JP4032126B1 - Mobile object position estimation system and mobile object position estimation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a mobile object position estimation system and a mobile object position estimation method capable of estimating a searchee's position in consideration of a searcher's position and search history.
Using the position information of the first moving object, the probability that the first moving object exists at each position in the coordinate system of the map data is calculated, and distributed to the coordinate system of the map data according to the probability. The first sampling means for sampling the particles for the first moving body to be detected and the probability that the first moving body at the position specified by the position information of the second moving body is the first at the other position. The probability that the first moving body exists at each position in the coordinate system of the map data is calculated so as to be lower than the probability that the moving body of the first moving body exists, and the first sampling means sampled by the first sampling means Second sampling means for re-sampling the moving object particles according to the probability, and estimating the position of the first moving object from the re-sampled particles.
[Selection] Figure 3

Description

  The present invention relates to a mobile body position estimation system and a mobile body position estimation method for estimating the position of a mobile body using position information obtained from a mobile body position measurement apparatus.

In recent years, it has become possible to easily acquire the position of a moving body using a moving body position measuring device such as a GPS. However, the position information has low accuracy, and an error of several tens of meters to several hundreds of meters is generated in the moving body position measuring device alone. In order to reduce the uncertainty, a moving body position estimation system is used in which the position of the moving body is estimated by a computer based on position information obtained from the moving body position measuring device. The following Non-Patent Document 1 proposes a technique for probabilistically handling a position using a particle filter in a moving body position estimation system. The particle filter is a method of approximating the entire probability distribution by a weighted sum of a plurality of particles, and the position is estimated by a computer as follows. (1) N particles are generated at random positions on the coordinate system of the map data. (2) Randomly walk particles. (3) The position information from the moving body position measuring device is expressed as a probability distribution. (4) The occurrence probability of each particle position is obtained according to the probability distribution. (5) The weight of each particle is updated in proportion to the occurrence probability. (6) Estimating the probability distribution for the position of the moving object from the weight and position of each particle. (7) The average of the probability distribution is estimated as the position of the moving object.
J. et al. Highwater, B.W. Borriello. “The Location Stack: A Layer Model for Location in Ubiquitous Computing”, 4th IEEE Workshop on Modular Computing Systems & Applications 2 (WMC). 22-28.2002.

  The above-described conventional mobile body position estimation system is used for searching for a distress or an elderly person who hesitates. In this case, the search target person is searched for by carrying the search terminal with an information terminal or the like having a GPS function and estimating the position of the information terminal or the like. In such a search, not only the position information of the searchee but also the position of the searcher and the search history are important factors, such as a high probability that the searchee does not exist at the position where the searcher is present or at the searched position. Come. However, the conventional mobile body position estimation system only estimates the searchee's position based only on the searchee's position information, and does not consider the searcher's position or search history at all.

  In addition, since the above-described conventional moving body position estimation system obtains one probability distribution from particles and estimates the average as the position of the moving body, there is only one estimated position. However, when the error of the moving body position measuring device is large, there is a problem that the estimation probability becomes low because the estimation probability dispersion becomes large if the estimation is performed at one estimated position.

  Therefore, the present invention can estimate the position of the searchee in consideration of not only the position information of the searchee but also the searcher's position and search history, and furthermore, the position of the mobile object can be determined with high accuracy. An object of the present invention is to provide a mobile object position estimation system and a mobile object position estimation method that can be estimated.

In the moving body position estimation system / method of the present invention, when the position information of the moving body is acquired from the moving body position measuring device, the particles that sample the particles that randomly walk to the coordinate system of the map data using the position information of the moving body In a moving body position estimation system / method that has a filter and estimates the position of the moving body using the sampled particles,
When the particle filter obtains the position information of the first moving body from the moving body position measuring device, the first moving body is positioned at each position in the coordinate system of the map data using the position information of the first moving body. Sampling means / step for sampling the particles for the first moving object that are distributed in the coordinate system of the map data according to the calculated probability;
When the position information of the second moving body is acquired, the probability that the first moving body exists at the position specified by the position information of the second moving body is the first moving body at another position. The probability that the first moving object is present at each position in the coordinate system of the map data is calculated so as to be lower than the probability, and the first moving object sampled by the first sampling means / step is calculated. A second sampling means / step for resampling the particles for use according to the probability,
It is characterized by comprising position estimation means / step for estimating the position of the first moving body from the particles for the first moving body sampled by the second sampling means / step.

  Here, the position information of the first moving body is position information obtained from the moving body position measuring device as the position of the mobile body information terminal carried by the search target person. Further, the position information of the second mobile object is position information for specifying the position of the searcher at that time or the position where the searcher has already searched, and the position information of the mobile object information terminal carried by the searcher is moved. It may be obtained from a body position measuring device, or may be one or a plurality of position information input as a search history of a searcher or the like.

  The mobile object position estimation system / step of the present invention is used when a searcher searches for a search target person. The first mobile body is a searcher carrying the information terminal, and the second mobile body is a searcher. At the time of searching, since the probability that the searchee exists is low at the position where the searcher is present or the searched position, the probability that the searchee is present is lowered at these positions. That is, according to the present invention, the first sampling means / step calculates the probability that the first moving body exists at each position in the coordinate system of the map data using the position information of the first moving body, The particles for the first moving body are sampled according to the probability. Then, by the second sampling means / step, the probability that the first moving body exists at the position specified by the position information of the second moving body is the probability that the first moving body exists at another position. The probability that the first moving object is present at each position in the coordinate system of the map data is calculated so as to be lower, and the particles for the first moving object are resampled according to the probability. As a result, for the position where the searcher exists or the searched position, the probability that the searched person exists can be lowered, and the position estimation can be performed in consideration of the position of the searcher.

  The second sampling means / step determines, from the position information of the second moving body, each position of the coordinate system of the map data as the probability that the first moving body exists at each position of the map data coordinate system. It is preferable to calculate the probability that the second moving object exists and use the inverse of the calculated probability.

  According to this invention, the second sampling means / step calculates the probability that the second moving body exists from the acquired position information of the second moving body, and obtains the reciprocal of the calculated probability. Then, the inverse number is used as the probability that the first moving body exists, and the particles for the first moving body are resampled according to the inverse number. Thereby, the position of the searcher can be reflected in the estimation of the position of the searchee in more detail.

  Preferably, the position estimating means / step clusters the first moving body particles sampled by the second sampling means / step, and an average value of each cluster is set as an estimated position of the first moving body. . One or more estimated positions may be calculated.

  When the position of the moving body is estimated from all the first moving body particles, when the error of the position information from the moving body position measuring device is large, the variance of the estimation probability becomes large and the estimation accuracy becomes low. According to the present invention, the first moving body particles are clustered, and the average value of each cluster is used as the estimated position. Therefore, dispersion of the estimation probability in each cluster can be suppressed, and the estimation accuracy can be increased. .

  The mobile body position estimation system / method of the present invention calculates a patrol route capable of patrol the estimated position of the first mobile body calculated by the position estimation means / step from the start point when the start point is given. It is preferable to include a traveling route calculation means / step for

  According to the present invention, the traveling route calculation means / step calculates a traveling route that starts from the given start point and travels around the estimated position, so that the searcher travels the estimated position according to the calculated traveling route. The person to be searched can be searched for.

Ｒｏｕｔｅは評価値Ｖを求めようとする経路における推定位置ｒ（スタート地点を除く各地点）の集合であり、ｒはその推定位置の１つを示す。Ｔ（ｒ）は、推定位置ｒに到達するまでの１つ前の地点（スタート地点又は推定位置）からの時間を求める関数である。Ｐ（ｒ）は、推定位置ｒの１つ前の地点までの全ての推定位置における移動体が存在する確率の総和である。 It is preferable that the cyclic route calculation means / step obtains an evaluation position V according to the following equation 1 for each cyclic route, and ranks the cyclic routes according to the evaluation value V.

Route is a set of estimated positions r (points excluding the start point) on the route for which the evaluation value V is to be obtained, and r indicates one of the estimated positions. T (r) is a function for obtaining the time from the previous point (start point or estimated position) until the estimated position r is reached. P (r) is the sum of the probabilities that there are moving objects at all estimated positions up to the point immediately before the estimated position r.

  According to the present invention, the traveling route calculation means / step is necessary for traveling the estimated position and the probability that the first moving body exists at each estimated position by ranking the traveling routes according to the above equation (1). It is possible to rank the traveling routes from the two points of view of time. Thereby, it is possible to present to the searcher a patrol route that can search the search target person quickly and reliably.

  According to the moving body position estimation stem / method according to the present invention, the first sampling means / step causes the first moving body to be located at each position in the coordinate system of the map data using the position information of the first moving body. The probability of existence is calculated, and particles are sampled according to the probability. Thereafter, the second sampling means / step calculates the probability that the first moving body exists so that the position specified by the position information of the second moving body is lower than the other positions, The particles are resampled according to the probability. This makes it possible to reduce the probability that the searched person exists for the position where the searcher or the like exists or the searched position, and estimate the position of the searched person considering the searcher's position or search history. Is possible.

  The second sampling means / step calculates the probability that the second moving body exists at each position in the coordinate system of the map data from the position information of the second moving body, and the calculated probability Is used as the probability that the first moving object exists in the coordinate system of the map data. By using the reciprocal, the position of the searcher can be reflected in more detail in the position estimation of the searchee.

  Further, if the position estimating means / step clusters the first moving body particles and uses the average value for each cluster as the estimated position, it is possible to suppress the dispersion of the estimation probability and improve the estimation accuracy. . Also, a plurality of estimated positions can be output, and more information can be given to the searcher.

  Further, when the traveling route calculation means / step calculates a traveling route that circulates the estimated position of the searchee, the searcher can search for the searchee by traveling according to the calculated traveling route.

  Further, if the traveling route calculation means / step ranks the traveling routes according to the above equation 1, the two viewpoints of the probability that the search target person exists at each estimated position and the time required to travel the estimated position Therefore, it is possible to rank the tour routes in an order effective for the searcher to search for the search target person. By patrolling on a high-order patrol route, the searcher can efficiently circulate the estimated position, and the searchee can be found more quickly.

  The present invention can be applied to various fields. For example, a mobile object such as a portable information terminal with a GPS function is carried by a person and used to search for an elderly person or a victim who is hesitant, It can be used to grasp the position of the vehicle and rush to ensure safety in an emergency.

(First embodiment)
The moving body position estimation system 1 according to the present embodiment obtains position information of a moving body from a moving body position measurement device, and samples particles that are dispersed in a coordinate system of map data according to the position information of the moving body. This system has a filter and estimates the position of a moving object using particles sampled by a particle filter.

  FIG. 1 is a block diagram showing a configuration of a moving object position estimation system 1 according to the first embodiment of the present invention. The mobile body position estimation system 1 according to the present embodiment connects an internal bus 11 with a communication interface 12, a CPU 13, a ROM 14, a RAM 15, a display 16, a keyboard / mouse 17, a drive 18, and a hard disk 19, and an address signal and a control signal. The computer system is configured to transmit the data or the like and realize the moving body position estimation system 1 according to the present embodiment.

  The communication interface 12 has various communication functions for communication with a mobile body position measurement device and a communication network such as the Internet, and receives position information of the mobile body from the mobile body position measurement device. It is also possible to download a program 19a or map data 19c that causes a computer to execute such position estimation. The CPU 13 controls the entire apparatus by the OS stored in the ROM 14 and controls the function of executing processing based on various application programs 19 a stored in the hard disk 19.

  The ROM 14 stores a program for controlling the entire apparatus, such as an OS, and has a function of supplying these to the CPU 13. The RAM 15 has a memory function used as a work area when the CPU 13 executes various programs.

  The display 16 has a function of graphically displaying map data, probability data, and an estimated position. The keyboard / mass 17 can input data such as characters, numbers, symbols, etc., and can operate the point position on the screen.

  The drive 18 is a drive unit for executing installation work from a recording medium such as a CD or DVD in which various programs and data are recorded. It is also possible to install a program 19a or map data 19c that causes a computer to function as the moving body position estimation system 1 from a storage medium.

  The hard disk 19 is an external storage device that stores various data such as a program 19a, a memory 19b, and map data 19c. The program 19a corresponds to a program stored from the communication interface 12, the drive 18 and the like described above in the execution format. The memory 19b is a storage unit that stores files such as execution results of various programs.

  For example, as shown in FIG. 2, the map data 19c has a coordinate system expressed by longitude / latitude, buildings / roads / rivers, etc., regions such as prefectures, municipalities, addresses, etc., and basic information such as position, direction, distance, etc. Data D1 is included.

  The map data 19c may include probability data D2 indicating the probability that a moving object exists so as to correspond to each area of the basic data D1. In the probability data D2, for example, as shown in FIG. 2, the area is divided into categories such that the building is category a, the road is category b, the river is category c, and the others are category d. Is set. This probability is a probability that the first moving object to be estimated exists in the region. In the case of a terminal carried by a person as an estimation target, the probability of building category a and road category b is set high, and the probability of river category c is set low. Further, the probability may be set for each area, such as for each building or each road, without being divided into categories, or may be set for an artificially divided area such as an address. Thereby, the action pattern and action range peculiar to a moving body can be reflected in estimation. The probability data D2 may be integrated as a part of the map information 19c, or may be stored as separate data.

  The probability data D2 is generated by pre-processing performed before step S1, which will be described later, and is stored in the hard disk 19. The probability may be set in the system 1 in advance, or may be arbitrarily set or changed by the user. For example, the basic data D1 of the map data 19c may be displayed graphically, the probability and category may be preset by the user for each building, road, etc., and the probability data D2 may be generated by this presetting. In addition, when the basic data D1 includes identification information for identifying an area such as a building or a road (for example, when a building or a road is color-coded or marked), the mobile object position estimation system 1 is The area may be categorized using the identification information, and the category may be used in the probability data D2. This eliminates the need for categorization and makes it possible to easily categorize areas using the existing map data 19c.

  The particle in the present invention is an index indicating the position of the map data 19c on the coordinate system, and the attribute includes at least position information capable of specifying the position in the coordinate system of the map data 19c. The particle filter has at least a function of sampling particles according to the position information from a plurality of particles dispersed in the coordinate system of the map data 19c when the position information obtained from the moving body position measuring device is acquired. In the embodiment, at least the following steps S1 to S9 are provided.

  FIG. 3 is a flowchart for explaining a moving object position estimation method using the moving object position estimation system 1, and FIG. 4 is an explanatory view for conceptually explaining. The mobile object position estimation system 1 includes a random walk means, a first sampling means, a second sampling means, and a position estimation means, and each means functions as follows to realize the mobile object position estimation method of the present invention. .

  (Step S1) The random walk means arranges a plurality of particles for the first moving body at random positions in the first coordinate system of the map data 19c. The arrangement of the particles is performed by setting the position information of each particle at random. As shown in FIG. 4 (1), a plurality of particles for the first moving body (gray circles in FIG. 4 (1)) are arranged at random positions.

  (Step S2) The random walk means periodically moves (random walks) all the particles arranged in the first coordinate system of the map data 19c. When the map data 19c includes the probability data D2, all particles are randomly walked depending on the probability. Probability data D2 is not essential, but by using probability data D2, it is possible to reflect the probability that a moving object exists for each area such as a road, a building, or a river, and the position of the moving object can be estimated with higher accuracy. It becomes. The random walk of particles is performed by changing the position information of each particle. For example, as shown in FIG. 5, the probability data D2 is obtained by converting the latitude / longitude coordinate system of the basic data D1 into pixel units, and in each area divided into categories a, b, c, and d in the basic data D1. The probability is set with. The particle movable range e is set as a constant radius circle centered on the particle position, and when the particles are randomly walked, the probability data of all the pixels in the movable range e are acquired, The next position is determined in proportion to the probability, and the particle is moved to that position. The timing for moving the particles is set in the moving body position estimation system 1 in advance, and the particles are randomly walked according to the probability data D2 at that timing. FIG. 4B is an explanatory diagram conceptually illustrating a state when particles are randomly walked according to the probability data D2. Many particles are arranged in high probability areas such as buildings and roads.

  (Step S3) While the position information of the first moving body is not acquired from the moving body position measuring device, the process returns to Step S2 and continues the random walk. When the information on the position of the first moving body is acquired from the moving body position measuring device, the process proceeds to step S4.

  (Step S4) When the first sampling means acquires the position information of the first moving body, the first moving body exists at each position on the coordinate system of the map data 19c using the acquired position information. Probability is calculated. The calculation of the probability is performed, for example, by converting the acquired position information into a probability distribution P1 (x). Since the format of the position information from the moving body position measuring device differs depending on the characteristics of the device, it is necessary to convert it into a probability distribution P1 (x) corresponding to the device. For this reason, the conversion rule by numerical formula etc. is preset for every moving body position measuring apparatus in the moving body position estimation system 1, and the conversion from position information to probability distribution P1 (x) is performed according to the rule. For example, in the case of GPS, latitude / longitude information is output, so that the latitude / longitude information obtained is converted into a normal distribution with an average. In addition, the position of the PHS is specified depending on which base station of the base stations located in each place has received data from the terminal. For this reason, the probability that the mobile body exists is uniform within the range in which the base station that has received the radio wave from the mobile body can receive the radio wave, and the probability is 0 in the other ranges. Therefore, when the position information is acquired from the PHS, the probability that the moving body exists within a range in which the received base station can receive radio waves is a constant value larger than 0, and the probability is 0 in other areas. Convert to distribution.

  (Step S5) According to the calculated probability, the particles are sampled from the population of particles for the first moving object that randomly walks to the map data 19c. FIG. 4C is an explanatory diagram for conceptually explaining the sampling of the particles for the first moving body. In the sampling method, first, one particle is extracted from the population of particles for the first moving body according to the probability distribution P1 (x) calculated in step S4, stored, and returned to the population. Similarly, the next particle is extracted again. By repeating this, the same number of particles as the population of particles for the first moving object are extracted according to the probability distribution P1 (x). FIG. 6A is an explanatory diagram illustrating a probability distribution P1 (x) when using GPS position information, a population of particles that randomly walk to the coordinate system of the map data 19c, and sampled particles. It is. The rectangles arranged on the horizontal axis represent individual particles. Particles are extracted from the population according to a probability distribution P1 (x) with the acquired position information as an average. As for the particles, the portion where the value of the probability distribution P1 (x) is high is extracted more, and the portion where the particle is lower is more difficult to extract. In the case of the normal distribution, many particles are likely to be extracted closest to the position specified by the acquired position information of the first moving object (that is, the average value of the probability distribution P1 (x)), and are more difficult to be extracted as they are farther away.

  (Step S6) When the position information of the second moving body is acquired, the process proceeds to Step S7.

  (Step S7) The second sampling means is configured such that the probability that the first moving body exists at the position specified by the position information of the second moving body is the probability that the first moving body exists at another position. The probability that the first moving body is present at each position in the coordinate system of the map data 19c is calculated so as to be lower than that of the map data 19c.

  The probability that the first moving body exists at each position in the coordinate system of the map data 19c is calculated as follows. First, the acquired position information of the second moving body is converted into a probability distribution P2 (x). When the position information of the second moving body is acquired from the moving body position measuring device, it is converted into a probability distribution P2 (x). The conversion method is the same as in step S4. When a searcher or the like inputs the position information of the second moving body using an input means such as a keyboard, the input position information is acquired, and the second moving body exists at a position specified by the position information. The probability is greater than 0, and the other regions are converted into a probability distribution with a probability of 0. The input position information may be information on a position at which the searcher is at that time, or may be a plurality of position information that is a search history of the searcher.

  (Step S8) Next, using the probability that the second moving object exists, the probability that the first moving object exists at each position in the coordinate system of the map data 19c is calculated. Since there is a high probability that the searcher is not present at the place where the searcher is present during the search, the probability that the first mobile object exists at the position specified by the position information of the second mobile object is the first probability at the other position. It is made lower than the probability that one moving body exists. For example, the inverse P3 (x) of the calculated probability distribution P2 (x) is calculated.

  (Step S9) According to the reciprocal P3 (x) calculated in Step S8, the particles for the first moving body sampled in Step S5 are sampled again (resampling). The resampling method is the same as in step S5. FIG. 6B is an explanatory diagram for explaining the probability distribution P2 (x), its reciprocal P3 (x), and the resampled particles when the GPS position information is used. FIG. 4 (4) is an explanatory diagram for conceptually explaining the dispersion of the resampled particles according to the reciprocal P3 (x). When the probability distribution P2 (x) is a normal distribution, particles are less likely to be extracted as they are closest to the position specified by the acquired position information of the second moving body, and more likely to be extracted as they are farther away. As a result, for the position where the searcher is present or the searched position, the probability that the searchee exists can be lowered, and the position of the searcher can be reflected in the position estimation of the searchee.

(Step S10) The position estimation means clusters the particles for the first moving body resampled in Step S9. FIG. 4 (5) is an explanatory diagram conceptually illustrating particle clustering. Sampled particles are clustered into multiple clusters. Here, clustering according to the present embodiment will be described. For the clustering of the present embodiment, a clustering method using a normal mixture distribution P (x) is used. The normal mixture distribution P (x) is a probability distribution represented by Equation 2. Pi (x) in Equation 2 represents a normal distribution. The clustering process is to obtain the number c of normal distributions in the formula 2, the weight α _i of each normal distribution, and the average value and variance of each normal distribution Pi (x). In the present embodiment, these values are collectively called parameters of the normal mixture distribution P (x). Clusters generated by clustering correspond to individual normal distributions Pi (x) constituting normal mixed distribution P (x). When the parameters of the normal mixture distribution P (x) are obtained, it becomes possible to obtain probabilistically which cluster each particle belongs to according to Equation 2. Furthermore, when an arbitrary position in the coordinate system of the map data 19c is given as an input, the probability that a moving object exists at that position can be obtained.

  FIG. 7 shows an example of clustering using the normal mixture distribution P (x). The rectangles arranged on the horizontal axis represent individual particles. P (x) represents a normal mixture distribution. The normal mixture distribution P (x) obtained from each particle has a higher probability (a mountain is higher) as the particles are gathered. In this example, two clusters are generated.

That is, in step S10, the position estimating means operates as follows. Expression 2 is set in advance in the moving object position estimation system 1, and the position estimation means uses the information on the particles resampled in step S9, and each parameter number c in Expression 2 and the weight α _{i of} each normal distribution. , And the average value and variance of the normal distribution Pi (x). The weight α _{i of} each normal distribution Pi (x) and the average value and variance of the normal distribution Pi (x) are calculated based on the EM algorithm using clusters to which individual particles belong as hidden variables. The EM algorithm is a method widely known as a maximum likelihood estimation method in the case where there are hidden variables, and it has been proved that the solution converges by performing repeated calculations. On the other hand, the number c of the normal mixture distribution P (x) is incremented one by one until the log likelihood value of each particle in the normal mixture distribution P (x) shown in Equation 3 is decreased with the initial value set to 1. Ask. The average value of each calculated normal distribution Pi (x) becomes the coordinates (x, y) of the estimated position of the first moving body.

  Furthermore, the moving body position estimation system 1 may calculate the probability P (x) that the first moving body exists at each estimated position (x, y). In this case, the calculated parameters are substituted into Equation 2, and the coordinates (x, y) of the calculated estimated positions are used as the input information (x) of Equation 2 to calculate P (x).

  Note that the clustering method is not limited to the above, and may be non-hierarchical clustering that classifies a fixed number of clusters, or by combining clusters in order from the closest distance to each other. Hierarchical clustering such as a C-Means method for dividing into a large number of groups may be used, and the clustering method is not limited.

  (Step S11) The average value of each cluster calculated in step S10, that is, the average value of each normal distribution Pi (x) is output as the estimated position (x, y). FIG. 4 (5) is an explanatory view for conceptually explaining the average value of each cluster. If there are a plurality of calculated estimated positions, the probability may be displayed at each estimated position (x, y), or the estimated positions (x, y) may be ranked and output in descending order of probability. The output destination may be the display 16 or may be output from the communication interface 12 to an information terminal (for example, a mobile phone) that can be connected via a network. Also, all of the calculated plurality of estimated positions may be output, or the top several may be output in order of probability.

  Since position estimation is performed by clustering particles, dispersion of estimation probabilities in the cluster can be suppressed, and estimation accuracy can be increased. Moreover, compared with the prior art in which the estimated position is output only, the present invention can obtain a plurality of candidates as estimated positions. For example, when the user is searching for a moving body, it is possible to search for a plurality of candidate sites.

(Second Embodiment)
When the mobile body position estimation system 1 is used for searching for an elderly person, the searcher needs to go around a plurality of estimated positions until a search target person is found. Therefore, the mobile object position estimation system 1 may include a traveling route calculation unit that calculates a traveling route of a plurality of estimated positions. FIG. 8 is a flowchart for explaining the position estimation processing of the present embodiment, and FIG. 9 is an explanatory view for conceptually explaining. A description of the same parts as those in the first embodiment will be omitted.

  (Step S1) A plurality of particles for the first moving body and a plurality of particles for the second moving body are respectively arranged at random positions in the first coordinate system of the map data 19c. The arrangement of the particles is performed by setting the position information of each particle at random. As shown in FIG. 9 (1), a plurality of particles for the first moving body (gray circles in FIG. 9 (1)) and a plurality of particles for the second moving body (in FIG. 9 (1)). White circles) are arranged at random positions.

  (Step 2 to Step 11) Step 2 to Step 11 are the same as those in the first embodiment.

  (Step S12 to Step S13) The start point of the tour is input before Step S14 of the tour route calculation. The start point may be arbitrarily input by a searcher or the like, or position information obtained from the mobile body position measurement device using a function such as GPS of an information terminal carried by the searcher may be input. In the present embodiment, a case where the position information of the second moving body obtained from the moving body position measuring device is used as the start point will be described as an example (step S12 and step S13).

  (Step S12) The traveling route calculation means samples the second moving body particles distributed in the first coordinate system of the map data 19c based on the probability distribution P2 (x) obtained in Step S7. The sampling method is the same as the particle sampling method for the first moving body. FIG. 9 (4) is an explanatory view for conceptually explaining the sampling of particles for the second moving body. Particles close to the position information of the second moving body are extracted with a high probability.

  (Step S13) The average value of the extracted particles for the second moving object is calculated (FIG. 9 (5)). This average value is taken as the starting point.

  (Step S14) From the position information of the start point and the calculated plurality of estimated positions and the probability that the moving object exists at the estimated position, the traveling route calculating means has a high possibility that the searcher will reach the moving object. In addition, a route that can be visited in a short time is calculated.

Ｒｏｕｔｅは評価値Ｖを求めようとする経路における推定位置ｒ（スタート地点を除く各地点）の集合であり、ｒはその推定位置の１つを示す。Ｔ（ｒ）は、推定位置ｒに到達するまでの１つ前の地点（スタート地点又は推定位置）からの時間を求める関数である。Ｐ（ｒ）は、推定位置ｒの１つ前の地点までの全ての推定位置における移動体が存在する確率の総和である。評価値Ｖの値が小さいほど、効率的な巡回経路であるといえる。 For example, the traveling route calculation means calculates the traveling route as follows. First, a search is made for a route that circulates all estimated positions (that is, vertices of normal distributions constituting a normal mixture distribution representing the estimated positions) from the start point. If the number of normal distributions constituting the normal mixed distribution is C, the path is its factorial C! Only exist. The following equation 1 is used as a function for evaluating these paths.

Route is a set of estimated positions r (points excluding the start point) on the route for which the evaluation value V is to be obtained, and r indicates one of the estimated positions. T (r) is a function for obtaining the time from the previous point (start point or estimated position) until the estimated position r is reached. P (r) is the sum of the probabilities that there are moving objects at all estimated positions up to the point immediately before the estimated position r. It can be said that the smaller the evaluation value V is, the more efficient the traveling route is.

  FIG. 10 is a graph showing an example of Equation 1. In FIG. 10, the horizontal axis indicates time, and the vertical axis indicates the probability that the moving object does not exist (1-P (x) in Equation 3), that is, the probability that the searcher cannot find the search target. Here, the number of normal distributions constituting the normal mixed distribution (that is, the number of estimated positions) C = 3 will be described as an example. The following processing is performed for all routes. The time required to reach the first normal distribution vertex (estimated position) p1 from the start point s and the probability that a moving object cannot be found at the reached vertex (estimated position) p1 are calculated. Further, the time required to reach the next vertex (estimated position) p2 from the first vertex (estimated position) p1 and the probability that a moving object at the reached vertex p2 cannot be found are calculated. In this way, the time and score are calculated in order of the waypoints that circulate from the start point. Then, with the horizontal axis representing time and the vertical axis representing the probability that a moving body cannot be found, the calculated time of each vertex (estimated position) and the probability that there is no moving body are plotted, and the integrated value (the hatched portion in FIG. 10). The size of (area) is the evaluation value V. This makes it possible to evaluate in consideration of the probability that the first moving body can be found and the traveling time. When outputting the cyclic route, only the route with the low evaluation value V may be presented, or all the cyclic routes may be output in the order of the evaluation value.

  As an example, as shown in FIG. 11A, calculation of a cyclic route when three estimated positions a, b, and c are obtained as a result of mobile object position estimation will be described. The starting point is the coordinates (0, 0), the estimated position a is the coordinates (5, 0), b is the coordinates (4, 3), c is the position of the coordinates (0, 5), and the movement at each estimated position The probability that a body exists is 0.5 for a, 0.1 for b, and 0.4 for c. FIG. 11B shows the result of calculating the evaluation value V for all the round routes. The result that the route that goes around in order of abc is the highest evaluation value is obtained, and abc is output as an effective tour route.

  As a result, the searcher can obtain information on a route through which the plurality of output estimated positions can be efficiently circulated. For example, in searching for an elderly person, it is possible to find the elderly person more quickly and ensure safety by patroling efficiently.

  Note that the present invention is not limited to the above-described embodiment, and can be modified without departing from the present invention. For example, the present invention may be applied to a conventional technique for performing maximum likelihood estimation of parameters without using clustering for position estimation. In this case, the moving body position estimation system 1 performs processing as follows. FIG. 12 is a flowchart thereof. The mobile object position estimation system performs step 1 to step 9 above, then performs parameter maximum likelihood estimation from the sampling result to calculate the probability distribution, calculates the estimated position from the obtained probability distribution (step S10), The estimated position is output (step S11). The parameter maximum likelihood estimation and the calculation of the estimated position from the probability distribution may be performed according to the type of the position measuring device. For example, in the case of GPS, parameter maximum likelihood estimation is performed to calculate a normal distribution, and an average value of the normal distribution is output as an estimated position. In the case of PHS, in step S4, the probability that a moving body exists within a range in which the received base station can receive radio waves is set to a constant value (greater than 0) greater than 0, and the probability is 0 in other ranges. Is converted into uniform distribution data. Except for step 4, the process is the same for both PHS and GPS.

  In addition, the present invention can be applied to various fields. For example, a mobile object such as an information portable terminal with GPS function can be carried by a person to search for an elderly person or a distressed person who is hesitant, or used to ascertain the position of a child who is going to and from school to ensure safety. it can. In addition, if the moving body is mounted on an automobile or the like, it can be applied to the position management of a taxi or a delivery car.

It is a block diagram which shows one structure of the moving body position estimation system by the 1st Embodiment of this invention. It is explanatory drawing which illustrates map data notionally. It is a flowchart explaining the mobile body position estimation method using the mobile body position estimation system in the said embodiment. It is explanatory drawing which illustrates notionally the position estimation process of the said embodiment. It is explanatory drawing explaining the setting of probability data in the said embodiment. (A) is explanatory drawing explaining probability distribution P1 (x) when using the positional information on GPS, the population of the particles disperse | distributed on the coordinate system of map data, and the extracted particle. (B) is explanatory drawing explaining probability distribution P2 (x) when using the positional information on GPS, its reciprocal number P3 (x), and the resampled particle. It is explanatory drawing which shows the example which represents the probability distribution of the particle extracted by the particle filter by normal mixture distribution. It is a flowchart explaining the process of 2nd Embodiment. It is explanatory drawing which illustrates notionally the position estimation process of the said embodiment. It is explanatory drawing explaining the function used for calculation of the cyclic route of the said embodiment. It is explanatory drawing which shows the example of calculation of the cyclic route of the said embodiment. It is a flowchart explaining the process of other embodiment.

Explanation of symbols

1 Mobile object position estimation system 19c Map data D1 Basic data D2 Probability data

Claims (10)

When the position information of the moving object is acquired from the moving object position measuring device, it has a particle filter that samples particles that randomly walk to the coordinate system of the map data using the position information of the moving object, and uses the sampled particles In the mobile object position estimation system for estimating the position of the mobile object,
When the particle filter obtains the position information of the first moving body from the moving body position measuring device, the first moving body is positioned at each position in the coordinate system of the map data using the position information of the first moving body. A first sampling means for sampling the particles for the first moving object distributed in the coordinate system of the map data according to the calculated probability;
When the position information of the second moving body is acquired, the probability that the first moving body exists at the position specified by the position information of the second moving body is the first moving body at another position. The probability that the first moving body is present at each position in the coordinate system of the map data is calculated so as to be lower than the probability, and for the first moving body sampled by the first sampling means Second sampling means for resampling the particles according to the calculated probability,
A moving body position estimation system comprising position estimation means for estimating the position of the first moving body from the first moving body particles sampled by the second sampling means.   The second sampling means determines the probability that the first moving body exists at each position in the coordinate system of the map data from the position information of the second moving body to each position in the coordinate system of the map data. The mobile body position system according to claim 1, wherein a probability that two mobile bodies exist is calculated, and an inverse number of the calculated probability is used.   The position estimation unit clusters the first moving body particles sampled by the second sampling unit, and uses an average value of each cluster as an estimated position of the first moving body. The moving body position estimation system according to claim 1 or 2.   The circuit further comprises a traveling route calculation unit that calculates a traveling route that can travel from the starting point to the estimated position of the first moving body calculated by the position estimation unit when the starting point is given. 4. A moving body position estimation system according to 3. The said cyclic route calculation means calculates | requires the evaluation value V for each cyclic route using following Formula 1, and ranks each cyclic route based on the said evaluation value V, It is characterized by the above-mentioned. Mobile object position estimation system.

Route is a set of estimated positions r (points excluding the start point) on the route for which the evaluation value V is to be obtained, and r indicates one of the estimated positions. T (r) is a function for obtaining the time from the previous point (start point or estimated position) until the estimated position r is reached. P (r) is the sum of the probabilities that there are moving objects at all estimated positions up to the point immediately before the estimated position r. It has a particle filter that acquires the position information of the moving body from the moving body position measuring device, samples particles that randomly walk to the coordinate system of the map data using the position information of the moving body, and uses the sampled particles In the mobile object position estimation method for estimating the position of the mobile object,
When the particle filter obtains the position information of the first moving body from the moving body position measuring device, the first moving body is positioned at each position in the coordinate system of the map data using the position information of the first moving body. A first sampling step of sampling the particles for the first moving object distributed in the coordinate system of the map data according to the calculated probability,
When the position information of the second moving body is acquired, the probability that the first moving body exists at the position specified by the position information of the second moving body is the first moving body at another position. The probability that the first moving object is present at each position in the coordinate system of the map data is calculated so as to be lower than the probability, and for the first moving object sampled in the first sampling step. A second sampling step for resampling particles according to the probability,
A moving body position estimating method comprising a position estimating step of estimating the position of the first moving body from the first moving body particles sampled in the second sampling step.   In the second sampling step, as the probability that the first moving body exists in the coordinate system of the map data, the second movement from the position information of the second moving body to each position in the coordinate system of the map data. The mobile body position estimation method according to claim 6, wherein a probability that a body exists is calculated, and an inverse number of the calculated probability is used.   The said position estimation step clusters the 1st particle | grains for mobile bodies sampled by the said 2nd sampling step, The average value of each cluster is made into the estimated position of the said 1st mobile body. The moving body position estimation method according to claim 6 or 7.   The circuit further comprises a tour route calculation step of calculating a tour route that can travel from the start point to the estimated position of the first moving body calculated by the position estimation method when the start point is given. 8. A moving object position estimation method according to 8. The said rounding route calculation step calculates | requires the evaluation value V using the following function 1 for every cyclic route, and ranks each cyclic route based on the said evaluation value V, It is characterized by the above-mentioned. Mobile object position estimation method.

Route is a set of estimated positions r (points excluding the start point) on the route for which the evaluation value V is to be obtained, and r indicates one of the estimated positions. T (r) is a function for obtaining the time from the previous point (start point or estimated position) until the estimated position r is reached. P (r) is the sum of the probabilities that there are moving objects at all estimated positions up to the point immediately before the estimated position r.

Images