertain Dynamics Fusing Camera Information

Research Article Design of an Attitude and Heading Reference System Based on Distributed Filtering for Small UAV

Attitude measurement system based on micro-silicon accelerometer array

On the Observability and Self-Calibration of Visual-Inertial Navigation Systems

Robust Attitude Estimation from Uncertain Observations of Inertial Sensors using Covariance Inflated Multiplicative Extended Kalman Filter

IMU Filter. Michael Asher Emmanuel Malikides November 5, 2011

Generalized mathematical model of a linear single-axis accelerometer as an integral part of the inclinometer

Sensors Fusion for Mobile Robotics localization. M. De Cecco - Robotics Perception and Action

COMBINED ADAPTIVE CONTROLLER FOR UAV GUIDANCE

Barometer-Aided Road Grade Estimation

ᓸᯏ㔚ㆇ ᗵ ℂ ᙥ ޕ ᑞ ᇷ ݾ ᔭ ஃ რऄתᖄ ៰ⷐ ᓸᯏ㔚ᛛⴚ ข ᓸᯏ㔚ㆇ ᗵ ℂ ᧂ น ਯ ᣂᙥ inemo ᘠᕈ㊂

Static temperature analysis and compensation of MEMS gyroscopes

The Fiber Optic Gyroscope a SAGNAC Interferometer for Inertial Sensor Applications

Design of Sliding Mode Attitude Control for Communication Spacecraft

Inertial Navigation and Various Applications of Inertial Data. Yongcai Wang. 9 November 2016

Sensors: a) Gyroscope. Micro Electro-Mechanical (MEM) Gyroscopes: (MEM) Gyroscopes. Needs:

Modeling Verticality Estimation During Locomotion

Motion Locus Analysis to Detect Rotation

Lecture. Aided INS EE 570: Location and Navigation. 1 Overview. 1.1 ECEF as and Example. 1.2 Inertial Measurements

Rao-Blackwellized Particle Filtering for 6-DOF Estimation of Attitude and Position via GPS and Inertial Sensors

Kalman Filter Enhancement for UAV Navigation

State observers for invariant dynamics on a Lie group

Fuzzy Adaptive Kalman Filtering for INS/GPS Data Fusion

Design and modelling of an airship station holding controller for low cost satellite operations

Research Article Relative Status Determination for Spacecraft Relative Motion Based on Dual Quaternion

Mathematical Modelling and Dynamics Analysis of Flat Multirotor Configurations

Simultaneous Adaptation of the Process and Measurement Noise Covariances for the UKF Applied to Nanosatellite Attitude Estimation

Calibration of a magnetometer in combination with inertial sensors

Vision and IMU Data Fusion: Closed-Form Determination of the Absolute Scale, Speed and Attitude

MEMS Gyroscope Control Systems for Direct Angle Measurements

Proprioceptive Navigation, Slip Estimation and Slip Control for Autonomous Wheeled Mobile Robots

MARINE biologists, oceanographers, and other ocean researchers

Model Reference Adaptive Control of Underwater Robotic Vehicle in Plane Motion

Hover Control for Helicopter Using Neural Network-Based Model Reference Adaptive Controller

VEHICLE WHEEL-GROUND CONTACT ANGLE ESTIMATION: WITH APPLICATION TO MOBILE ROBOT TRACTION CONTROL

Theory of Vibrations in Stewart Platforms

Adaptive Two-Stage EKF for INS-GPS Loosely Coupled System with Unknown Fault Bias

SOFTWARE ALGORITHMS FOR LOW-COST STRAPDOWN INERTIAL NAVIGATION SYSTEMS OF SMALL UAV

Nonlinear Landing Control for Quadrotor UAVs

In Use Parameter Estimation of Inertial Sensors by Detecting Multilevel Quasi-static States

Attitude Estimation and Control of VTOL UAVs

1 Kalman Filter Introduction

Angle estimation using gyros and accelerometers

Further results on global stabilization of the PVTOL aircraft

MODELING OF DUST DEVIL ON MARS AND FLIGHT SIMULATION OF MARS AIRPLANE

Transcription:

An Adaptive Filter for a Small Attitude and eading Reference System Using Low Cost Sensors Tongyue Gao *, Chuntao Shen, Zhenbang Gong, Jinjun Rao, and Jun Luo Department of Precision Mechanical Engineering of Shanghai University P.. Bo 8, Yanchang Rd 49, Shanghai 7, P.R. China gty@shu.edu.cn Abstract. Small Attitude And eading Reference System (ARS) based on MEMS are small size, light weight, low power consumption, and low cost by the inclusion of micro sensors. Small ARS have many potential uses beyond air- and spacecraft applications. owever, the MEMS sensors have large noise, bias and scale factor errors due to drift..an etended Kalman filter with adaptive gain was used to build a small ARS system based on a stochastic model. The adaptive filter tunes its gain automatically based on the system dynamitic sensed by the movement state to yield optimal performance. Keywords: adaptive Kalman filter, Attitude And eading Reference System, low cost sensors. Introduction Small Attitude And eading Reference System (ARS) based on Micro-Electro- Mechanical Systems (MEMS) are small size, light weight, low power consumption, and low cost by the inclusion of micro sensors. Small ARS have many potential uses beyond air- and spacecraft applications. But the MEMS sensors have large noise, bias and scale factor errors due to drift. The traditional algorithm using a low-cost MEMS sensor is difficultly satisfying the ARS performance reuirements. Recently, a lot of efforts have been directed at developing low-cost systems for ARS. Although, they effort to achieve the best performance of ARS. They all do not think about the state of carrier movement. In this paper, An etended Kalman filter with adaptive gain was used to build a small ARS system. The adaptive filter tunes its gain automatically based on the system dynamitic sensed by the movement state to yield optimal performance. System Design Small ARS system block diagram is figure. ARS consists of three parts. One part is sensors which are made up of triple ais gyroscopeâtriple ais accelerometerâtriple ais magnetometer and temperature sensor. The seconded part * Corresponding author. Y. Wu (Ed.): Advances in Computer, Communication, Control & Automation, LNEE, pp. 9. springerlink.com Springer-Verlag Berlin eidelberg

T. Gao et al. is MCU which realizes data AcuisitionÂSensor Calibration ÂKalman Algorithm and Communication. The third part is user interface. Fig.. The ARS system block diagram The Model of Sensors. The Mode of Gyroscope Gyroscope measures the rotational velocity. The angle of rotation could be obtained through integration of this sensor signal. owever, as integration is reuired, even the smallest constant bias can make the error grow to infinity. It is known as drift that is biggest error of rate gyro. In the following analysis we denote by ω : the biased compensated gyro measurements: ω ω ω () where, ω :the output of gyroscope,ω :the estimated bias drift by Kalman filter. Further, ω ω ω T ω ω ω T ω ω ω T (). The Mode of Accelerometer Accelerometer measures the sum of all inertial forces and gravity. But the high translational accelerations will happen only in transient modes and by small time intervals. So the accelerometer can be used as inclinometers:

An Adaptive Filter for a Small Attitude and eading Reference System A arcsina /g () A arcsina,a (4) where, A : the pitch angle by the data of accelerometer A : the roll angle by the data of accelerometer A, A,A : the output of accelerometer g:the gravity. The Mode of Magnetometer Magnetometer measures the sum of the earth magnetic and other stuff magnetic. When the values of other stuff magnetic surrounding the ARS is very large, the corresponding arithmetic will compensate the output of magnetometer. M arctany,x (5) where, Y : the level component of the earth magnetic X : the level component of the earth magnetic M : the yaw angle by the data of magnetometer When the ARS is inclining, M must be compensated by the pitch and roll angle: X Xcos Y sin cos Zcos cos (6) where, X,Y,Z: the output of the magnetometer. Y Y cos Zsin (7) 4 The Adaptive Kalman Filter 4. The Adaptive Kalman Filter Thought In this paper, by Kalman filter, the attitudeâheading and gyro bias is estimated. Furth more, the attitude is calculated by integrating angular rates measurements given by gyroscope. Accelerometers and magnetometer will work as observation data to correct the predicted attitudeâheading and gyro bias. A diagram of the Kalman filter is shown in Figure. Fig.. The diagram of Kalman Filter In order to obtain better accuracy, the adaptive gain was gotten by adjust observer noise matri R based on the movement state such as non-acceleration mode, acceleration mode and high dynamic mode. In one word, the adaptive filter tunes its gain automatically based on the system dynamitic sensed by the movement state to yield optimal filter performance.

4 T. Gao et al. 4. The Adaptive Kalman Filter Design In order to achieve the etended kalman filter on computer, the following parameter must be gotten: A. State transition Matri / The uaternion differential euations computation is relatively small, and there are no singularities. So we set up the Kalman state euation by uaternion. uaternion differential epression: 4 ω ω ω ω ω ω ω (8) ω ω ω ω ω ω 4 where, 4 T is uaternion. Based on the formulation () and (8), the state euation can be gotten: ω m ω my ω mz ω X AXWt m ω mz ω my 4 ω my ω mz ω m 4 ω mz ω my ω 4 m Wt (9) ω e ω ey ω ez Assume the state vector () The seven states are four uaternion elements and three bias errors for the gyroscopes. So the system state euation is: ω m ω my ω mz χ χ ω X AXWt m ω mz ω my χ χ χ χ ω my ω mz ω m χ χ χ ω mz ω my ω m χ χ χ Wt () χ χ χ The above euation is Kalman filter state euation. Above nonlinear state euation can be linearized and discrete along the currently estimated trajectory χ. So the state transition matri: ω m ω my ω mz χ χ ω m ω mz ω my χ χ 4 / ω my ω mz ω m χ χ ω mz ω my ω t () m χ 4 χ

An Adaptive Filter for a Small Attitude and eading Reference System 5 B. Measurement matri Such as above description, the observer vector is: Y A A M () Then the observer euation is Y XVt (4) Above nonlinear state euation can be linearized and discrete along the currently estimated trajectory χ. So the observer euation is A A A A A A A χ χ χ χ χ χ χ χ χ A A A A A A χ A Y Vt χ χ χ χ χ χ χ χ Vt χ M M M M M M M χ χ χ χ χ χ χ χ χ (5) According the relationship between the Euler angle and uaternion, the measurement matri can be gotten. Based on the information, the Direction Cosine Matri (DCM) is: cc sc s DCM sc css cc sss cs (6) sc csc cs ssc cc where, c and s are the abbreviation of cos and sin. uaternion also can be converted to DCM: 4 4 4 DCM 4 4 4 (7) 4 4 Based on the formulation (6) and(7 ),then 4 atan asin 4 4 atan 4 According to above relationship, the partial derivatives between Euler angle and uaternion in the measurement matri can be gotten. If we define the three variables: (9) 4 (8) () 4

6 T. Gao et al. 4 () Then measurement matri () C. Process Noise Matri Process noise matri () Each element on the diagonal corresponding to each state vector component, said the state vector components of the noise situation. Each element of process noise matri can be determined eperimentally using measurement data from the ARS sensors. D. Observer Noise Matri Adaptive Regulation Traditional Kalman filtering algorithm uses fied noise covariance matri, and can t obtain the optimal attitude angle in various conditions. This paper provides a kind of adaptive noise matri euation, namely adaptive unscented Kalman filter algorithm. In static state, observation noise covariance matri R k can be obtained through the analysis of observational data, and then determine the Kalman filter gain for optimal attitude estimation. In dynamic state, the measurements of the accelerometer not only include the acceleration of gravity but also the linear acceleration. Therefore, this paper proposed regulating the value of R according to the motion condition, and the gain of Kalman filter is adjusted to get the optimal attitude estimation. The state parameter is defined as follows: η A A y A z g (4) = = k 4 5 6 7 k =

An Adaptive Filter for a Small Attitude and eading Reference System 7 When ησ σ σ, the system is considered in static state. Where σ, σ and σ are the noise variance of the accelerometer in the stationary state. In stationary state the noise variance matri R is as follows: σ R σ (4) σ When σ σ σ ηt, the system is considered in low-dynamic state. Where Th is the threshold parameter of the state. In low-dynamic state the noise variance matri R is as follows: αη σ R αη σ (5) αη σ Where α is an adjustable scale factor. When ηt that the system is in high dynamic condition. In high dynamic state the noise variance matri R is as follows: T R T (6) T Where T ÂT and T are the threshold of the maimum observation noise. 5 Eperiment In order to verify the effect of adaptive Kalman filter algorithm proposed in this paper, the system is tested in three different conditions with turntable systems. The accelerometer attitude angle (AAA), estimated attitude angle (EAA) and real attitude angle (RAA) are compared respectively. The error is represented by the mean error (ME) and mean suare error (MSE). The test details are as follows: (a) static state (b) accelerated motion (c) high freuency motion Fig.. Filtering effect in stationary state

8 T. Gao et al. Figure (a) shows the effect of the filter that the sensor module is in static state. The optimal attitude angle has no drift, and the accuracy is greatly improved compared to only use the accelerometer. Table. Stationary state error error ME MSE Acceleration-based angle error.88.44 Filter-based angle error.98.8 Figure (b) shows the effect of the filter that sensor module doing accelerated motion in the horizontal plane. The accelerometer attitude angle is impact by the linear acceleration, while using the adaptive Kalman filter the error of the pitch angle is significantly improved. Table. Linear acceleration state error error ME MSE Acceleration-based angle error 4.67.678 Filter-based angle error. 54.85 Figure (c) shows the effect of the adaptive Kalman filter that the sensor module doing high freuency movement. The accuracy of the EAA angle is greatly improved compared to only use the accelerometer. Table 4. igh freuency motion state error error ME MSE Acceleration-based angle error 5.7.47 Filter-based angle error.57. 6 Conclusion ARS based on MEMS have many potential uses beyond air- and spacecraft applications. owever, the MEMS sensors have large noise, bias and scale factor errors due to drift. The adaptive Kalman filter is put forward. The adaptive filter tunes its gain automatically based on the system dynamitic sensed by the movement state to yield optimal performance. Finally the eperiment tests the good performance of the ARS based on the adaptive Kalman filter in three situations. In other word, he result proves the effectiveness of the adaptive Kalman filter using the low cost sensors. Acknowledgment. This project is supported by National Natural Science Foundation of China (No.595 and 558).

An Adaptive Filter for a Small Attitude and eading Reference System 9 References. Euston, M., Coote, P., Mahony, R., Kim, J., amel, T.: A Complementary Filter for Attitude Estimation of a Fied-Wing UAV with a Low-Cost IMU. In: 6th International Conference on Field and Service Robotics, FSR 7 (July 7). Kim., S., Park., M., Anumas., S., Yoo, J.: ead Mouse System Based on Gyro- and Opto- Sensors. In: IEEE International Conference on Biomedical Engineering and Informatics (). Bo, A., Borges, G.: Low Cost D Localization System for Applications on Aerial Robots. In: ABCM Symposium Series in Mechatronics, vol., pp. 55 6 (8) 4. Gao, T., Gong, Z., Luo, J., Ding, W., Feng, W.: An Attitude Determination System For A Small Unmanned elicopter Using Low-Cost Sensors. In: IEEE International Conference on Robotics and Biomimetics, Kunming, China, December 7- (6) 5. Mahony, R., amel, T., Pflimlin, J.: Complementary filter design on the special orthogonal group SO(). In: IEEE Conference on Decisoin and Control, pp. 477 484 (December 5) 6. Simon, D.: Optimal state estimation, pp. 95 457. John Wiley & Sons, Inc., oboken (6) 7. Woodman:, An introduction to inertial navigation University of Cambridge Technical Report, UCAM-CL-TR-696, ISSN 476-986 (August 7) 8. Gareth Evans, D., Drew, R., Blenkhorn, P.: Controlling Mouse Pointer Position Using an Infrared ead-operated Joystick. IEEE Transactions on Rehabilitation Engineering 8() (March )