Position estimation in kalman filter
1 like128 views
The document discusses the use of Kalman filters for position estimation, particularly utilizing smartphone sensors for indoor positioning applications. It emphasizes the importance of precise indoor location tracking, given that people spend most of their time indoors, and outlines the methodology for processing data from smartphones like the Huawei P10. The study concludes with potential improvements for accuracy and the importance of various calibration methods.
Position estimation in kalman filter
- 1. 1 Position Estimation in Kalman Filter Presented By David Sestak, Zhang zhen bing, Ahmed Mohamed Reda & Mohamed Freeshah January, 2018
- 2. 2 Position Estimation in Kalman Filter Zhang zhen bing Presented By Ahmed Mohamed Reda Mohamed Ahmed FreeshahDavid Sestak
- 3. 3 ▪ Introduction ▪ Motivation ▪ Equipment ▪ Methodology ➢ Data source ➢ KF Processing ▪ Results Project Schedule
- 4. 4 ▪ Precise position estimation is one of the fundamental technical dependencies in our society today. It has influenced our lives in countless ways most people do not think about. ▪ It is used in a variety of applications including: ▪ GNSS (and all its varying applications) ▪ Self driving vehicles ▪ Mobile robots ▪ Internet of Things (IoT) and many others Introduction
- 5. 5 ▪ Indoor positioning has become an area of increasing interest utilizing the various signals that our devices use such as Bluetooth and WiFi. ▪ People spend about 80% of their time indoors, so it is of great significant to achieve precise indoor position estimation. ▪ This is why we are using a smartphone, a device everyone carries, for real time position estimation only using its inertial sensors. Motivation
- 6. 6 ▪ Smartphone is a versatile device with many sensors such as: MEMS, Camera, WiFi, Bluetooth, GPS, etc. , which provides a high potential for indoor position estimation for many users . ▪ Data source ▪ The HUAWEI P10 smartphone was utilized to acquire the original data, including: the value of the accelerometer, gyroscope, magnetometer ▪ From the accelerometer , we can get the value of acceleration along the X axis and Y axis. ▪ From the magnetometer and gyroscope , we can get the value of angle and angle rate along the way. Equipment
- 7. 7 HUAWEI P10 Sensors: ▪ Fingerprint Sensor, G-Sensor, Gyroscope Sensor, Compass, Ambient Light Sensor, Proximity Sensor, Hall Sensor Software: ▪ HIPE2.3 ▪ MATLAB Equipment
- 8. 8 ▪ The Kalman Filter (KF) is a time-varying linear optimal estimation algorithm , which is ideally suited for position estimation in modern multi-sensor systems. ▪ An important aspect of KF is its recursive nature. The position estimation is only based on the previously calculated state and current input. ▪ This makes it ideal for real time application ▪ KF puts more weight on higher certainty values improving its estimation precision. Why KF ?
- 9. 9 ▪ Our Observation Variables: ▪ ax = x-direction acceleration ▪ ay = y-direction acceleration ▪ θ = direction ▪ ω = angular rate ▪ These are the observation values that our smartphone inertial sensors provides. Methodology
- 10. 10 ▪ Kalman Filter Variables: ▪ A = system function matrix ▪ H = observation function matrix ▪ Q = system noise matrix for compensation ▪ R = observation noise matrix ▪ xk = system state variables ▪ zk = observation variables ▪ Pk = covariance matrix ▪ Note: ▪ - = predicted value ▪ ^ = estimated value Methodology
- 12. 12 Methodology 0 1 2 3 4 5 6 类别 1 类别 2 类别 3 类别 4
- 13. 13 ▪ KF Processing In our project: Assumption:(Gaussian distribution) First: define the state variables Second: confirm the system function according to the physic principle. Third: given the observation equation. Fourth: confirm the matrix of Q, R, and P according to the experience. Methodology
- 14. 14 Methodology K+1 K 0 0 t 0 0 0 0 0 0 0 0 t 0 0 0 0 0 0 1 0 t 0 0 0 0 0 0 1 0 t 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 t 0 0 0 0 0 0 0 1 θ ω θ ω = System function
- 15. 15 Methodology K+1 K+1 θ ω 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 θ ω = Observation function
- 16. 16 Methodology K+1 K+1 θ ω 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 θ ω = Observation function
- 17. 17 Methodology 0.3 0 0 0 0 0 0 0 0 0.3 0 0 0 0 0 0 0 0 0.01 0 0 0 0 0 0 0 0 0.01 0 0 0 0 0 0 0 0 0.01 0 0 0 0 0 0 0 0 0.01 0 0 0 0 0 0 0 0 16 0 0 0 0 0 0 0 0 16 Q = Covariance matrix of the system noise
- 18. 18 Methodology 0.1 0 0 0 0 0.1 0 0 0 0 36 0 0 0 0 36 R = Covariance matrix of the measurement noise
- 19. 19 Methodology 0.01 0 0 0 0 0 0 0.01 0 0 0 0 0 0 0.1 0 0 0 0 0 0 0.1 0 0 0 0 0 0 36 0 0 0 0 0 0 36 R = Covariance matrix of the measurement noise
- 20. 20 Methodology 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0.1 0 0 0 0 0 0 0 0 0.1 0 0 0 0 0 0 0 0 0.1 0 0 0 0 0 0 0 0 0.1 0 0 0 0 0 0 0 0 36 0 0 0 0 0 0 0 0 36 P = Error covariance matrix of the estimation of the state variables
- 21. 21 Observations acceleration (x) Estimated value Observed value
- 25. 25 Observations Velocity (x , y)
- 26. 26 Results Distances (x , y)
- 27. 27 Results Sys. Variables Variance
- 28. 28 Results Final Track Part of play ground in front of LIESMARS (Straight line = 100 m, Curve = 50 m)
- 29. 29 ▪ Errors: ▪ Unreasonable velocity increase ▪ Improvement used: ▪ Counteracting unreasonable velocity increase ▪ Ways to Improve: ▪ Ground Truth ▪ Inertial sensor position update Discussion
- 30. 30 Conclusion Availability Continuity Accuracy Integrity (Ways to Improve by four criteria)
- 31. 31 ▪ Leick, A. L. Rapoport, D. Tatarnikov., 2016, GPS Satellite Surveying, 4th Edition, New York: John Wiley and Sons ▪ G. Welch, G. Bishop., 2001, An Introduction to the Kalman Filter, University of North Carolina at Chapel Hill, Department of Computer Science, Chapel Hill, NC 27599-3175 ( http://www.cs.unc.edu ) ▪ Kalman, R.E. Trans ASME. 1960; 82:35–45. Crossref | Scopus (12465) ▪ http://consumer.huawei.com/en/phones/p10/specs/ Major References:
- 32. 32 Thank You !