JP6438354B2 - Self-position estimation apparatus and mobile body equipped with self-position estimation apparatus - Google Patents

Description

  The technology disclosed in this specification relates to a moving object. Specifically, the present invention relates to a technique for estimating the position of a moving object.

  2. Description of the Related Art A moving body that moves within a predetermined moving area and that estimates its own position within the moving area is known. For example, in the moving body of Non-Patent Document 1, the weighted average value of self-position estimation by Monte Carlo Localization (hereinafter referred to as MCL) and self-position estimation by a vision sensor is used as the final self-position. .

Daichi Yoshikawa, Masaru Adachi, Kenichi Yasuda, “Robust Self-Location Estimation Using Multiple Sensors for Indoor Robots”, 31st Annual Conference of the Robotics Society of Japan, September 4-6, 2013

  In the technique of Non-Patent Document 1, the weighting coefficient used for self-position estimation is changed depending on whether or not the dispersion of the MCL particle set exceeds a threshold value. Here, the degree of dispersion of the MCL particle set varies depending on the moving region in which the moving body moves. For this reason, if the movement region changes, the threshold value must be changed accordingly. In the technique of Non-Patent Document 1, since the threshold value is changed based on the experience and intuition of the setter, there is a problem that the variation is large and the self-position estimation accuracy is not stable.

  In the present specification, a technique capable of stabilizing the self-position estimation accuracy of a moving object as compared with the conventional technique is disclosed.

  The present specification discloses a self-position estimation apparatus that estimates the final self-position of a moving body that moves within a predetermined movement area for each step. The self-position estimation device includes a first self-position estimation unit, a second self-position estimation unit, an abnormality occurrence probability calculation formula storage unit, an abnormality occurrence probability calculation unit, and a final self-position estimation unit. The first self-position estimation unit includes an environment map of the moving area, “a distance from the moving object to an object existing in the moving area” and “an orientation of the object with respect to the moving object” observed in the current step, Is used to update the probability distribution of the state of the moving object to the latest state, and the first self-position is estimated based on the probability distribution of the latest state. The second self-position estimating unit obtains the moving body from the previous step to the current step acquired by odometry at the final self-position of the moving body in the previous step estimated by the final self-position estimating unit. The second self-position is estimated by adding the movement distance and the movement direction. The abnormality occurrence probability calculation formula storage unit calculates an abnormality occurrence probability obtained by machine learning of the learning data acquired when the first self-position estimation unit estimates the first self-position during the learning movement of the moving body. Store the expression. The abnormality occurrence probability calculation unit inputs a plurality of variables acquired when the first self-position estimation unit estimates the first self-position during the main movement of the moving object to the abnormality occurrence probability calculation formula, and the abnormality occurrence probability Calculate The final self-position estimation unit calculates a weighted average value calculated using the first self-position acquired from the first self-position estimation unit and the second self-position acquired from the second self-position estimation unit, The final self-position in this step. The learning data includes a plurality of variables acquired when the first self-position estimation unit estimates the first self-position during learning movement, and a classification when the first self-position estimation result at that time is classified as normal or abnormal It has a plurality of data that associates the result. The weighting coefficient used in the final self-position estimation unit is a function of the abnormality occurrence probability acquired by the abnormality occurrence probability calculation unit.

  In the above self-position estimation apparatus, the weighting factor used when estimating the final self-position of the moving object is a function of the probability of occurrence of abnormality. The abnormality occurrence probability is calculated from an abnormality occurrence probability calculation formula. The abnormality occurrence probability calculation formula is obtained by machine learning of learning data acquired when the moving body learns and moves. That is, the abnormality occurrence probability calculation formula is generated by statistically analyzing the learning data. For this reason, since the value of a weighting coefficient is not determined based on experience and intuition, the estimation accuracy of the final self-position of the moving body can be stabilized.

  Details of the technology disclosed in this specification and further improvements will be described in detail in the detailed description and examples.

The system configuration of a mobile object is shown. A system configuration when the mobile body performs the final self-position estimation processing and the like is shown. 1 shows a system configuration when a moving body performs learning data generation processing. An example of the movement area | region where a moving body learns and moves is shown. An example of learning data is shown. The system configuration | structure when a mobile body implements abnormality-occurrence probability calculation formula generation processing is shown. The graph of abnormality probability calculation formula is shown. The graph of the weighting coefficient depending on abnormality occurrence probability is shown.

  The main features of the embodiments described below are listed. The technical elements described below are independent technical elements and exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. Absent.

(Characteristic 1) In the self-position estimation apparatus for a moving body disclosed in this specification, the weighting coefficient of the first self-position decreases and the weighting coefficient of the second self-position increases as the abnormality occurrence probability increases. Also good. According to the above configuration, when the probability of occurrence of abnormality is high (that is, when the estimation accuracy of the first self-position is low), the weighting coefficient of the second self-position increases, so that the estimation accuracy of the final self-position decreases. Can be suppressed.

(Feature 2) In the mobile body self-position estimation device disclosed in the present specification, when the first self-position estimation unit does not exceed the predetermined threshold in the abnormality occurrence probability in the previous step, the first The probability distribution of the state of the moving body may be updated to the latest state by further using the first self-position of the moving body in the previous step estimated by the self-position estimating unit. When the abnormality occurrence probability in the previous step exceeds a predetermined threshold value, a predetermined range including the second self-position estimated by the second self-position estimation unit in the current step You may update so that a mobile body may be located in. In this configuration, when the abnormality occurrence probability exceeds a predetermined threshold value in the previous step (that is, when the estimation accuracy of the first self-position is lower than the threshold value), in the current step, the second self-position estimation is performed. The probability distribution of the state of the moving body is updated so that the moving body is positioned within a predetermined range including the second self-position estimated by the unit. For this reason, it is possible to appropriately update the probability distribution of the state of the moving object.

(Characteristic 3) In the mobile self-position estimation apparatus disclosed in the present specification, the first self-position estimation result of the learning data includes the first self-position estimated by the first self-position estimation unit during the learning movement, and the movement When the position error with respect to the actual position of the body is less than or equal to a predetermined value, it may be classified as normal, and when the position error exceeds a predetermined value, it may be classified as abnormal. In this configuration, the first self-position estimation result is classified as normal or abnormal based on the actual position of the moving object. For this reason, a highly reliable classification result can be acquired, the reliability of learning data is improved, and as a result, a highly accurate abnormality occurrence probability calculation formula can be acquired.

(Feature 4) In the mobile body self-position estimation apparatus disclosed in the present specification, the odometry may be any of wheel odometry, visual odometry using a camera, or laser odometry using a laser sensor.

(Feature 5) In the mobile self-position estimation device disclosed in the present specification, the first self-position estimation unit uses a method of either Monte Carlo Localization (MCL) or an extended Kalman filter. One self-position may be estimated.

(Characteristic 6) The self-position estimation apparatus disclosed in the present specification can be mounted on a moving body that moves within a predetermined moving region. The moving body may include a movement information acquisition unit, an observation information acquisition unit, an environment map storage unit, and a self-position estimation device. The movement information acquisition unit may acquire the movement distance and movement direction of the moving body by odometry. The observation information acquisition unit may observe an object existing in the moving region, and acquire a distance from the moving object to the object and an orientation of the object with respect to the moving object. The environment map storage unit may store an environment map of the moving area. According to this configuration, it is possible to realize a mobile object that has more stable self-position estimation accuracy than the conventional one.

  The moving body 10 will be described with reference to FIGS. The moving body 10 is a wheel drive type moving body having wheels on the left and right, and autonomously travels within a predetermined moving area. The left and right wheels of the moving body 10 are driven independently, and the moving body 10 can change the traveling direction by changing the rotational speed of the left and right wheels. The “current step” to be described later means the latest step while the moving body 10 is traveling. In addition, “the previous step” and “the next (next) step” are respectively “the next step from the current step” and “the next step after the current step (next)”. Means "step". The moving body 10 travels from the start point to the destination based on the final self-position estimated for each step. Hereinafter, the traveling in which the mobile body 10 estimates the final self-position is referred to as “main traveling”. Strictly speaking, the mobile object 10 estimates its own position and orientation, but hereinafter, the position and orientation of the mobile object 10 are collectively referred to as “self-position”.

As shown in FIG. 1, the moving body 10 includes an encoder 12, a laser range finder 14 (hereinafter referred to as LRF 14), a triangulation sensor 16 (described later), and a computer 18.
The encoder 12 detects the rotation angles of the left and right wheels, and outputs them to a first self-position estimation unit 28 and a second self-position estimation unit 29 (described later) of the computer 18. That is, the left and right wheels are each driven by a motor (not shown). The motors that drive the left and right wheels are each provided with an encoder 12, and each encoder 12 detects the rotation angle of the left and right wheels. The rotation angles of the left and right wheels detected by the encoder 12 are input to the computer 18.
The LRF 14 emits laser light and measures the time until the emitted laser light is reflected by an object and returned. From the time measured by the LRF 14, the distance from the LRF 14 (that is, the moving body 10) to the object is measured. Further, since the direction in which the laser beam is emitted from the LRF 14 (that is, the incident angle of the laser beam reflected from the object) is known, the orientation of the object with respect to the LRF 14 can be determined. Hereinafter, the distance to the object and the direction of the object are also referred to as “observation information”. The observation information acquired by the LRF 14 is output to the first self-position estimation unit 28 of the computer 18. The LRF 14 emits laser light in a predetermined angle range with respect to the traveling direction of the moving body 10 (for example, 60 ° in the left-right direction with respect to the traveling direction).
The computer 18 includes a CPU 20 that performs arithmetic processing, a RAM 22 that temporarily stores arithmetic processing data, and a ROM 24 that stores arithmetic programs executed by the CPU. The computer 18 executes various processes described later. When the CPU 20 executes the arithmetic program stored in the ROM 24, the CPU 20 functions as the first self-position estimating unit 28, the second self-position estimating unit 29, and the like. Since the traveling control of the moving body 10 by the computer 18 can be performed by a known method, processing such as final self-position estimation by the computer 18 will be described in detail here.

(First self-position estimation process: normal)
As shown in FIG. 2, the computer 18 includes an environment map storage unit 26 and a first self-position estimation unit 28. The first self-position estimation unit 28 includes a determination unit 58, a normal time estimation unit 59, and an abnormal time estimation unit 60. The first self-position estimating unit 28 switches between processing in the normal time estimating unit 59 and processing in the abnormal time estimating unit 60 for each step based on the determination result in the determining unit 58. Both the normal time estimation unit 59 and the abnormal time estimation unit 60 estimate the self-position of the moving body 10. Hereinafter, the self-position estimated by the normal time estimation unit 59 and the abnormal time estimation unit 60 are both referred to as “first self-position”. Hereinafter, the first self-position estimation process in the normal time estimation unit 59 will be described first, and the determination process in the determination unit 58 and the first self-position estimation process in the abnormality time estimation unit 60 will be described later.

The environment map storage unit 26 is provided in the ROM 24 and stores an environment map in which the position and height of an object in the moving area are recorded on the xy coordinate plane. Examples of the objects stored in the environment map include obstacles (for example, walls, floor surfaces, etc.) in the moving area.
The normal time estimation unit 59 determines the first self-position of the moving body 10 for each step by Monte Carlo localization (hereinafter referred to as MCL) based on a known motion model stored in advance in the ROM 24. To estimate. In the first self-position estimation process using MCL, the first self-position in the current step is estimated using the first self-position estimated in the previous step. Here, when the current step is the first step after the moving body 10 starts moving from the start point, the “first self-position estimated in the previous step” is the position of the start point. Represents. This process is performed by the CPU 20. MCL is a known technique in which a moving body estimates its own position using a particle filter. Specifically, the normal time estimation unit 59 (1) first generates an initial particle set. Each particle is a vector that represents a position candidate of the moving body 10, and includes, as vector elements, the x coordinate, y coordinate, and yaw angle of the moving body 10 in the xy coordinate system. (2) Next, the normal time estimation unit 59 determines the moving distance and moving direction for one step of the moving body 10 (hereinafter referred to as “moving distance and moving direction” based on the rotation angle of each wheel acquired from the encoder 12. Also referred to as “movement information”). Specifically, the normal time estimation unit 59 calculates movement information of the moving body 10 in the current step from the rotation angle of each wheel acquired from the encoder 12 in the current step. Then, by obtaining the difference between the movement information of the moving body 10 in the current step and the movement information of the moving body 10 that has already been calculated in the previous step, the movement information for one step of the moving body 10 is obtained. calculate. Here, when the current step is the first step described above, the “movement information of the moving body 10 already calculated in the previous step” is zero. Then, movement information for one step of the moving body 10 is input to the movement model, and each particle is moved according to the movement model. (3) Subsequently, the normal time estimation unit 59 matches the observation information input from the LRF 14 with the environment map acquired from the environment map storage unit 26 for each particle, and the likelihood (probability) of each particle. Calculate (4) Then, particles are selected so that the number of particles is large in the high likelihood region and small in the low likelihood region, and noise is added to generate a new set.

  The normal time estimation unit 59 increases the likelihood of the particles by repeating the processes (2) to (4) described above, finds a particle that best matches the observation information, and sets the matching particle as the first self of the moving object 10. Estimated position. In other words, the normal time estimation unit 59 includes the actual position of the moving body 10 (hereinafter, also referred to as a true value) in the particle set (that is, the true value matches one of the particles constituting the particle set). ) Or until any particle in the particle set is very close to the true value, the above processes (2) to (4) are repeated. That is, in the normal first self-position estimation process, the particle set distribution is updated a plurality of times in one step. As a result, the normal time estimation unit 59 estimates the first self-position at the current step.

(Second self-position estimation process)
As shown in FIG. 2, the computer 18 includes a second self-position estimation unit 29 and a final self-position storage unit 56 in addition to the first self-position estimation unit 28. The final self-position storage unit 56 is provided in the ROM 24, and stores the final self-position (strictly, coordinates on the environment map) estimated by the final self-position estimation unit 54 (described later) for each step. . The second self-position estimating unit 29 adds movement information for one step of the moving body 10 to the final self-position in the previous step among the final self-positions stored in the final self-position storage unit 56. The value calculated by this is estimated as the self-position of the moving body 10 in the current step. As is clear from the above description, the second self-position estimating unit 29 estimates the second self-position without using the environment map. Note that the second self-position itself may include an element depending on the environment map (that is, the final self-position in the previous step includes the first self-position as a component).

  In addition, the above-mentioned “final self-position in the previous step” is, in other words, the latest final self-position stored in the final self-position storage unit 56. Hereinafter, the self-position of the moving body 10 estimated by the second self-position estimation unit 29 is referred to as a “second self-position”. This process is performed by the CPU 20. The method for obtaining the movement information for one step of the moving body 10 in the second self-position estimating unit 29 is the same as the method for obtaining the normal time estimating unit 59 of the first self-position estimating unit 28. For this reason, movement information for one step of the moving body 10 may be shared between the normal time estimation unit 59 and the second self-position estimation unit 29. When the current step is the first step after the moving body 10 starts moving from the start point, the “final self position in the previous step” means the position of the start point.

(Learning data generation process)
The mobile body 10 performs a travel called a learning travel before performing the main travel. Below, the process which produces | generates learning data based on the learning driving | running | working of the moving body 10 and the data obtained at the time of learning driving | running | working is demonstrated.

  As shown in FIG. 3, in addition to each unit used in the above-described processing, the computer 18 includes a true value calculation map storage unit 30, a true value calculation unit 32, a position error calculation unit 34, and learning data generation. A unit 36 and a learning data storage unit 38 are provided. The environment map storage unit 26 stores an environment map of a moving area in which the moving body 10 performs learning travel. The moving area in which the moving body 10 performs the learning travel may be different from the moving area in which the main traveling is performed. For this reason, when a movement area differs at the time of learning driving | running | working and the time of this driving | running | working, the environment map memory | storage part 26 has memorize | stored the environment map of both movement area | regions. In the following, in order to distinguish the movement area during the learning run from the movement area during the main run, the former is referred to as a “first movement area” and the latter is referred to as a “second movement area”. The environment map of the first movement area is referred to as “first environment map”, and the environment map of the second movement area is referred to as “second environment map”.

  FIG. 4 shows a first movement area 40 in which the moving body 10 performs a learning run. As shown in FIG. 4, a ridge 42 as an obstacle is arranged in a part of the first movement region 40. The collars 42 are arranged so as to face each other, and reflectors 44 are fixed to the opposing surface (inner surface) of the collar 42 at equal intervals. The travel of the moving body 10 during the learning travel is controlled by a remote controller (remote controller). A broken line indicates a travel route of the moving body 10 during the learning travel. As indicated by a broken line, the moving body 10 travels both in a region where the rod 42 is not disposed and in a region between the opposite rods 42.

  The true value calculation map storage unit 30 is provided in the ROM 24 and stores a map in which the positions of the ridges 42 and the reflectors 44 are recorded in the first environment map stored in the environment map storage unit 26. . In other words, the first environment map does not store information on the artificially arranged basket 42 and the reflector 44.

  The triangulation sensor 16 emits laser light in all directions (360 °), detects the laser light reflected and returned by the reflector 44, and thereby the distance to the reflector 44 and the orientation of the reflector 44 with respect to the moving body 10. And outputs this information to the true value calculation unit 32 of the computer 18. The triangulation sensor 16 is mounted on the moving body 10.

  The true value calculation unit 32 specifies the relative position of the moving body 10 with respect to the reflector 44 (strictly, including the relative azimuth) from the distance and azimuth information acquired from the triangulation sensor 16. Since the position of the reflector 44 is recorded on the map acquired from the map storage unit 30 for true value calculation, the position of the moving body 10 on the map is determined by mapping the relative position of the moving body 10 on the map. Can be identified. In this way, the true value calculation unit 32 calculates the actual position (true value) of the moving body 10. This process is performed by the CPU 20. The true value of the moving body 10 is output to the position error calculation unit 34.

On the other hand, the normal time estimation unit 59 of the first self-position estimation unit 28 estimates the first self-position of the moving body 10 in the first movement region 40 using the first environment map, and the result is the position error calculation unit 34. Output to. The first self-position estimation process and the true value calculation process are performed at the same timing. Further, the normal time estimation unit 59 determines the standard deviation σ _{x in} the x direction and the standard deviation σ in the y direction from the distribution of the latest particle set (that is, the particle set when the first self-position is estimated at the current step). _y and the maximum likelihood value w _max of the particles are calculated, and these _pieces of information (standard deviation σ _x , standard deviation σ _y , maximum likelihood w _max ) are output to the learning data generation unit 36. The standard deviation σ _x , the standard deviation σ _y , and the maximum likelihood w _max correspond to an example of “variable”.

  The position error calculation unit 34 calculates a difference (position error) between the true value acquired from the true value calculation unit 32 and the first self-position acquired from the normal time estimation unit 59 and outputs the difference to the learning data generation unit 36. . This process is performed by the CPU 20.

  When the position error acquired from the position error calculation unit 34 is 15 cm or less, the learning data generation unit 36 considers that the normal time estimation unit 59 has succeeded in estimating the first self-position, and determines the first self-position estimation result as “ Classified as “normal”. On the other hand, when the position error exceeds 15 cm, the normal time estimation unit 59 regards the first self-position estimation failure and classifies the first self-position estimation result as “abnormal”. In the learning run, the situation where the position error is classified as “abnormal” is intentionally created. That is, in the first movement area 40, the ridge 42 that is not recorded in the first environment map is arranged. For this reason, when the moving body 10 travels between the eaves 42, the degree of matching between the observation information acquired by the LRF 14 and the first environment map becomes low, and it becomes difficult to find suitable particles. This means that even if a true value is included in the particle set, the moving body 10 cannot find a particle that matches the true value as a matching particle. For this reason, when the moving body 10 travels between the eaves 42, a particle different from the true value is regarded as a compatible particle, and the possibility that the particle is estimated as the first self-position becomes extremely high. As a result, the position error between the first self position and the true value becomes large, and when the position error exceeds 15 cm, it is classified as “abnormal”.

  As described above, in the learning data generation unit 36, whether the first self-position estimation result is normal based on the position error, that is, based on the shift amount of the first self-position of the moving body 10 with respect to the “true value”. Classification is made as to whether it is abnormal. For this reason, a highly reliable classification result can be acquired. In this embodiment, the normal / abnormal classification reference value of the first self-position estimation result is set to 15 cm, but the classification reference value is not limited to this. The lower the classification reference value, the higher the abnormality prediction accuracy of the abnormality occurrence probability calculation formula.

FIG. 5 shows an example of learning data generated by the learning data generation unit 36. As shown in FIG. 5, the learning data generation unit 36 includes the time at the time of the first self-position estimation, the position error acquired from the position error calculation unit 34, the classification result of the first self-position estimation result, and the normal time estimation. A data set group (learning data) in which data sets in which the three variables (standard deviation σ _x , standard deviation σ _y , maximum likelihood w _max ) acquired from the unit 59 are associated in time series is generated (σ ₍ The alphabets of _x , σ _y and w _max are actually numerical values). By referring to the learning data, it is possible to know the position error at each time, normality / abnormality of the first self-position estimation result, and variables when the first self-position is estimated. For example, at time t = 2s, since the position error between the first self position and the true value is 7 cm, the first self position estimation result is classified as “normal”, and the variables σ _x and σ _{y at} that time are classified. , W _max are b, g, and l, respectively. At time t = 10 s, since the position error between the first self position and the true value is 20 cm, the first self position estimation result is classified as “abnormal”, and the variables σ _x and σ _{y at} that time are classified. , W _max are c, h, and m, respectively. The learning data generation process is performed by the CPU 20. As described above, the mobile body 10 itself generates the learning data, so that the system configuration can be simplified as compared with the configuration in which the learning data is generated outside the mobile body 10.

  The learning data storage unit 38 is provided in the ROM 24 and stores the learning data generated by the learning data generation unit 36. The learning data stored in the learning data storage unit 38 is output to an abnormality occurrence probability calculation formula generation unit 46 described later. In this way, by storing the learning data by the mobile body 10 itself, it is possible to verify whether there is an error by analyzing the log of the learning data when a malfunction occurs in the mobile body 10 or the like. Become. For this reason, it becomes easy to discover the cause when the malfunction occurs in the moving body 10, and the reliability of the moving body 10 can be improved. In this embodiment, the data set includes the time at the time of self-position estimation. However, the present invention is not limited to this configuration. For example, the data set may not include time, and the learning data may be configured by a data group in which the data set is arranged regardless of the first self-position estimation time. Further, the learning data may not include a position error.

（Ｚ：目的変数、Ｘ_１〜Ｘ_３：説明変数、ｂ_０：定数、ｂ_１〜ｂ_３：回帰係数）に学習データを適用して、定数ｂ_０及び回帰係数ｂ_１〜ｂ_３の値を求める。学習データを適用する際は、目的変数Ｚに分類結果（正常：０、異常：１）を用い、説明変数Ｘ_１、Ｘ_２、Ｘ_３に学習データの標準偏差σ_ｘ、標準偏差σ_ｙ、最大尤度ｗ_ｍａｘをそれぞれ用いる。これにより、次の数式で表される異常発生確率計算式；

が生成される（図７参照）。ｐは、移動体１０の本走行時において第１自己位置推定部２８により推定される第１自己位置（即ち、通常時推定部５９により推定される第１自己位置と、後述する異常時推定部６０により推定される第１自己位置）が異常となる確率（以下、異常発生確率とも称する）である。異常発生確率ｐは、第１自己位置が真値からどれだけ離れているかを示す尺度であり、第１自己位置が真値から遠ざかるにつれて大きくなる。第１自己位置の推定精度が低くなるほど、異常発生確率ｐが大きくなる。説明変数Ｘ_１〜Ｘ_３に用いる変数（σ_ｘ、σ_ｙ、ｗ_ｍａｘ）は互いに独立しているため、回帰係数ｂ_１〜ｂ_３は、項目間の影響が排除された係数となる。従って、異常発生確率を算出する際に、より正確に項目ごとの影響度を量ることができる。異常発生確率計算式生成処理は、ＣＰＵ２０で行われる。
異常発生確率計算式記憶部４８は、ＲＯＭ２４に設けられており、異常発生確率計算式生成部４６で生成された異常発生確率計算式を記憶する。 (Abnormality occurrence probability formula generation processing)
A process in which the moving body 10 generates the abnormality occurrence probability calculation formula will be described. As shown in FIG. 6, the computer 18 includes an abnormality occurrence probability calculation formula generation unit 46 and an abnormality occurrence probability calculation formula storage unit 48 in addition to the units used in the above-described processing.
The abnormality occurrence probability calculation formula generation unit 46 acquires the learning data stored in the learning data storage unit 38 and performs logistic regression analysis on the learning data to generate an abnormality occurrence probability calculation formula. Logistic regression analysis is a method of multivariate analysis, and when the result is binary, it is a method of generating a statistical regression model that can predict the probability that the result will occur. Specifically, a logistic model represented by the following formula:

(Z: objective _variable, X 1 to X _3: an explanatory variable, _{b 0: constant,} b 1 ~b _3: regression coefficient) by applying the learning data, the value of the constant _{b 0} and regression coefficient _b 1 ~b ₃ Ask for. When applying the learning data, the classification result (normal: 0, abnormal: 1) is used for the objective variable Z, and the standard deviation σ _x , standard deviation σ _y of the learning data is used for the explanatory variables X ₁ , X ₂ , X ₃ , Each maximum likelihood w _max is used. As a result, an abnormality occurrence probability calculation formula represented by the following formula:

Is generated (see FIG. 7). p is a first self-position estimated by the first self-position estimation unit 28 during the main traveling of the mobile body 10 (that is, a first self-position estimated by the normal time estimation unit 59 and an abnormal time estimation unit described later). The first self-position estimated by 60) is a probability of abnormality (hereinafter also referred to as abnormality occurrence probability). The abnormality occurrence probability p is a measure indicating how far the first self-position is from the true value, and increases as the first self-position moves away from the true value. As the estimation accuracy of the first self-position decreases, the abnormality occurrence probability p increases. Since the variables (σ _x , σ _y , w _max ) used for the explanatory variables X _{1 to} X ₃ are independent from each other, the regression coefficients b _{1 to} b ₃ are coefficients in which the influence between items is eliminated. Therefore, when calculating the abnormality occurrence probability, the degree of influence for each item can be measured more accurately. The abnormality occurrence probability calculation formula generation process is performed by the CPU 20.
The abnormality occurrence probability calculation formula storage unit 48 is provided in the ROM 24 and stores the abnormality occurrence probability calculation formula generated by the abnormality occurrence probability calculation formula generation unit 46.

(First self-position estimation process: when resetting)
As described above, the first self-position estimation unit 28 switches between the process in the normal time estimation unit 59 and the process in the abnormal time estimation unit 60 for each step based on the determination result in the determination unit 58. Below, the 1st self-position estimation process in the abnormal time estimation part 60 is demonstrated.
The abnormal time estimation unit 60 is different from the normal time estimation unit 59 in that a motion model is not used when generating a particle set. Specifically, the abnormal time estimation unit 60 first acquires the second self-position at the current step from the second self-position estimation unit 29 and generates a particle set within a predetermined range including the second self-position. To do. Within this predetermined range, the moving body 10 (that is, the true value) is included. The number of particles constituting the particle set is substantially the same as the number of particles in the normal time estimation unit 59.
Next, the abnormal time estimation unit 60 executes the processes (3) and (4) in the normal time estimation unit 59. The abnormal time estimation unit 60 performs the above (3) and (4) until the true value matches any of the particles constituting the particle set or until any particle in the particle set becomes very close to the true value. Repeat the process. That is, also in the first self-position estimation process at the time of resetting, the particle set distribution is updated a plurality of times in one step. Thereby, in the abnormal time estimation part 60, the 1st self-position in the present step is estimated. That is, the abnormal time estimation unit 60 estimates the first self-position in the current step using the second self-position in the current step, not the first self-position in the previous step. This process is performed by the CPU 20. The particles constituting the particle set may be randomly distributed within the predetermined range, may be uniformly distributed, or may be distributed according to a predetermined standard.

(Abnormality occurrence probability calculation process)
The process in which the moving body 10 calculates the abnormality occurrence probability p will be described. The computer 18 includes an abnormality occurrence probability calculation unit 50 in addition to the units used in the processing described above.
When the moving body 10 performs the main traveling in the second movement area, the normal time estimation unit 59 or the abnormal time estimation unit 60 of the first self position estimation unit 28 estimates the first self position using the second environment map. To do. The three variables σ _x , σ _y , and w _max when the first self-position at the current step is estimated by the normal time estimation unit 59 or the abnormal time estimation unit 60 are output to the abnormality occurrence probability calculation unit 50.
The abnormality occurrence probability calculation unit 50 acquires the three variables σ _x , σ _y , w _max in the current step acquired from the normal time estimation unit 59 or the abnormality time estimation unit 60 from the abnormality occurrence probability calculation formula storage unit 48. Anomaly occurrence probability p at the current step is obtained by inputting each to X ₁ , X ₂ , X ₃ of the abnormality occurrence probability calculation formula. The abnormality occurrence probability p is calculated for each step. This process is performed by the CPU 20. The abnormality occurrence probability p calculated by the abnormality occurrence probability calculation unit 50 is output to the determination unit 58 of the first self-position estimation unit 28 and the weighting coefficient calculation unit 52. The abnormality occurrence probability p at the current step is used for determination processing (described later) in the next step in the determination unit 58, and is used in the weight coefficient calculation processing in the current step in the weight coefficient calculation unit 52.

(Determination process)
A process of determining whether the mobile body 10 performs the process of estimating the first self-position in the current step by the normal time estimation unit 59 or the abnormal time estimation unit 60 will be described. As described above, the first self-position estimation unit 28 includes the determination unit 58. The determination unit 58 acquires the abnormality occurrence probability p in the previous step from the abnormality occurrence probability calculation unit 50. The determination unit 58 causes the normal time estimation unit 59 to execute the first self-position estimation process when the abnormality occurrence probability p satisfies 0 <p ≦ 0.8, and when the abnormality occurrence probability p satisfies 0.8 <p The unit 60 causes the first self-position estimation process to be executed. In the first step after the moving body 10 starts moving from the start point, the abnormality occurrence probability p that is a criterion for determination has not been calculated yet, so the normal time estimation unit 59 performs the first self-position estimation process. Do. That is, the determination process is performed from the second and subsequent steps. Further, the threshold value of the abnormality occurrence probability p serving as a reference for determining whether the first self-position estimation process is performed by the normal time estimation unit 59 or the abnormal time estimation unit 60 is not limited to 0.8. Or less than 0.8.

(Final self-position estimation process)
The mobile body 10 performs the above-described learning data generation processing and abnormality occurrence probability calculation formula generation processing before performing the main traveling in the second movement region. And when the mobile body 10 performs this driving | running | working, a weighting coefficient is calculated based on the abnormality occurrence probability calculated from the abnormality occurrence probability calculation formula, and the final self-position is estimated using the weighting coefficient. Below, the last self-position estimation process at the time of this driving | running | working of the moving body 10 is demonstrated. As shown in FIG. 2, the computer 18 includes a weight coefficient calculation formula storage unit 51, a weight coefficient calculation unit 52, and a final self-position estimation unit 54 in addition to the units used in the above-described processing.

（ｐ：異常発生確率、α：重み係数）で表される（図８参照（後述））。
重み係数計算部５２は、重み係数計算式記憶部５１から取得した重み係数計算式に、異常発生確率計算部５０から取得した現在のステップにおける異常発生確率ｐを入力して重み係数αを求める。この処理は、ＣＰＵ２０で行われる。重み係数計算部５２で算出された重み係数αは、最終自己位置推定部５４に出力される。 The weighting factor calculation formula storage unit 51 is provided in the ROM 24 and stores a preset weighting factor calculation formula. The weighting factor calculation formula is the following sigmoid function:

(P: abnormality occurrence probability, α: weighting coefficient) (see FIG. 8 (described later)).
The weighting factor calculation unit 52 obtains the weighting factor α by inputting the abnormality occurrence probability p in the current step acquired from the abnormality occurrence probability calculation unit 50 to the weighting factor calculation equation acquired from the weighting factor calculation equation storage unit 51. This process is performed by the CPU 20. The weighting factor α calculated by the weighting factor calculation unit 52 is output to the final self-position estimation unit 54.

（ｒ_ｆ：最終自己位置、ｒ_１：第１自己位置、ｒ_２：第２自己位置）に従って計算を行い、移動体１０の現在のステップにおける最終自己位置を算出する。重み係数αは第１自己位置の重み係数として使用され、１から重み係数αを減じた係数１−αは第２自己位置の重み係数として使用される。この処理は、ＣＰＵ２０で行われる。最終自己位置推定部５４で算出された最終自己位置は、最終自己位置記憶部５６に出力される。最終自己位置記憶部５６に出力され記憶される最終自己位置は、１つ後のステップにおける第２自己位置推定処理に用いられる。 The final self-position estimation unit 54 uses the weighting coefficient α obtained from the weighting coefficient calculation unit 52 and the first self-position obtained from the first self-position estimation unit 28 and the second self-position estimation unit 29 obtained from the second self-position estimation unit 29. The weighted average value of the self position is calculated, and the value is estimated as the final self position. Specifically, the final self-position estimating unit 54 calculates the following formula:

Calculation is performed according to (r _f : final self-position, r ₁ : first self-position, r ₂ : second self-position), and the final self-position of the moving body 10 in the current step is calculated. The weighting coefficient α is used as a weighting coefficient for the first self-position, and a coefficient 1−α obtained by subtracting the weighting coefficient α from 1 is used as the weighting coefficient for the second self-position. This process is performed by the CPU 20. The final self-position calculated by the final self-position estimation unit 54 is output to the final self-position storage unit 56. The final self-position output and stored in the final self-position storage unit 56 is used for the second self-position estimation process in the next step.

  Here, the weighting coefficient calculation formula will be described with reference to the above function (Equation 3) and FIG. The weighting factor α calculated by the weighting factor calculation formula is a function of the abnormality occurrence probability p. The weighting factor α, which is the weighting factor of the first self-position, decreases as the abnormality occurrence probability p increases (that is, the estimation accuracy of the first self-position decreases). Specifically, the value of the weighting coefficient α is close to 1 when the abnormality occurrence probability p satisfies 0 <p <0.1, is 0.5 when p = 0.2, and is 0. It is approximated to 0 when .3 <p is satisfied. For this reason, the final self-position substantially coincides with the first self-position when the abnormality occurrence probability p is less than 10%, and is intermediate between the first self-position and the second self-position when the abnormality occurrence probability p is 20%. It becomes a point, and it can be seen that when the abnormality occurrence probability exceeds 30%, it almost coincides with the second self-position. As is apparent from the above description, in the present embodiment, when the abnormal time estimation unit 60 performs the first self-position estimation process, the second self-position and the final self-position substantially coincide. Note that the constants (50, 0.2) in the weight coefficient calculation formula are not limited to the above values. Further, the weight coefficient calculation formula is not limited to the sigmoid function. Various functions can be used depending on how the value of the weighting factor α is associated with the value of the abnormality occurrence probability p.

  Next, the determination process will be described in more detail. When the moving body 10 travels in the vicinity of an obstacle that is not recorded in the second environment map in the second movement area, the degree of matching between the observation information acquired by the LRF 14 and the environment map becomes low (the observation information is unnecessary). Therefore, the estimation accuracy of the first self-position is lowered. For this reason, the abnormality occurrence probability p increases. That is, even if a true value is included in the particle set, the moving object 10 cannot regard a particle corresponding to the true value as a conforming particle, and regards another particle as a conforming particle. In the moving body 10 according to the embodiment, the normal time estimation unit 59 estimates the first self-position at the current step using the first self-position one step before in accordance with the MCL technique. For this reason, when the mobile body 10 continues to travel in the vicinity of an obstacle that is not recorded in the second environment map, the first self-position estimated by the normal time estimation unit 59 is further away from the true value. Eventually, the true value is not included in the particle set. Once the particle set no longer contains the true value, even if the moving body 10 subsequently travels in the area where no obstacle is located in the second movement area (that is, useful observation information is displayed). Even if it can be acquired), the MCL method makes it impossible to generate a particle set in a range including a true value. That is, the normal time estimation unit 59 cannot accurately estimate the first self-position.

  When the abnormality occurrence probability p exceeds 80%, the determination unit 58 of the first self-position estimation unit 28 causes the abnormality-time estimation unit 59 to perform the first self-position estimation process instead of the normal-time estimation unit 59. The abnormal time estimation unit 60 generates a particle set within a predetermined range including the second self-position estimated by the second self-position estimation unit 29 in the current step. A true value is included in the predetermined range. For this reason, if the moving body 10 can acquire useful observation information by the abnormal time estimation unit 60, the abnormal time estimation unit 60 can regard the particle corresponding to the true value in the particle set as a suitable particle, In the step, the first self-position can be estimated with high accuracy. That is, when the determination unit 58 switches the first self-position estimation method based on the abnormality occurrence probability p, the first self-position estimation accuracy in the first self-position estimation unit 28 can be improved again. When the estimation accuracy of the first self-position is improved and the abnormality occurrence probability p in the current step becomes 80% or less, the determination unit 58 normally performs the first self-position estimation process in the next step from the abnormality time estimation unit 60. Switching to the time estimation unit 59 (described later).

  On the other hand, consider a case where the abnormal time estimation unit 60 cannot acquire useful observation information, for example, when the moving body 10 continues to travel in the vicinity of an obstacle that is not recorded in the second environment map. In this case, even if the particle set is generated within the range including the true value, the abnormal time estimation unit 60 regards a particle different from the particle corresponding to the true value as the matching particle, and accurately determines the first self-position. There is a possibility that it cannot be estimated. Thereby, when the abnormality occurrence probability p still exceeds 80%, the determination unit 58 causes the abnormality estimation unit 60 to perform the first self-position estimation process in the next step.

  On the other hand, when the abnormality occurrence probability p is 80% or less, the determination unit 58 causes the normal time estimation unit 59 to perform the first self-position estimation process. The inventors of the present application show that, if the probability of occurrence of an abnormality p is 80% or less, even if the moving body 10 is traveling near an obstacle that is not recorded in the second environment map, the true value is obtained. Has been confirmed to be included in the particle set. For this reason, if the normal time estimation unit 59 can acquire useful observation information, the normal time estimation unit 59 can regard the particle corresponding to the true value in the particle set as the matching particle, and the first step in the current step. The self-position can be estimated with high accuracy. As a result, the abnormality occurrence probability p in the current step does not exceed 80%, and the determination unit 58 causes the normal time estimation unit 59 to perform the first self-position estimation process in the next step.

  On the other hand, let us consider a case where the mobile object 10 cannot acquire useful observation information in a plurality of consecutive steps, for example, by traveling for a long time near an obstacle that is not recorded in the second environment map. In this case, the first self-position estimation accuracy of the normal time estimator 59 decreases as each step is followed. When the abnormality occurrence probability p finally exceeds 80%, the determination unit 58 switches the first self-position estimation from the process in the normal time estimation unit 59 to the process in the abnormality time estimation unit 60. Thereby, as above-mentioned, the estimation precision of a 1st self-position can be improved again.

  The effect of the moving body 10 of an Example is demonstrated. In this mobile object 10, the weighting factor α used when the final self-position estimating unit 54 estimates the final self-position of the mobile object 10 is a function of the abnormality occurrence probability p. The abnormality occurrence probability calculation formula for calculating the abnormality occurrence probability p is obtained by analyzing learning data acquired by the moving movement of the mobile body 10 using a method of logistic regression analysis which is one method of machine learning. Is an expression. Therefore, the weighting factor α is not affected by factors other than the abnormality occurrence probability calculation formula. Therefore, for example, even if the moving area is changed or the administrator of the moving object 10 is changed, the reference for determining the value of the weighting coefficient α does not change. As a result, the estimation accuracy of the final self-position of the moving body 10 can be stabilized.

The second self-position is obtained by adding movement information for one step calculated based on the rotation angles of the left and right wheels acquired from the encoder 12 to the final self-position in the previous step. . The information detected by the encoder 12 may include an error due to slipping of each wheel or the like. However, when the detection distance by the encoder 12 is relatively short, the above error can be almost ignored. Since the movement distance for one step is extremely short, the accuracy of the movement information for one step is high. The second self-position is estimated without using the second environment map. For this reason, even when the moving body 10 travels in the vicinity of an obstacle that is not recorded on the second environment map in the second movement region, the estimation accuracy of the second self-position is reduced due to this. Never do. Therefore, if the estimation accuracy of the final self-position one step before is high, the second self-position in the current step can maintain high estimation accuracy.
In the moving body 10 of the present embodiment, the weighting factor α of the first self-position decreases and the weighting coefficient 1-α of the second self-position increases as the abnormality occurrence probability p increases. According to this configuration, when the abnormality occurrence probability p is high (that is, when the estimation accuracy of the first self-position is low), the ratio of the first self-position to the final self-position is reduced and the ratio of the second self-position is set. Can be increased. As described above, the estimation accuracy of the second self-position is high. For this reason, according to this structure, even if the estimation accuracy of the first self-position is lowered, it is possible to suppress the estimation accuracy of the final self-position from being lowered. Further, since the weighting coefficient calculation formula is a continuous smooth function, the weighting coefficient α varies smoothly according to the transition of the abnormality occurrence probability p. For this reason, the final self-position is a value that can continuously transition between the first self-position and the second self-position. Accordingly, the final self-position estimation accuracy is further improved.

  Further, in the conventional mobile object, once the true value is not included in the particle set, the subsequent steps have to switch from the MCL method to another method (odometry, GPS, etc.). However, in the moving body 10 of the embodiment, when the true value is not included in the particle set, the generation method of the particle set is changed so that the true value is included in the particle set again. For this reason, the first self-position estimation process using the MCL technique can be resumed, and the first self-position estimation accuracy can be improved again. Therefore, even after the ratio of the second self-position occupying the final self-position becomes extremely high due to a decrease in the estimation accuracy of the first self-position, the final self-position is improved by improving the estimation accuracy of the first self-position again. It is possible to increase again the ratio of the first self-position in the area. In this way, by utilizing the functions of the first self-position estimation unit 28 and the second self-position estimation unit 29 in a well-balanced manner, the final self-position is estimated until the moving body 10 reaches the destination from the start point. The accuracy can be kept high.

  The inventors of the present application conducted an experiment comparing the final self-position estimation method disclosed in this specification with the conventional MCL method. In the experiment, the self-position estimation apparatus using the above-described method was mounted on each of the two moving bodies, and each moving body was caused to travel in a moving area in which an obstacle that was not recorded in the environmental map was placed. After running, the average position error for each step from the starting point to the destination was calculated for each moving object. As a result of the experiment, the average value of the position error was reduced by about 30% in the method disclosed in the present specification as compared with the conventional method. From this, it was found that the moving body 10 can estimate the final self-position with more stable accuracy than in the past.

  Further, in the present embodiment, when the first self-position estimation process is switched from the process in the normal time estimation unit 59 to the process in the abnormal time estimation unit 60, the method of generating the particle set is changed. Specifically, the particle set is generated in a predetermined range including the second self position in the current step. The number of particles is substantially the same as the number of particles used in the normal time estimation unit 59. On the other hand, in the conventional method, when re-dispersing particles, a large amount of particles are dispersed over the entire moving region. For this reason, the accuracy of self-position estimation is low, and processing takes time. In this embodiment, a particle set is generated with reference to a specific position (that is, the second self position (which is also the final self position in this embodiment)) located relatively close to the true value. As a result, the first self-position can be estimated accurately and in a short time without increasing the number of particles.

In the present embodiment, information (that is, variables such as a self-position estimation result and a standard deviation σ _x ) that is the basis of learning data is acquired by causing the moving body 10 to actually travel through the first moving area 40. . For this reason, the first self-position estimation performance of the moving body 10 is directly reflected in the learning data. Thus, by generating learning data based on information obtained by actually traveling in a specific area, a more accurate abnormality occurrence probability calculation formula can be generated.

  As mentioned above, although the Example of the technique which this specification discloses was described in detail, these are only illustrations, and the mobile body which this specification discloses contains what changed and changed the said Example variously. It is.

  For example, the first self-position estimation unit 28 may perform the first self-position estimation process using an extended Kalman filter instead of the MCL. In this case, using the values obtained from the estimated error covariance matrix (for example, the variance value, the value of each element of the error covariance matrix, the major axis and minor axis of the error ellipse) as explanatory variables, Can be generated. Moreover, normal / abnormality of the first self-position estimation result can be accurately determined by inputting these variables obtained during the actual traveling into the abnormality occurrence probability calculation formula.

  In the above embodiment, the movement information is acquired using the encoder 12 provided in the motor. However, the method of acquiring the movement information is not limited to this. For example, the movement information may be acquired using visual odometry using a camera or laser odometry using a laser sensor.

  In addition, the moving body 10 may not have the learning data generation unit 36, and the learning data may be generated outside the moving body 10. Further, the moving body 10 may not have the learning data storage unit 38. That is, the learning data may be deleted once the abnormality occurrence probability calculation formula is generated based on the learning data. Further, the mobile object 10 may not have the abnormality occurrence probability calculation formula generation unit 46, and the abnormality occurrence probability calculation expression may be generated outside the mobile object 10.

Further, the abnormality occurrence probability calculation formula generation unit 46 may use a support vector machine instead of multivariate analysis. The support vector machine is a well-known technique in pattern recognition modeling, and is a technique for constructing an identification function that outputs a class to which a feature vector belongs in binary when a feature vector with an unknown class assignment is input. The discriminant function is composed of a training sample set with known class membership. For this reason, a discriminant function (abnormality occurrence probability calculation formula) can be configured by learning the learning data with the support vector machine using the learning data of the embodiment as a sample set for training. Then, the normal / abnormality of the first self-position estimation result can be determined by inputting the three variables σ _x , σ _y , and w _max acquired from the first self-position estimation unit 28 to the identification function. . Also by this method, the same effect as the embodiment can be obtained.

Further, the abnormality occurrence probability calculation formula generation unit 46 may use a neural network instead of the multivariate analysis. The neural network is a well-known technique in pattern recognition modeling, and is a technique for constructing a discriminant function that outputs a class to which a variable belongs in binary when a variable with unknown class assignment is input. In this method, the discrimination function can be configured in two ways, supervised learning and unsupervised learning. In supervised learning, a learning example and a target output for the learning example are given, and the weight of the function in the initial state is adjusted so that the output obtained by inputting the learning example to the function in the initial state matches the target output Thus, an identification function is configured. For this reason, the three variables σ _x , σ _y , and w _max of the learning data of the embodiment are taken as learning examples, and the learning result is learned by a neural network with the classification result (normal / abnormal) as a target output. The discriminant function (abnormality occurrence probability calculation formula) can be constructed. Then, the normal / abnormality of the first self-position estimation result can be determined by inputting the three variables σ _x , σ _y , and w _max acquired from the first self-position estimation unit 28 to the identification function. . Also by this method, the same effect as the embodiment can be obtained.

  When analyzing learning data, discriminant analysis may be used instead of logistic regression analysis.

  In addition, the learning travel may be performed by simulation instead of actually traveling a specific area on the moving body.

  Moreover, the moving body 10 may not be a wheel drive type, for example, may be a bipedal walking type. Further, the moving body 10 may have a plurality of LRFs 14.

  Further, in the true value calculation unit 32, the true value of the moving body 10 is calculated using the triangulation sensor 16 and the reflector 44, but the method of calculating the true value is not limited to this. For example, the true value may be calculated using a motion capture or magnetic rail technique.

  Specific examples of the present invention have been described in detail above, but these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and changes of the specific examples illustrated above. The technical elements described in this specification or the drawings exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. In addition, the technology illustrated in the present specification or the drawings achieves a plurality of objects at the same time, and has technical utility by achieving one of the objects.

10: mobile body, 12: encoder, 14: laser range finder (LRF), 26: environment map storage unit, 28: first self-position estimation unit, 29: second self-position estimation unit, 36: learning data generation unit, 38: learning data storage unit, 46: abnormality occurrence probability calculation formula generation unit, 48: abnormality occurrence probability calculation formula storage unit, 50: abnormality occurrence probability calculation unit, 51: weighting factor calculation formula storage unit, 52: weighting factor calculation unit 54: Final self-position estimation unit 56: Final self-position storage unit 58: Determination unit 59: Normal time estimation unit 60: Abnormal time estimation unit

Claims (7)

An apparatus that estimates the final self-position of a moving body that moves within a predetermined moving area for each step,
A first self-position estimation unit, a second self-position estimation unit, an abnormality occurrence probability calculation formula storage unit, an abnormality occurrence probability calculation unit, and a final self-position estimation unit,
The first self-position estimation unit includes an environment map of the moving area, “a distance from the moving object to an object existing in the moving area” and “the moving object relative to the moving object” observed in the current step. Update the probability distribution of the state of the moving body to the latest state using at least the azimuth of the object, and estimate the first self-position based on the probability distribution of the latest state;
The second self-position estimating unit, a final self-position of the movable body in the final self-position preceding steps estimated by the estimation unit is obtained by O Dometori, the current from the previous step Estimating the second self-position by adding the moving distance and moving direction of the moving body up to the step;
The abnormality occurrence probability calculation formula storage unit is obtained by machine learning of learning data acquired when the first self-position estimation unit estimates the first self-position during the learning movement of the mobile body Memorize abnormality occurrence probability formula,
The abnormality occurrence probability calculation unit inputs, into the abnormality occurrence probability calculation formula, a plurality of variables acquired when the first self-position estimation unit estimates the first self-position during the main movement of the moving body. To calculate the probability of anomaly occurrence,
The final self-position estimation unit is a weighted average calculated using the first self-position acquired from the first self-position estimation unit and the second self-position acquired from the second self-position estimation unit Let the value be the final self-position in the current step,
The learning data includes a plurality of variables acquired when the first self-position estimating unit estimates the first self-position during learning movement, and a first self-position estimation result at that time is classified as normal or abnormal Have multiple data related to the classification results
The mobile body self-position estimation apparatus, wherein the weighting factor used in the final self-position estimation unit is a function of the abnormality occurrence probability acquired by the abnormality occurrence probability calculation unit.   The mobile body self-position estimation apparatus according to claim 1, wherein the weighting coefficient of the first self-position decreases and the weighting coefficient of the second self-position increases as the abnormality occurrence probability increases. The first self-position estimation unit
When the abnormality occurrence probability in the previous step does not exceed a predetermined threshold, the first self-position of the moving body in the previous step estimated by the first self-position estimation unit is further used. Update the probability distribution of the state of the moving body to the latest state,
When the abnormality occurrence probability in the previous step exceeds a predetermined threshold, the second self-position estimated by the second self-position estimation unit in the current step is the probability distribution of the state of the moving object. The mobile body self-position estimation apparatus according to claim 1, wherein the mobile body is updated so that the mobile body is positioned within a predetermined range.   The first self-position estimation result of the learning data is obtained when a position error between the first self-position estimated by the first self-position estimation unit during learning movement and the actual position of the moving body is a predetermined value or less. The mobile body self-position estimation apparatus according to any one of claims 1 to 3, wherein the mobile body is classified as normal and is classified as abnormal when the position error exceeds a predetermined value.   5. The mobile body self-position estimation device according to claim 1, wherein the odometry is any one of wheel odometry, visual odometry using a camera, and laser odometry using a laser sensor.   The said 1st self-position estimation part estimates the said 1st self-position using the method of either a Monte Carlo location identification method (Monte Carlo Localization, MCL) or an extended Kalman filter, The any one of Claims 1-5. The self-position estimation apparatus of the moving body as described in 2. A moving body that moves within a predetermined movement area,
A movement information acquisition unit that acquires a movement distance and a movement direction of the moving body by odometry;
An observation information acquisition unit for observing an object existing in the moving region and acquiring a distance from the moving object to the object and an orientation of the object with respect to the moving object;
An environmental map storage unit storing an environmental map of the moving area;
A mobile body comprising: the self-position estimation apparatus according to any one of claims 1 to 6.

Images