CN107990900A - A kind of particle filter design methods of pedestrian's indoor positioning data - Google Patents
A kind of particle filter design methods of pedestrian's indoor positioning data Download PDFInfo
- Publication number
- CN107990900A CN107990900A CN201711194526.8A CN201711194526A CN107990900A CN 107990900 A CN107990900 A CN 107990900A CN 201711194526 A CN201711194526 A CN 201711194526A CN 107990900 A CN107990900 A CN 107990900A
- Authority
- CN
- China
- Prior art keywords
- mrow
- mtd
- msup
- mtr
- msub
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
- G01C21/206—Instruments for performing navigational calculations specially adapted for indoor navigation
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Automation & Control Theory (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Navigation (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention provides a kind of particle filter design methods of pedestrian's indoor positioning data, step includes:Obtain the real time positioning data of RSS alignment systems;Walking distance is calculated according to the gathered data of mems accelerometer;Designed for fusion mems accelerometer and the particle filter of the nonlinear transformations of building map.The particle filter design methods of pedestrian's indoor positioning data are according to RSS measurement data, mems accelerometer measurement data and cartographic information establish particle filter model, mems accelerometer and cartographic information are added in alignment system, walking distance is estimated based on the motion model of zero crossing algorithm, the nonlinear transformations from mems accelerometer and building map are integrated using particle filter, avoid the accumulated error as caused by sensor noise, it is significantly improved compared with Kalman filtering on the average value and standard deviation of evaluated error, and relative to the estimation result of wrong step-length it is in robustness.
Description
Technical field
The present invention relates to a kind of filter model design method, especially a kind of particle filter of pedestrian's indoor positioning data
Device design methods.
Background technology
Positioning and navigation system are in personnel safety, assets and personnel tracking, intelligently guiding, location aware multimedia service etc.
Huge success is achieved in extensive use based on location-based service (LBS).In general, these systems can be divided into three groups:Satellite
Alignment system, communications localization system and sensor alignment system.
For example famous GPS of global position system, the Big Dipper, Galileo system, it is primarily adapted for use in outdoor positioning.However, work as
When being positioned indoors under environment, satellite system is subject to the decay of building and wall, reflection and refraction.
Communications localization system utilizes existing communication network infrastructure, such as Wireless LAN (WLAN), ultra wide band (UWB)
Or DECT network, use received signal strength (RSS), arrival time (TOA), reaching time-difference (TDOA) and angle of arrival (AOA)
To calculate the position of user, the advantages of communications localization system is can to dispose that outdoor can also be deployed in indoors.It is in addition, logical
The communication network that letter alignment system uses need not additionally complete the construction of hardware, and cost of installation and maintenance is than relatively low.It is but logical
Letter alignment system is vulnerable to the puzzlement of the noise characteristic of wireless channel and multipath distortion, causes the reduction of positioning accuracy.
Sensor alignment system senses the relevant informations such as absolute position, such as magnetic transducing using various sensor specials
Device, laser sensor, ultrasonic wave, infrared sensor, inertial sensor etc., wherein inertial sensor are mainly used to sensing the feedback of position phase
Close the change of information.Since inertial sensor can only provide relative information, they usually require to be combined with other alignment systems
Complete positioning.For example, GPS/INS solutions correct the accumulated error of inertial sensor using GPS as manager, it is another
Aspect, inertial navigation system (INS) can also improve the performance of GPS, particularly such as in building or other cause gps signal
In the case of temporary block, traditional inertial navigation system is huge and costly, and which has limited they and indoor locating system
It is integrated.It is therefore desirable to design a kind of particle filter model, come handling to various location datas, to improve room
Interior positioning accuracy.
The content of the invention
The technical problem to be solved in the present invention is traditional inertial navigation system to limit they and indoor locating system
It is integrated, cause indoor position accuracy low.
In order to solve the above technical problem, the present invention provides a kind of particle filter model of pedestrian's indoor positioning data
Design method, includes the following steps:
Step 1, the real time positioning data of RSS alignment systems is obtained, specific calculation procedure is:
Step 1.1, in off-line case, the reception power vector of multiple access points at calibration point is measured, and will
Each measured value is recorded as each calibration point fingerprint;
Step 1.2, in on-line case, the power vector of reception and calibration point fingerprint are carried out using distance metric formula
Compare, select the distance average of the K calibration point close with receiving power vector as RSS measured values, distance metric formula
For:
In formula,For the measurement power of access point q, Pq(x) it is q-th of element of the fingerprint at calibration point x, Q is access
The quantity of point;
Step 1.3, RSS is calculated using the real coordinate position of each RSS measured values and each corresponding calibration point to determine
The real time positioning data of position system;
Step 2, calculating walking distance according to the gathered data of mems accelerometer is:
D (k)=Step_Size × Num_Steps (2)
In formula, Num_Steps counts step number using zero crossing algorithm, since the normal acceleration of mems accelerometer is believed
Number each step all will extend over zero curve twice, then the quantity of zero cross point divided by 2 be walking step number Num_Steps, step-length
The calculation formula of Step_Size is:
In formula, AmaxFor each step peak acceleration of pedestrian, AminFor each step minimum acceleration of pedestrian, C is steady state value, by
Obtained in the ambulation training data of different pedestrians;
Step 3, designed for fusion mems accelerometer and the particle filter of the nonlinear transformations of building map, tool
Body step is:
Step 3.1, the state value of Z (k) and the posterior probability density function pdf of x (k) are estimated using particle filter,
The formula of particle filter is:
In formula, xi(k) it is i-th of particle of posterior probability, wi(k) it is the weights of particle;
Step 3.2, each particle is obtainedNew particle xi(k+1), specific formula
For:
In formula, Δ t is the time that each particle derivation is new particle, if the movement of pedestrian is by passing through Gauss acceleration noise
The inertia of superposition dominates, i.e., a is sampled from normal distributionx(k) and ay(k), and assume that the RSS obtained by RSS alignment systems is measured
Value is the Gaussian Profile around actual position, then calculates new particle xi(k+1) weight is:
In formula, xRSSRefer to the field strength change of X-direction, yRSSRefer to the field strength change in y-axis direction;
Step 3.3, when using mems accelerometer, the walking distance between two RSS samples is calculated using formula (2)
D (k), since walking distance d (k) needs sigmoidal function, then the prediction samples equation used is:
In formula, di(k) from normal distributionIn sampled, be walking distance d (k) and standard deviationAverage value, θi(k) sampling is completed from equally distributed (0~2 π), formula (6) is reused and completes granular Weights Computing,
Based on formula (6) and (7), merge mems accelerometer using particle filter and RSS is measured;
Step 3.4, using particle filter fusion RSS, mems accelerometer and cartographic information, impossible grain is deleted
Son carrys out improved estimator, is by the particle filter model that equation (6) is improved to after optimization by weighting:
In formula, σaccFor standard deviation.
Further, in step 3.1, the execution step of particle filter is:
Initialization step, initializes pdf functions, using p (x (0)) to N number of particle { xi(0), i=1...N } adopted
Sample;
Prediction samples step, for each particle xi(k), from the state transition equation p (x of posterior probability density function pdf
(k+1)|xi(k)) new particle x is obtained ini(k+1);
Importance sampling step, for each new particle xi(k+1), w is calculatedi(k+1)=p (z (k+1) | xi(k+1));
Normalization and resampling steps, by weight normalization and final resampling, and are deleted in resampling steps low
The particle of weight, and repeat the particle with high weight so that each particle has identical weight.
The beneficial effects of the present invention are:According to RSS measurement data, mems accelerometer measurement data and cartographic information
Particle filter model is established, mems accelerometer and cartographic information are added in alignment system, based on zero crossing algorithm
Motion model estimates walking distance, is integrated using particle filter non-thread from mems accelerometer and building map
Property information, avoid the accumulated error as caused by sensor noise, the particle filter compared with Kalman filtering, estimation miss
It is significantly improved on the average value and standard deviation of difference, and is in robustness relative to the estimation result of wrong step-length.
Brief description of the drawings
Fig. 1 is flow chart of the method for the present invention;
Fig. 2 is straight line detour run trace schematic diagram;
Fig. 3 is variable velocity straight trip track schematic diagram;
Fig. 4 is 90 ° of steering run trace schematic diagrames;
Fig. 5 is 180 ° of steering run trace schematic diagrames;
Fig. 6 is 45 ° of steering run trace schematic diagrames;
Fig. 7 is one schematic diagram of annular run trace;
Fig. 8 is two schematic diagram of annular run trace;
Fig. 9 is random walk track schematic diagram;
Figure 10 is the average value schematic diagram of evaluated error in simulation;
Figure 11 is the standard deviation schematic diagram of evaluated error in simulation;
Mean error schematic diagram when Figure 12 is using different step-lengths;
Figure 13 is test result schematic diagram compared with simulation test result.
Embodiment
The embodiment of the present invention is described further below in conjunction with the accompanying drawings.
As shown in Figure 1, the present invention provides a kind of particle filter design methods of pedestrian's indoor positioning data, bag
Include following steps:
Step 1, mems accelerometer and cartographic information are increased in typical WLAN pedestrian's alignment system, based on RSS
It is pattern match or K- arest neighbors (KNN) algorithm that location algorithm is used in system, obtains the real-time positioning number of RSS alignment systems
According to specific calculation procedure is:
Step 1.1, in off-line case, the reception power vector of multiple access points at calibration point is measured, and will
Each measured value is recorded as each calibration point fingerprint;
Step 1.2, in on-line case, the power vector of reception and calibration point fingerprint are carried out using distance metric formula
Compare, select the distance average of the K calibration point close with receiving power vector as RSS measured values, distance metric formula
For:
In formula,For the measurement power of access point q, Pq(x) it is q-th of element of the fingerprint at calibration point x, Q is access
The quantity of point;
Step 1.3, RSS is calculated using the real coordinate position of each RSS measured values and each corresponding calibration point to determine
The real time positioning data of position system;
Step 2, calculating walking distance according to the gathered data of mems accelerometer is:
D (k)=Step_Size × Num_Steps (2)
In formula, Num_Steps counts step number using zero crossing algorithm, since the normal acceleration of mems accelerometer is believed
Number each step all will extend over zero curve twice, then the quantity of zero cross point divided by 2 be walking step number Num_Steps, step-length
The calculation formula of Step_Size is:
In formula, AmaxFor each step peak acceleration of pedestrian, AminFor each step minimum acceleration of pedestrian, C is steady state value, by
Obtained in the ambulation training data of different pedestrians;
Comparative maturity, miniaturization, cheap mems accelerometer are widely used current MEMS technology, MEMS acceleration
Meter is the device for measuring the acceleration of mobile object.In theory, a kind of method is by being integrated to acceleration signal
Calculate and obtain translational speed and distance, but for indoor walking, since acceleration is small, institute in this way the defects of be can hardly
By translational speed, distance and sensor noise, offset drifts and the phase separation of tilt variation data.Another method is detection walking
Distance, when people walk, its locomotory mechanism, normal acceleration periodically fluctuates, this cyclical signal represents people's step
Row distance, therefore, we obtain walking distance using definite pedestrian's step number and step number distance, such as formula (2) and formula (3) institute
Show.
Step 3, designed for fusion mems accelerometer and building map nonlinear transformations particle filter,
In indoor locating system, when pedestrian is by consecutive tracking, due to its mobility, the fluctuation of RSS measured values can be caused, this will lead
The reduction of location data precision is caused, by particle filter optimization smooth motion trajectories can be helped simultaneously to reduce evaluated error,
Frequency of use is higher for Kalman filter in target following, but is that its operation depends on line the shortcomings that Kalman filter
Property model, and linear model is substantially not present in practical application;Extended Kalman filter (EKF) and unlimited karr
Graceful wave filter (UKF) solves nonlinear estimation problem by the way that nonlinear model is converted into linear model, but EKF or UKF
For solving the problems, such as that part nonlinear system is also relatively difficult;Currently, since particle filter is to be based on Monte Carlo
Sampling, has and handles the non-linear and characteristic of non-gaussian estimation problem, thus using particle filter can with such as walking away from
Integrated, concretely comprised the following steps from all kinds of nonlinear transformations such as cartographic information:
Step 3.1, the state value of Z (k) and the posterior probability density function pdf of x (k) are estimated using particle filter,
The formula of particle filter is:
In formula, xi(k) it is i-th of particle of posterior probability, wi(k) step is performed for the weights of particle, the particle filter
It is rapid to be specially:
Initialization step, initializes pdf functions, using p (x (0)) to N number of particle { xi(0), i=1...N } adopted
Sample;
Prediction samples step, for each particle xi(k), from the state transition equation p (x of posterior probability density function pdf
(k+1)|xi(k)) new particle x is obtained ini(k+1);
Importance sampling step, for each new particle xi(k+1), w is calculatedi(k+1)=p (z (k+1) | xi(k+1));
Normalization and resampling steps, by weight normalization and final resampling, and are deleted in resampling steps low
The particle of weight, and repeat the particle with high weight so that each particle has identical weight;
View of the above, it will be seen that for the particle filter, pdf state transition equations p (x (k+1) | xi(k))
With pdf output density functions p (z (k+1) | xi(k+1)) should possess the function of being predicted sampling and weight calculation, herein
Pdf functions are not necessarily required to possess Gaussian characteristics;
Step 3.2, each particle is obtainedNew particle xi(k+1), specific formula
For:
In formula, Δ t is the time that each particle derivation is new particle, if the movement of pedestrian is by passing through Gauss acceleration noise
The inertia of superposition dominates, i.e., a is sampled from normal distributionx(k) and ay(k), and assume that the RSS obtained by RSS alignment systems is measured
Value is the Gaussian Profile around actual position, then calculates new particle xi(k+1) weight is:
In formula, xRSSRefer to the field strength change of X-direction, yRSSRefer to the field strength change in y-axis direction;
Step 3.3, when using mems accelerometer, the walking distance between two RSS samples is calculated using formula (2)
D (k), since walking distance d (k) needs sigmoidal function, then the prediction samples equation used is:
In formula, di(k) from normal distributionIn sampled, be walking distance d (k) and standard deviationAverage value, θi(k) sampling is completed from equally distributed (0~2 π), formula (6) is reused and completes granular Weights Computing,
Based on formula (6) and (7), merge mems accelerometer using particle filter and RSS is measured;
Step 3.4, using particle filter fusion RSS, mems accelerometer and cartographic information, impossible grain is deleted
Son carrys out improved estimator, is by the particle filter model that equation (6) is improved to after optimization by weighting:
In formula, σaccFor standard deviation.
Building map is another highly useful information source, because can be extracted from fabric structure information a large amount of
Location dependent data, such as the distance between floor, the position of wall, door or elevator, for tracking problem, which helps
In the uncertainty for reducing foot path.RSS, accelerometer and cartographic information can be merged using particle filter, by deleting
Except can not possibly particle carry out improved estimator (such as through the particle of wall), pass through weighting, it is proposed that particle filter Optimized model.
The experimental result of particle filter Optimized model and analysis:In order to preferably assess different filtering technique and particle
The performance of filter optimization model, we are tested in simulated environment and actual WLAN environment respectively.
First, simulation test
It has selected the wireless signal transmission model of multiple wall obstacles in emulation experiment environment, envisioned based on the pattern die
RSS in public room environmental is distributed, and the position of AP is marked with asterisk in the wireless signal transmission model, and reference point is with 1 meter of resolution
Rate uniform design, is used to carry out initial estimation first by KNN algorithms, then carrys out smooth track using different wave filters and subtract
Small site error.In order to improve the covering surface of test and accuracy, eight different foot paths are have selected in testing, such as scheme
Straight line detour shown in 2-9 is walked (Fig. 2), is kept straight on (Fig. 3) with variable velocity, is turned to and walked (Fig. 4) with 90 °, is turned to and gone with 180 °
(Fig. 5) is walked, is turned to and walked (Fig. 6) with 45 °, is walked in an annular (Fig. 7 and 8) and random walk (Fig. 9), estimates to miss in simulation
The average value and standard deviation of difference as shown in FIG. 10 and 11, algorithm parameter for details see attached table 1.
Indoor pedestrian's walk test result of the actual WLAN environment of table 1
Algorithm types | Mean error (m) | Standard deviation (m) |
KNN | 6.44 | 6.84 |
Kalman Filter(KF) | 5.81 | 4.07 |
Particle Filter(PF) | 5.57 | 3.9 |
PF+Accelerometer | 4.54 | 3.52 |
PF+Accelerometer+Map | 4.30 | 2.80 |
It was found from being analyzed from simulation result as shown in FIG. 10 and 11, when only considering RSS measurements, particle filter and card
Thalmann filter performance is suitable, and after walking distance information is combined, experimental result, which has, to be significantly improved;When only map is believed
When breath is added to RSS information, also there is certain improvement;Particle filter, accelerometer, cartographic information, walking distance etc. are added
The simulation result being added in same model averagely lifts more than 40% than KNN result, is averagely lifted than Kalman filtered results
30%.
In emulation experiment, we perform in the different step-lengths of Fig. 9, intend the step-size estimation of simulation mistake, and Figure 12 is provided
Mean error when using different step-lengths in fig.9, test result indicates that, the step error of 10-20% will not cause with
The deviation of best orientation precision is excessive.
2nd, indoor pedestrian's walk test of actual WLAN environment
Using all algorithms in real RSS and acceleration analysis verification text, test environment and the phase described in simulation
Together, foot path is identical with the track simulated in Fig. 9, and acceleration information is by motion model algorithm process, and identified step
Row distance is as shown in figure 12.
Indoor pedestrian's walk test result of actual WLAN environment is compared with simulation test result, such as Figure 13 institutes
Show, compared using filtering algorithm-Kalman filter or particle filter-arest neighbors positioning based on RSS, positioning accuracy can
To improve about 10%.When particulate filter obtains extraneous information from acceleration transducer and building map, can obtain about
25% further optimization lifting.
The particle filter design methods of pedestrian's indoor positioning data proposed by the present invention, according to RSS measurement data,
Mems accelerometer measurement data and cartographic information establish particle filter model, and MEMS acceleration is added in alignment system
Degree meter and cartographic information, estimate walking distance based on the motion model of zero crossing algorithm, are come using particle filter to integrate
From the nonlinear transformations of mems accelerometer and building map, the accumulated error as caused by sensor noise is avoided, the grain
Subfilter is significantly improved compared with Kalman filtering on the average value and standard deviation of evaluated error, and relative to mistake
The estimation result of step-length is in robustness by mistake.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on, should all be included in the protection scope of the present invention.
Claims (2)
1. a kind of particle filter design methods of pedestrian's indoor positioning data, it is characterised in that include the following steps:
Step 1, the real time positioning data of RSS alignment systems is obtained, specific calculation procedure is:
Step 1.1, in off-line case, the reception power vector of multiple access points at calibration point is measured, and will be each
Measured value is recorded as each calibration point fingerprint;
Step 1.2, in on-line case, the power vector of reception and calibration point fingerprint are compared using distance metric formula
Compared with the distance average of the selection K calibration point close with receiving power vector is as RSS measured values, distance metric formula:
<mrow>
<mi>d</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mi>Q</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>q</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>Q</mi>
</munderover>
<mo>{</mo>
<msup>
<mrow>
<mo>(</mo>
<msup>
<msub>
<mi>P</mi>
<mi>q</mi>
</msub>
<mi>m</mi>
</msup>
<mo>-</mo>
<msub>
<mi>P</mi>
<mi>q</mi>
</msub>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>}</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula,For the measurement power of access point q, Pq(x) it is q-th of element of the fingerprint at calibration point x, Q is access point
Quantity;
Step 1.3, RSS positioning system is calculated using the real coordinate position of each RSS measured values and each corresponding calibration point
The real time positioning data of system;
Step 2, calculating walking distance according to the gathered data of mems accelerometer is:
D (k)=Step_Size × Num_Steps (2)
In formula, Num_Steps counts step number using zero crossing algorithm, since the vertical acceleration signal of mems accelerometer is every
One step all will extend over zero curve twice, then the quantity of zero cross point divided by 2 be walking step number Num_Steps, step-length Step_
The calculation formula of Size is:
<mrow>
<mi>S</mi>
<mi>t</mi>
<mi>e</mi>
<mi>p</mi>
<mo>_</mo>
<mi>S</mi>
<mi>i</mi>
<mi>z</mi>
<mi>e</mi>
<mo>&ap;</mo>
<mroot>
<mrow>
<msub>
<mi>A</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>A</mi>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
</mrow>
<mn>4</mn>
</mroot>
<mo>&times;</mo>
<mi>C</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, AmaxFor each step peak acceleration of pedestrian, AminFor each step minimum acceleration of pedestrian, C is steady state value, by difference
Obtained in the ambulation training data of pedestrian;
Step 3, designed for fusion mems accelerometer and the particle filter of the nonlinear transformations of building map, specific step
Suddenly it is:
Step 3.1, the state value of particle filter estimation Z (k) and the posterior probability density function pdf of x (k), particle are utilized
The formula of wave filter is:
<mrow>
<mi>p</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mo>|</mo>
<mi>Z</mi>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>&ap;</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>N</mi>
</munderover>
<msup>
<mi>w</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mi>&delta;</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mo>-</mo>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, xi(k) it is i-th of particle of posterior probability, wi(k) it is the weights of particle;
Step 3.2, each particle is obtainedNew particle xi(k+1), specific formula is:
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mrow>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msup>
<mi>y</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<msup>
<mi>v</mi>
<mi>i</mi>
</msup>
<mi>x</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<msup>
<mi>v</mi>
<mi>i</mi>
</msup>
<mi>y</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mi>&Delta;</mi>
<mi>t</mi>
</mrow>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mi>&Delta;</mi>
<mi>t</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mrow>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msup>
<mi>y</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<msup>
<mi>v</mi>
<mi>i</mi>
</msup>
<mi>x</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<msup>
<mi>v</mi>
<mi>i</mi>
</msup>
<mi>y</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>+</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mfrac>
<mrow>
<msup>
<mi>&Delta;t</mi>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mfrac>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mfrac>
<mrow>
<msup>
<mi>&Delta;t</mi>
<mn>2</mn>
</msup>
</mrow>
<mn>2</mn>
</mfrac>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>&Delta;</mi>
<mi>t</mi>
</mrow>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mi>&Delta;</mi>
<mi>t</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>a</mi>
<mi>x</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>a</mi>
<mi>y</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, Δ t is the time that each particle derivation is new particle, if the movement of pedestrian by Gauss acceleration noise by being superimposed
Inertia dominate, i.e., a is sampled from normal distributionx(k) and ay(k), and assume that the RSS measured values that are obtained by RSS alignment systems are
Around the Gaussian Profile of actual position, then new particle x is calculatedi(k+1) weight is:
<mrow>
<msup>
<mi>w</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>p</mi>
<mrow>
<mo>(</mo>
<mi>z</mi>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
<mo>|</mo>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<msqrt>
<mrow>
<mn>2</mn>
<mi>&pi;</mi>
</mrow>
</msqrt>
<mi>&sigma;</mi>
</mrow>
</mfrac>
<msup>
<mi>e</mi>
<mrow>
<mo>&lsqb;</mo>
<mrow>
<mo>-</mo>
<mfrac>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>x</mi>
<mrow>
<mi>R</mi>
<mi>S</mi>
<mi>S</mi>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<msup>
<mrow>
<mo>(</mo>
<msup>
<mi>y</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>y</mi>
<mrow>
<mi>R</mi>
<mi>S</mi>
<mi>S</mi>
</mrow>
</msub>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
<mrow>
<mn>2</mn>
<msup>
<mi>&sigma;</mi>
<mn>2</mn>
</msup>
</mrow>
</mfrac>
</mrow>
<mo>&rsqb;</mo>
</mrow>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, xRSSRefer to the field strength change of X-direction, yRSSRefer to the field strength change in y-axis direction;
Step 3.3, when using mems accelerometer, the walking distance d between two RSS samples is calculated using formula (2)
(k), since walking distance d (k) needs sigmoidal function, then the prediction samples equation used is:
<mrow>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mtd>
</mtr>
<mtr>
<mtd>
<msup>
<mi>y</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msup>
<mi>x</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
<mo>+</mo>
<msup>
<mi>d</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<msup>
<mi>&theta;</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mtd>
</mtr>
<mtr>
<mtd>
<msup>
<mi>y</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
<mo>+</mo>
<msup>
<mi>d</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mi>s</mi>
<mi>i</mi>
<mi>n</mi>
<msup>
<mi>&theta;</mi>
<mi>i</mi>
</msup>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
In formula, di(k) from normal distributionIn sampled, be walking distance d (k) and standard deviation's
Average value, θi(k) sampling is completed from equally distributed (0~2 π), formula (6) is reused and completes granular Weights Computing, be based on
Formula (6) and (7), merge mems accelerometer using particle filter and RSS are measured;
Step 3.4, using particle filter fusion RSS, mems accelerometer and cartographic information, impossible particle is deleted
Improved estimator, is by the particle filter model that equation (6) is improved to after optimization by weighting:
In formula, σaccFor standard deviation.
2. the particle filter design methods of pedestrian's indoor positioning data according to claim 1, it is characterised in that
In step 3.1, the execution step of particle filter is:
Initialization step, initializes pdf functions, using p (x (0)) to N number of particle { xi(0), i=1...N } sampled;
Prediction samples step, for each particle xi(k), from the state transition equation p (x (k+1) of posterior probability density function pdf
|xi(k)) new particle x is obtained ini(k+1);
Importance sampling step, for each new particle xi(k+1), w is calculatedi(k+1)=p (z (k+1) | xi(k+1));
Normalization and resampling steps, by weight normalization and final resampling, and delete low weight in resampling steps
Particle, and repeat the particle with high weight so that each particle has identical weight.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711194526.8A CN107990900A (en) | 2017-11-24 | 2017-11-24 | A kind of particle filter design methods of pedestrian's indoor positioning data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711194526.8A CN107990900A (en) | 2017-11-24 | 2017-11-24 | A kind of particle filter design methods of pedestrian's indoor positioning data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107990900A true CN107990900A (en) | 2018-05-04 |
Family
ID=62032043
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711194526.8A Pending CN107990900A (en) | 2017-11-24 | 2017-11-24 | A kind of particle filter design methods of pedestrian's indoor positioning data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107990900A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109121080A (en) * | 2018-08-31 | 2019-01-01 | 北京邮电大学 | A kind of indoor orientation method, device, mobile terminal and storage medium |
CN110691326A (en) * | 2019-09-10 | 2020-01-14 | 东南大学 | Indoor hybrid positioning semi-physical simulation method and platform |
CN111256695A (en) * | 2020-01-14 | 2020-06-09 | 电子科技大学 | UWB/INS combined indoor positioning method based on particle filter algorithm |
CN111578938A (en) * | 2019-02-19 | 2020-08-25 | 珠海格力电器股份有限公司 | Target object positioning method and device |
CN112729301A (en) * | 2020-12-10 | 2021-04-30 | 深圳大学 | Indoor positioning method based on multi-source data fusion |
CN113534222A (en) * | 2020-04-17 | 2021-10-22 | 宝马股份公司 | Method for vehicle positioning, device for vehicle positioning and vehicle |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107339989A (en) * | 2017-06-23 | 2017-11-10 | 江苏信息职业技术学院 | A kind of pedestrian's indoor orientation method based on particle filter |
-
2017
- 2017-11-24 CN CN201711194526.8A patent/CN107990900A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107339989A (en) * | 2017-06-23 | 2017-11-10 | 江苏信息职业技术学院 | A kind of pedestrian's indoor orientation method based on particle filter |
Non-Patent Citations (1)
Title |
---|
WANG HUI,ETC: "WLAN-Based pedestrian tracking using particle filters and low-cost MEMS sensors", 《WPNC"07: 4TH WORKSHOP ON POSITIONING NAVIGATION AND COMMUNICATION 2007, WORKSHOP PROCEEDINGS》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109121080A (en) * | 2018-08-31 | 2019-01-01 | 北京邮电大学 | A kind of indoor orientation method, device, mobile terminal and storage medium |
CN109121080B (en) * | 2018-08-31 | 2020-04-17 | 北京邮电大学 | Indoor positioning method and device, mobile terminal and storage medium |
CN111578938A (en) * | 2019-02-19 | 2020-08-25 | 珠海格力电器股份有限公司 | Target object positioning method and device |
CN110691326A (en) * | 2019-09-10 | 2020-01-14 | 东南大学 | Indoor hybrid positioning semi-physical simulation method and platform |
CN111256695A (en) * | 2020-01-14 | 2020-06-09 | 电子科技大学 | UWB/INS combined indoor positioning method based on particle filter algorithm |
CN113534222A (en) * | 2020-04-17 | 2021-10-22 | 宝马股份公司 | Method for vehicle positioning, device for vehicle positioning and vehicle |
CN112729301A (en) * | 2020-12-10 | 2021-04-30 | 深圳大学 | Indoor positioning method based on multi-source data fusion |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107990900A (en) | A kind of particle filter design methods of pedestrian's indoor positioning data | |
CN107339989A (en) | A kind of pedestrian's indoor orientation method based on particle filter | |
CN107289941B (en) | Inertial navigation-based indoor positioning method and device | |
EP2673593B1 (en) | Method and device for indoor positioning | |
Wang et al. | WLAN-based pedestrian tracking using particle filters and low-cost MEMS sensors | |
Kim et al. | Indoor positioning system using geomagnetic anomalies for smartphones | |
CN104977006A (en) | Indoor positioning method based on fuzzy theory and multi-sensor fusion | |
CN104215238A (en) | Indoor positioning method of intelligent mobile phone | |
CN105263113A (en) | Wi-Fi location fingerprint map building method and system based on crowd-sourcing | |
Liu et al. | Indoor localization using smartphone inertial sensors | |
CN105547305A (en) | Pose solving method based on wireless positioning and laser map matching | |
Suski et al. | Using a map of measurement noise to improve UWB indoor position tracking | |
CN107014375B (en) | Indoor positioning system and method with ultra-low deployment | |
CN104075711B (en) | A kind of IMU/Wi Fi signal tight integration indoor navigation methods based on CKF | |
Marquez et al. | Accurate UWB and IMU based indoor localization for autonomous robots | |
CN102288938B (en) | Effective three-dimensional positioning method for wireless sensor network node | |
CN104394588A (en) | Indoor positioning method based on Wi-Fi fingerprints and multi-dimensional scaling analysis | |
CN106643711A (en) | Indoor positioning method and indoor positioning system based on hand-held device | |
KR101264306B1 (en) | Apparatus of tracking user indoor using user motion model learning and recording media therefor | |
CN104202818A (en) | Floor identification method based on building open edge distance weighting | |
CN106197418A (en) | The indoor orientation method that a kind of fingerprinting based on sliding window merges with sensor | |
Jose et al. | Taylor series method in TDOA approach for indoor positioning system. | |
Umetani et al. | Probabilistic localization of mobile wireless LAN client in multistory building based on sparse Bayesian learning | |
Lategahn et al. | Robust pedestrian localization in indoor environments with an IMU aided TDoA system | |
CN108332749B (en) | Indoor dynamic tracking and positioning method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180504 |