JP2019207177A - Self-position estimation apparatus - Google Patents

Abstract

To enable recognition of a relationship between a result of estimating a position and a true value.SOLUTION: A self-position estimation apparatus (100) includes: generation means (11) for generating plural position candidates of a mobile body (1); extraction means (12) for extracting plural probable position candidates from the plural position candidates by comparing the plural position candidates with map information; first calculation means (13, 14) for calculating respective degrees of reliability of the plural position candidates by comparing among measurement results of measurement means (20) that measures surroundings environment of the mobile body, the map information and the plural position candidates; a second calculation means (14) for calculating respective weighting of the plural probable position candidates on the basis of the respective degrees of reliability of the plural position candidates; and estimation means (15, 16) for estimating a position of the mobile body on the basis of a provisional position indicated by each of the plural probable position candidates and the weighting of the provisional position, and determining reliability of the estimated position on the basis of the respective weighting of the plural probable position candidates and a degree of reliability of the weighting.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a technical field of a self-position estimation apparatus that estimates the position of a mobile object.

  As a technique used in this type of device, for example, a positioning position of the host vehicle is calculated using GPS (Global Positioning System), and a map matching process is performed for comparing the positioning position with road data included in the map data. A technique for correcting the positioning position and estimating the current position of the host vehicle has been proposed (see Patent Document 1). Other related technologies include those described in Patent Documents 2 and 3.

JP 2011-058909 A JP 2017-062539 A Japanese Unexamined Patent Publication No. 2016-110576

  In the technique described in Patent Document 1, the reliability, which is an index indicating the accuracy of the current position, is based on the degree of multipath for radio waves received from GPS satellites, the influence of terrain, the confidence of map data, and the like. It is determined. However, this reliability is not an indication of how close the estimated current position is to the true value (ie, actual position) (ie, accuracy), but a concept that can be paraphrased as estimation error or estimation accuracy. It is. That is, the technique described in Patent Document 1 cannot know the relationship between the estimated current position and the true value. For this reason, the technique described in Patent Document 1 has a technical problem that when the estimated current position is deviated from the true value, the apparatus cannot recognize the deviation.

  This invention is made | formed in view of the said problem, and makes it a subject to provide the self-position estimation apparatus which can recognize the relationship between an estimation result and a true value.

  A self-position estimation apparatus according to an aspect of the present invention includes a measuring unit that measures a surrounding environment of a moving body, a movement detecting unit that detects the movement of the moving body, a map database that stores map information, and (i) The position of the moving body at a first time, and (ii) the movement detected by the motion detection means in a period from the first time to a second time after the first time. Generating means for generating a plurality of position candidates indicating the temporary position of the mobile body at the second time based on body movement, comparing the plurality of position candidates with the map information, Extracting means for extracting a plurality of likely position candidates from a plurality of position candidates, comparing the measurement result by the measuring means, the map information, and the plurality of position candidates, and determining the reliability of each of the plurality of position candidates. First calculating means for calculating; Second calculation means for calculating a weight of each of the plurality of probable position candidates based on the reliability of each of the plurality of position candidates; a temporary position indicated by each of the plurality of probable position candidates; Estimating the position of the moving body based on the weight of each of the probable position candidates and determining the reliability of the estimated position based on the weight and the reliability of each of the plurality of probable position candidates Means.

It is a block diagram which shows the structure of the self-position estimation apparatus which concerns on embodiment. It is a conceptual diagram which shows the concept of the calculation method of a position candidate. It is a conceptual diagram which shows the concept of the extraction method of a likely position candidate. It is a conceptual diagram which shows the concept of correct / incorrect determination. It is a block diagram which shows the structure of the self-position estimation apparatus which concerns on an application example.

  An embodiment according to a self-position estimation apparatus will be described with reference to FIGS. In the following embodiment, a self-position estimation device mounted on a vehicle as a moving body is taken as an example.

(Constitution)
The configuration of the self-position estimation apparatus according to the embodiment will be described with reference to FIG. FIG. 1 is a block diagram illustrating a configuration of the self-position estimation apparatus according to the embodiment.

  In FIG. 1, a self-position estimation apparatus 100 is mounted on a vehicle 1. The self-position estimation apparatus 100 includes an arithmetic unit 10, an external sensor 20, an internal sensor 30, and a map database 40 (hereinafter referred to as “map DB 40” as appropriate).

  The external sensor 20 includes, for example, a laser sensor, LIDAR (Light Detection and Ranging), a camera that captures the periphery of the vehicle 1, and the like. The inner sensor 30 includes, for example, an encoder, a gyro, and the like. The map DB 40 stores map information. The map information includes, for example, a map including information on features such as white lines, lane marks, and utility poles on the road, and a point cloud map.

  The arithmetic device 10 includes a position candidate calculation unit 11, a matching unit 12, a correctness determination unit 13, a reliability comparison unit 14, a position as a processing block that is logically realized therein or as a processing circuit that is physically realized. An estimation unit 15 and a reliability determination unit 16 are provided. The arithmetic unit 10 estimates the position of the vehicle 1 and the reliability of the position using the sensor observation values acquired by the external sensor 20 and the internal sensor 30 and the map information stored in the map DB 40.

(Position estimation)
Position estimation in the self-position estimation apparatus 100 (in other words, operation of the arithmetic apparatus 10) will be described.

1. Position Candidate Calculation A position candidate calculation method in the position candidate calculation unit 11 will be described with reference to FIG. FIG. 2 is a conceptual diagram showing a concept of a position candidate calculation method.

  The position candidate calculation unit 11 calculates a position candidate of the vehicle 1 at time t using a calculation method described below. The calculation method includes the position of the vehicle 1 at time t-1 (see the black circle in FIG. 2), and the inner-field sensor observation value acquired by the inner-field sensor 30 during the period from time t-1 to time t (that is, Based on the motion of the vehicle 1 (see arrow “Detected motion” in FIG. 2) and dead reckoning (autonomous navigation) for determining the position of the vehicle 1 at time t.

  Here, the actual position of the vehicle 1 often deviates from the position of the vehicle 1 obtained by dead reckoning due to disturbance factors such as road surface unevenness, wind, and tire distortion. Therefore, in this calculation method, in addition to the position of the vehicle 1 at time t-1 and the observed value of the internal sensor, a random number corresponding to a disturbance factor is used, and the position candidate of the vehicle 1 at time t (see the white circle in FIG. 2) ) Are calculated.

  Each of the plurality of position candidates includes, in addition to the position information indicating the position, the posture information indicating the attitude of the vehicle 1, the position of the vehicle 1 at time t-1 (that is, the position used to calculate the position candidate at time t-1). ) Information indicating the reliability of the vehicle 1 (hereinafter referred to as “reliability at time t−1” as appropriate), and information indicating the weight of the position of the vehicle 1 at time t−1 (hereinafter referred to as “weight at time t−1” as appropriate). "). Note that the initial position of the vehicle 1 (that is, the position at time t = 0) may be obtained by using an existing positioning technique such as GPS, or may be manually input by the user. The reliability and weight will be described later.

2. The operation of the matching matching unit 12 will be described with reference to FIG. FIG. 3 is a conceptual diagram showing a concept of a method for extracting a likely position candidate. In FIG. 3, in order to make the concept of the extraction method easy to understand, the variation of position candidates (white circles) is exaggerated.

  For example, the matching unit 12 refers to the position of the vehicle 1 at time t−1 and acquires a map around the vehicle 1 from the map DB 40. As shown in FIG. 3B, the matching unit 12 superimposes a plurality of position candidates (here, the position candidates of the vehicle 1 at time t) calculated by the position candidate calculation unit 11 on the acquired map. . In an actual apparatus, the process of overlapping the map and the position candidate does not have to be performed.

  For example, the matching unit 12 leaves (that is, extracts) position candidates that can be logically explained from the history of the inner-field sensor observation values (that is, the movement history of the vehicle 1), while from the history of the inner-field sensor observation values. Exclude position candidates that cannot be explained. As a result, for example, position candidates that overlap with sidewalks and structures, and position candidates on the opposite lane beyond the median are excluded.

  Such an operation of the matching unit 12 can be realized by the following processing, for example. Assuming that the inner-field sensor observation values have variations according to the normal distribution, the likelihood distribution is obtained (see concentric circles in FIG. 3B). Next, of the obtained likelihood distribution, for example, the likelihood corresponding to a sidewalk or a structure on the map is changed to zero. Based on the likelihood indicated by the likelihood distribution, a matching rate, which is an index indicating the likelihood of each of the plurality of position candidates, is calculated. Then, position candidates whose matching rate is larger than a predetermined value are extracted, while position candidates whose matching rate is smaller than the predetermined value are excluded. When the matching rate is equal to the predetermined value, it can be included in either one. More specifically, the operation of the matching unit 12 can be realized by, for example, a particle filter.

  The position candidates extracted by the matching unit 12 are hereinafter referred to as “probable position candidates” as appropriate.

3. The operation of the correctness determination correctness determination unit 13 will be described with reference to FIG. FIG. 4 is a conceptual diagram showing the concept of correct / incorrect determination.

  The correctness determination unit 13 classifies the plurality of position candidates calculated by the position candidate calculation unit 11 separately from the matching unit 12 described above. Specifically, the correctness determination unit 13 first performs map matching using the external sensor observation values (here, point clouds) acquired by the external sensor 20 and the point cloud map acquired from the map DB 40, Matching data indicating the result is generated (see FIG. 4A).

  The external sensor observation value at time t depends on the actual position of the vehicle 1 at time t. Focusing on this point, map matching is a technique for obtaining the position of the vehicle 1 at time t by matching (that is, aligning) an external sensor observation value and a point cloud map. By the way, the position of the vehicle 1 at the time t obtained by map matching using the external sensor observation value at the time t and the point cloud map is caused by an error in the external sensor observation value, an error in the point cloud map, or the like. The actual position of the vehicle 1 is not necessarily shown. That is, the position of the vehicle 1 obtained by map matching is not always correct.

  The correctness determination unit 13 assumes that the position indicated by each of the plurality of position candidates is the position obtained by map matching, and the position indicated by each of the plurality of position candidates is the actual position of the vehicle 1 (that is, Correct / incorrect determination is performed to determine whether the value corresponds to (true value). “Corresponding to the actual position (that is, true value)” is not limited to the fact that the position indicated by the position candidate matches the actual position, and the position indicated by the position candidate is a predetermined range near the actual position. It also means being within (see “predetermined range A” in FIG. 4B).

  As described above, the external sensor observation value at time t depends on the actual position of the vehicle 1 at time t. For this reason, if the actual position of the vehicle 1 is the position indicated by the position candidate, the external sensor observation value should also correspond to the position indicated by the position candidate. In the above correct / incorrect determination, based on such a concept, a position candidate that does not contradict the external sensor observation value (or matching data) at time t acquired by the external sensor 20 is “correct” (that is, corresponds to the actual position). The position candidate inconsistent with the external sensor observation value is determined to be “false” (that is, does not correspond to the actual position). As a result of the correctness determination, a plurality of position candidates are classified.

  Such correctness determination can be realized by the following configuration, for example. For example, the external sensor observation value when the position corresponding to the true value is obtained by map matching and the external sensor observation value when the position deviating from the true value is obtained by map matching are used as learning data. A learned neural network is constructed in advance. The above-described correctness determination is performed by inputting the above-described matching data and the position indicated by each of the plurality of position candidates to the neural network constructed in this manner. The neural network outputs the result of correct / incorrect determination as a probability. That is, the neural network outputs a continuous value as a probability indicating a possibility of being “positive”, for example, instead of a binary value of “positive” or “false”. Here, it is assumed that the probability is higher as the position indicated by the position candidate is more likely to correspond to the actual position. The result of correctness / incorrectness determination is output to the reliability comparison unit 14 after being associated with the corresponding position candidate. Note that the result of the correctness / incorrectness determination may be a probability indicating the possibility of being “correct”, or may indicate “correct” or “false” determined based on the probability indicating the possibility of being “correct”. It may be information.

4). The reliability comparison reliability comparison unit 14 calculates the reliability of each of the plurality of position candidates. Specifically, the reliability comparison unit 14 first calculates the likelihood for the correct / incorrect determination result (that is, the probability of obtaining one result related to the correct / incorrect determination) using the following equation.

In the following equation, “d _t ” is the result of correctness determination at time t for one position candidate among the plurality of position candidates, and “r _t ” is the reliability at time t for the one position candidate. “X _t ” is a position at the time t for the one position candidate, “z _t ” is an internal sensor value at the time t, and “m” is a map. “D _pos ” and “d _neg ” are arbitrary coefficients, “T” is a symbol indicating transposition, and “p _pos ” indicates that the correct / incorrect determination result is correct (that is, the correct / incorrect determination result is incorrect). (P _neg) is a probability distribution that determines the likelihood distribution with respect to the reliability, and “p _neg ” is a case where the correct / incorrect determination result is incorrect (ie, the correct / incorrect determination result is incorrect). This is a probability distribution that determines the likelihood distribution for the reliability. “X _t ”, “z _t ”, and “m” are vector quantities, and the other variables are scalar quantities.

The above “p _pos ” and “d _neg ” are specifically represented by the following equations. In the following equation, “a” and “b” are arbitrary coefficients, “B ()” is a beta function, and “unif” is a uniform distribution.

Next, the reliability comparison unit 14 estimates the position at time t for one position candidate among the plurality of position candidates based on the likelihood “p (d _t | r _t , x _t , z _t , m)”. The success probability (i.e., reliability r _t ) is obtained by, for example, a binary Bayes filter. When the binary Bayes filter is expressed by logit (ie, log odds), the following equation is obtained.

In the following expression, the likelihood “p (d _t | r _t , x _t , z _t , m)” is simply _expressed as “p (d _t | r _t )”. “R _t−1 ” is the reliability at time t−1 for the one position candidate. The reliability at time t-1 is the reliability of the position at time t-1 used to calculate the one position candidate at time t, as described in “1. Calculation of position candidates”. .

By transforming this equation, the reliability r _t at time t for one position candidate among the plurality of position candidates is obtained from the following equation. The reliability comparison unit 14 performs such reliability calculation for each of a plurality of position candidates.

Next, the reliability comparison unit 14 calculates the weight of each of the plurality of probable position candidates output from the matching unit 12 using the following formula based on the reliability calculated as described above. Here, “ω _t ” and “ω _t−1 ” are the weight at time t and the weight at time t−1, respectively. “I” is a subscript for identifying each of a plurality of probable position candidates, and indicates the “i” th probable position candidate among the plurality of probable position candidates. The weight at time t−1 is the position at time t−1 used to calculate the “i” th most likely position candidate at time t, as described in “1. Calculation of position candidates”. Is the weight.

Through a series of processes in the reliability comparison unit 14, the reliability and weight at each time t of a plurality of probable position candidates are calculated. Here, according to the research of the present inventors, the following has been found. That is, when the reliability at time t-1 (that is, the reliability of the position at time t-1 used for calculating the position candidate at time t) is relatively high, the position indicated by the position candidate is true. It is considered that the value is relatively close to the value (that is, the actual position of the vehicle 1) (that is, the possibility that the result of the correctness / incorrectness determination is “false” is extremely low). For this reason, even when the reliability at time t-1 is relatively high, the correctness / incorrectness determination result is “incorrect” (or when the probability indicating the possibility of being “correct” is relatively low). It is considered that the correctness determination by the correctness determination unit 13 is not appropriate (that is, there is an error in the correctness determination). In the present embodiment, as described above, the likelihood (or reliability or weight) for the result of correctness determination is calculated. For example, although the reliability at time t-1 is relatively high, correctness determination is not performed. The weight of a position candidate whose result is “false” (that is, there is a relatively high possibility of an error in correct / incorrect determination) is relatively small. For this reason, the influence on the position estimation mentioned later of the position candidate with a comparatively high possibility of an error in correct / incorrect determination can be suppressed.

5. The position estimation position estimation unit 15, for example, based on the weight at each time t of the plurality of likely position candidates calculated by the reliability comparison unit 14 and the position indicated by each of the plurality of likely position candidates, for example, The position of the vehicle 1 at time t is estimated by obtaining a weighted average of the positions indicated by the position candidates.

6). The reliability determination reliability determination unit 16 calculates the time t of the most probable position candidate having the largest weight based on the reliability and the weight at the time t of each of the plurality of possible position candidates calculated by the reliability comparison unit 14. Is determined as the reliability of the position of the vehicle 1 at time t estimated by the position estimation unit 15.

  In the self-position estimation apparatus 100, the position of the vehicle 1 is sequentially estimated by repeatedly performing the process related to position estimation.

(Technical effect)
One of the problems for realizing so-called fully automatic driving is to accurately estimate the position of the host vehicle. As a position measuring method, for example, a satellite positioning system (GNSS) that receives a radio wave from a satellite and estimates its own position, map matching, and the like have been proposed. However, the GNSS cannot in principle estimate the position in a section where a radio wave from a satellite cannot be received, such as a tunnel. In map matching, in order to improve position estimation accuracy, it is necessary to use a point cloud map composed of high-density point clouds each having highly accurate position information, and it is difficult to construct the point cloud map. is there.

  In the self-position estimation apparatus 100, the position transition of the vehicle 1 is treated stochastically in consideration of disturbance factors (see “1. Calculation of position candidates” described above), while the likelihood of each of a plurality of position candidates is increased. , By extracting a plurality of likely position candidates from a plurality of position candidates, the range in which the vehicle 1 will exist is narrowed (see “2. Matching” above). In the self-position estimation apparatus 100, the reliability of each of the plurality of position candidates is calculated separately from the extraction of the likely position candidates, and the weight of each of the plurality of likely position candidates is calculated based on the reliability. (See “4. Reliability Comparison” above.) Then, the self-position estimation apparatus 100 estimates the position of the vehicle 1 from the position indicated by each of the plurality of probable position candidates using the weights of each of the plurality of probable position candidates, and The reliability of the likely position candidate having the largest weight among the likely position candidates is determined as the reliability related to the estimated position of the vehicle 1 (see “5. Position estimation” and “6. Reliability determination described above”). "reference). That is, in the self-position estimation apparatus 100, the position of the vehicle 1 and its reliability are obtained simultaneously.

  If a relatively large number of position candidates are calculated (for example, 1000 or more), a relatively large number of position candidates are often collected near the actual position of the vehicle 1. For this reason, it can be expected that a relatively large number of possible position candidates are extracted. If a position indicated by a probable position candidate is used, it can be expected that a position with a relatively small degree of deviation from the actual position is estimated. That is, according to the self-position estimation apparatus 100, the position of the vehicle 1 can be estimated relatively accurately. Thus, according to the self-position estimation apparatus 100, the position of the vehicle 1 can be estimated relatively accurately by a relatively simple method.

  As described above, the “reliability” in the self-position estimation apparatus 100 indicates the success probability of position estimation. That is, the higher the reliability, the closer the estimated position is to the actual position (that is, the true value). In other words, the lower the reliability, the more the estimated position deviates from the actual position. Therefore, the self-position estimation apparatus 100 can easily know how accurate the estimated position is by referring to the reliability. That is, according to the self-position estimation apparatus 100, the relationship between the estimated position and the actual position (that is, the true value) can be easily known.

<Modification>
The self-position estimation device 100 may include a plurality of arithmetic devices 10. In this case, it is desirable that the external sensor observation values used in the correctness determination unit 13 are different from each other between one arithmetic device 10 and another arithmetic device 10. Specifically, for example, in the correctness determination unit 13 of one arithmetic device 10, an observation value acquired by a laser sensor as the external sensor 20 is used, and in the correctness determination unit 13 of another arithmetic device 10, the external sensor 20 It is desirable to use an observation value acquired from an image picked up by the camera. Then, the position of the vehicle 1 and its reliability are finally determined from the position and reliability obtained by each of the plurality of arithmetic devices 10 (for example, among the positions obtained by each of the plurality of arithmetic devices 10, The position with the highest reliability may be determined as the position of the vehicle 1).

  If fusion is performed by fusing a plurality of position estimation results or using a plurality of sensor information, the reliability of the entire system decreases. This is because the reliability of the entire system is defined as the multiplication of the reliability of a plurality of modules (corresponding to the arithmetic device 10) that perform position estimation. However, in this modification, since fusion is not performed, even if the self-position estimation apparatus 100 includes a plurality of arithmetic apparatuses 10, the reliability of the self-position estimation apparatus 100 does not decrease.

<Application example>
An application example of the self-position estimation apparatus 100 configured as described above will be described with reference to FIG. FIG. 5 is a block diagram illustrating a configuration of a self-position estimation apparatus according to an application example.

  In FIG. 5, the self-position estimation device 200 includes a failure determination unit 50 and a vehicle control device 60 in addition to the arithmetic device 10, the external sensor 20, the internal sensor 30, and the map DB 40 described above. The vehicle control device 60 is configured to be capable of automatically driving the vehicle 1 based at least on the position estimated by the arithmetic device 10.

  The failure determination unit 50 determines whether the position estimation by the calculation device 10 has failed based on the reliability of the position estimated by the calculation device 10. Specifically, the failure determination unit 50 determines that the position estimation has failed when the reliability is smaller than a predetermined threshold. On the other hand, when the reliability is greater than the predetermined threshold, the failure determination unit 50 determines that the position estimation has not failed (that is, succeeded). If the reliability is equal to the predetermined threshold value, it may be included in either case.

  When it is determined that the position estimation by the arithmetic device 10 has failed, the failure determination unit 50 may transmit a signal indicating that the vehicle 1 is to be stopped to the vehicle control device 60. If comprised in this way, since the vehicle 1 will be stopped automatically, even if it is a case where position estimation fails, safety can be ensured.

  Or when it determines with the position estimation by the arithmetic unit 10 having failed, the failure determination part 50 may transmit the signal which shows that a position estimation is performed to the arithmetic unit 10 again. In this case, for example, the computing device 10 estimates the position of the vehicle 1 again by deleting a position candidate having a relatively low weight or reliability from the likely position candidates and calculating a new position candidate. Good.

  Alternatively, when it is determined that the position estimation by the calculation device 10 has failed, the failure determination unit 50 determines the degree of divergence between the estimated position and the actual position from the reliability of the position estimated by the calculation device 10. It may be estimated. Then, the failure determination unit 50 may correct the position estimated by the arithmetic device 10 based on the estimated degree of deviation. The relationship between the reliability and the degree of deviation between the estimated position and the actual position may be obtained in advance by, for example, experiments or simulations.

  On the other hand, when it is determined that the position estimation by the calculation device 10 has not failed (that is, succeeded), the failure determination unit 50 sends a signal indicating the position estimated by the calculation device 10 to the vehicle control device 60. Send.

  Various aspects of the invention derived from the embodiments and modifications described above will be described below.

  A self-position estimation apparatus according to an aspect of the invention includes: a measuring unit that measures a surrounding environment of a moving body; a movement detecting unit that detects the movement of the moving body; a map database that stores map information; The position of the moving body at time 1 and (ii) the moving body detected by the motion detection means in a period from the first time to a second time after the first time. A plurality of position candidates indicating a temporary position of the mobile object at the second time based on the movement of the second position, and comparing the plurality of position candidates with the map information, Extracting means for extracting a plurality of probable position candidates from the position candidates, comparing the measurement result by the measuring means, the map information and the plurality of position candidates, and calculating the reliability of each of the plurality of position candidates First calculating means for A second calculating means for calculating a weight of each of the plurality of likely position candidates based on the reliability of each of the plurality of position candidates; a temporary position indicated by each of the plurality of likely position candidates; Estimating the position of the moving body based on the weight of each of the probable position candidates and determining the reliability of the estimated position based on the weight and the reliability of each of the plurality of probable position candidates Means.

  In the above-described embodiment, “vehicle 1” corresponds to an example of “moving body”, “external sensor 20” corresponds to an example of “measuring means”, and “inner sensor 30” corresponds to “motion detection means”. “Position candidate calculation unit 11” corresponds to an example of “generation unit”, “Matching unit 12” corresponds to an example of “extraction unit”, “correction determination unit 13” and “reliability” The “comparison unit 14” corresponds to an example of “first calculation unit”, the “reliability comparison unit 14” corresponds to an example of “second calculation unit”, and the “position estimation unit 15” and the “reliability determination unit 16”. "Corresponds to an example of" estimating means ".

  In the self-position estimation apparatus, a plurality of likely position candidates are extracted by comparing a plurality of position candidates with map information. As a result, the range in which the moving body will actually exist is narrowed down. In the self-position estimation apparatus, the positioning position acquired from the measurement result by the measuring means and the map information is compared with the plurality of position candidates, and the reliability of each of the plurality of position candidates is calculated. Here, the reliability indicates the success probability of calculation (or estimation) related to the temporary position indicated by each position candidate.

  In the self-position estimation device, the weight of each of a plurality of likely position candidates is calculated based on the reliability of each of the plurality of position candidates, and then the position of the moving object is estimated using the calculated weight. And the reliability of the estimated position is determined. Since the reliability indicates the success probability as described above, if the reliability is relatively high, it can be said that the actual position (that is, the true value) of the moving object has been successfully estimated, but the reliability is compared. If it is low, it can be said that the actual position of the moving object has not been estimated successfully (that is, failed). Therefore, according to the self-position estimation apparatus, the relationship between the estimation result and the true value can be recognized.

  The present invention is not limited to the above-described embodiment, and can be appropriately changed without departing from the spirit or idea of the invention that can be read from the claims and the entire specification, and self-position estimation with such a change. The apparatus is also included in the technical scope of the present invention. For example, the self-position estimation apparatus according to the present invention is applicable not only to a vehicle but also to a moving body such as a robot capable of autonomous traveling.

  DESCRIPTION OF SYMBOLS 1 ... Vehicle, 10 ... Arithmetic unit, 20 ... External sensor, 30 ... Internal sensor, 40 ... Map database, 50 ... Failure determination part, 60 ... Vehicle control apparatus, 100, 200 ... Self-position estimation apparatus

Claims (1)

Measuring means for measuring the surrounding environment of the moving body;
Motion detection means for detecting motion of the moving body;
A map database for storing map information;
(I) the position of the moving body at a first time, and (ii) detected by the motion detection means in a period from the first time to a second time after the first time. Generating means for generating a plurality of position candidates indicating the temporary position of the moving body at the second time based on the movement of the moving body;
An extraction means for comparing the plurality of position candidates with the map information and extracting a plurality of likely position candidates from the plurality of position candidates;
A first calculation unit that compares the measurement result of the measurement unit, the map information, and the plurality of position candidates, and calculates the reliability of each of the plurality of position candidates;
Second calculation means for calculating a weight of each of the plurality of likely position candidates based on the reliability of each of the plurality of position candidates;
Based on the temporary position indicated by each of the plurality of probable position candidates and the weight of each of the plurality of probable position candidates, the position of the moving object is estimated, and the weight of each of the plurality of probable position candidates Estimating means for determining reliability of the estimated position based on reliability;
A self-position estimation apparatus comprising: