Program, information processing apparatus, and information processing method

Abstract

To calculate an absolute position of a moving body on a map with higher accuracy.SOLUTION: A program of an embodiment causes a computer to function as a section calculation unit and a second absolute position calculation unit. The section calculation unit calculates, from a first absolute position of a moving object based on a GlobalPositioningSystem (GPS) and a relative position of the moving object based on information other than the GPS, an error variation indicating a variation of an error between a position indicated by the first absolute position and a position indicated by the relative position, and calculates at least one section on a map in which the error variation is equal to or less than a first threshold value. The second absolute position calculation unit calculates the second absolute position so that a constant difference loss for causing a difference between the first absolute position and the second absolute position on the map to approach a constant for each section becomes smaller.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] FIELD Embodiments of the present invention relate to a program, an information processing device, and an information processing method. [Background technology]

[0002] In order to visualize infrastructure maintenance and inspections such as defects on roads, as well as the movement of people and objects, technology is being developed to collect driving data from mobile objects equipped with sensing devices such as cameras or LiDAR (Light Detection and Ranging) and accurately estimate the location of the above-mentioned objects on a map from the driving data. Using a sensing device such as a camera or LiDAR on a mobile object makes it possible to calculate the relative position from the mobile object to the object, but it is not possible to determine the object's absolute position on a map (latitude, longitude, etc.). To determine the absolute position of the object, it is first necessary to determine the absolute position of the mobile object. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7328178 Summary of the Invention [Problem to be solved by the invention]

[0004] However, with conventional techniques, it has been difficult to calculate the absolute position of a moving object on a map with high accuracy. [Means for solving the problem]

[0005] According to an embodiment, the program causes a computer to function as a section calculation unit and a second absolute position calculation unit. The section calculation unit calculates an error variation indicating a variation in error between a position indicated by a first absolute position of a moving object based on a Global Positioning System (GPS) and a relative position of the moving object based on information other than the GPS, and calculates at least one section on a map in which the error variation is equal to or less than a first threshold. The second absolute position calculation unit calculates the second absolute position so that a constant difference loss for making the difference between the first absolute position and a second absolute position on the map closer to a constant for each section is smaller. [Brief explanation of the drawings]

[0006] [Figure 1] FIG. 2 is a diagram showing an example of the functional configuration of the information processing apparatus according to the embodiment. [Figure 2] FIG. 1 is a diagram showing an overview of route matching. [Figure 3] 5A to 5C are diagrams for explaining an example of processing (when landmarks are used) performed by the corrected position acquisition unit 102 according to the embodiment. [Figure 4A] 5 is a diagram for explaining an example of processing (when lane numbers are used) performed by the corrected position acquisition unit 102 according to the embodiment. FIG. [Figure 4B] FIG. 2 is a diagram showing an example of a camera image according to an embodiment. [Figure 5] 6A and 6B are diagrams for explaining an example of processing by a section calculation unit according to the embodiment; [Figure 6] FIG. 4 is a diagram showing an example of the functional configuration of a second absolute position calculation unit according to the embodiment. [Figure 7] FIG. 10 is a diagram for explaining a constant error loss according to the embodiment. [Figure 8] 1 is a flowchart illustrating an example of an information processing method according to an embodiment. [Figure 9] FIG. 1 is a diagram illustrating an example of a hardware configuration of an information processing apparatus according to an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0007] Hereinafter, embodiments of a program, an information processing device, and an information processing method will be described in detail with reference to the accompanying drawings.

[0008] A common method is to use GPS to obtain the absolute position of a moving object such as a car. However, low-cost GPS installed in devices such as drive recorders has low positioning accuracy due to its standalone positioning, resulting in errors of 1 to 10 meters. There is also a technology called route matching (used in car navigation systems, etc.) that uses GPS and road information to correct the GPS value to the nearest road, but the widely used road information only includes road centerlines, making it impossible to estimate the vehicle's position with high accuracy.

[0009] Conventional self-estimation requires the acquisition of a corrected position every time the environment or the vehicle's behavior changes. Since it is not possible to control conditions such as changes in the vehicle's behavior and the number of GPS positioning satellites, it is necessary to install many landmarks. However, installing many such landmarks over a wide area is costly and unrealistic.

[0010] In the following embodiment, the relative position calculated from camera images captured by an on-board camera, road information such as road centerlines, and GPS values ​​are combined in offline processing, enabling lane-level map position estimation from a smaller number of GPS corrected positions acquired independently of changes in the environment or the vehicle itself.

[0011] [Example of functional configuration] 1 is a diagram illustrating an example of the functional configuration of an information processing device 100 according to an embodiment. The information processing device 100 according to the embodiment is mounted on a moving body such as an automobile. The information processing device 100 according to the embodiment includes a first absolute position acquisition unit 101, a corrected position acquisition unit 102, a relative position acquisition unit 103, a section calculation unit 104, and a second absolute position calculation unit 105.

[0012] The first absolute position acquisition unit 101 acquires a first absolute position indicating the mobile object's own position on a map based on a GPS. The GPS-based own position on a map may be the GPS value used as is, or may be an absolute position obtained by correcting the GPS value using road information including at least one of the number of lanes, lane centerlines, and road width. An example of correction using road information is route matching used in car navigation systems, etc.

[0013] Figure 2 shows an overview of route matching. Route matching is a technology that uses time-series GPS to correct the vehicle's position to the nearest road. Route matching generally only moves the vehicle closer to the road center line 1, so it cannot estimate the vehicle's position at the lane level, such as which lane it is traveling in. Furthermore, even if the vehicle changes lanes, the corrected position after route matching remains straight along the road center line 1.

[0014] The absolute position corrected by such a technique as route matching may be acquired by the first absolute position acquisition unit 101. The information processing device 100 according to the embodiment receives a self-position (first absolute position) having an error as input, and ultimately outputs a self-position (second absolute position) with higher accuracy than the first absolute position.

[0015] Returning to FIG. 1, the corrected position acquisition unit 102 corrects the first absolute position for one or more times to acquire the corrected position of the first absolute position.

[0016] Next, a method for acquiring the corrected position by the corrected position acquisition unit 102 will be described in detail. In this embodiment, it is sufficient that there is one or more corrected positions. Since the corrected position is an absolute position on a map that is acquired using means other than GPS, various acquisition methods are possible. Four examples of acquisition methods will be introduced below.

[0017] (1) When using a landmark The first example of the corrected position acquisition method is a method of acquiring a corrected position of the first absolute position by using the absolute positions of landmarks registered on a map and a camera image and associating the landmarks on the camera image with the landmarks on the map. Here, the landmarks are often objects that appear more frequently in the environment where the position is estimated. Examples of landmarks include traffic lights, signs, crosswalks, white lines, stop lines, and buildings.

[0018] 3 is a diagram illustrating an example of processing (using landmarks) by the corrected position acquisition unit 102 according to the embodiment. The corrected position acquisition unit 102 uses a camera image 301, a three-dimensional point cloud 303, a GPS value 307, and absolute positions of landmarks 308 registered in advance on a map in processing to acquire a corrected position.

[0019] For example, the 3D point cloud 303 is a point cloud acquired by a laser sensor such as LiDAR. Alternatively, the 3D point cloud 303 is a 3D point cloud of the surrounding environment calculated from a camera image using Visual SLAM (Simultaneous Localization and Mapping). Alternatively, the 3D point cloud 303 is a 3D point cloud obtained from the camera 306 when the camera 306 is a stereo camera.

[0020] The position of the three-dimensional point cloud 303 is expressed on a coordinate system with the position of the camera 306 (or LiDAR sensor) as the origin at the time included in the GPS value 307.

[0021] First, the corrected position acquisition unit 102 detects and tracks the position of a landmark on the camera image 301 using a detector that detects (recognizes) an object on the camera image 301, and identifies the area of ​​the landmark 302 on the image.

[0022] Next, the corrected position acquisition unit 102 uses the area of ​​the landmark 302 on the camera image 301 and the 3D point group 303 to extract a 3D point group 304 included in the detection area on the camera image 301. Then, the corrected position acquisition unit 102 obtains a representative point 305 of the 3D point group 304. In this case, the representative point may be the average value of the 3D point group 304 or the median value obtained for each coordinate axis.

[0023] Next, the corrected position acquisition unit 102 converts the three-dimensional position (representative point 305) of the landmark into an initial position 309, which is an absolute position on the map, using a GPS value 307 at a certain time. The representative point 305 is a value on a coordinate system whose origin is the position of the camera 306 (or LiDAR sensor) at the time indicated by the GPS value 307, and therefore is a relative position from the camera 306 (or LiDAR sensor). The corrected position acquisition unit 102 determines the initial position 309 of the landmark 302 on the map from the relative position obtained from the camera image 301 and the absolute position indicated by the GPS value 307. The corrected position acquisition unit 102 associates the closest landmarks of the same type with each other based on the initial position 309 of the landmark 302 and the absolute position of a pre-registered landmark 308. This determines an association 310 between the landmark 308 on the map and the landmark 302 on the camera image 301.

[0024] Next, the corrected position acquisition unit 102 obtains a corrected position of the first absolute position (GPS value 307) from this correspondence 310. Specifically, when there are only a few correspondences, such as one, the corrected position acquisition unit 102 moves only the first absolute position while maintaining the relative positional relationship between the initial position 309 of the landmark 302 obtained from the camera image 301 and the GPS value 307 and the first absolute position (GPS value 307). Then, the corrected position acquisition unit 102 determines the position where the positional error between the pre-registered landmark 308 and the initial position 309 of the landmark 302 is smallest as the final corrected position of the first absolute position. For example, when there is only one correspondence, the corrected position is obtained by adding a difference vector between the initial position 309 of the landmark 302 and the position of the pre-registered landmark 308 to the first absolute position (GPS value 307).

[0025] When there are multiple correspondences, the corrected position acquisition unit 102 moves at least one of the first absolute position and the orientation at the first absolute position while maintaining the relative positional relationship between the initial position 309 of the landmark 302 and the first absolute position (GPS value 307). Then, the corrected position acquisition unit 102 uses an ICP algorithm or the like to find the position and orientation that minimizes the positional error between the pre-registered landmark 308 and the initial position 309 of the landmark 302. The position obtained in this way becomes the final corrected position of the first absolute position.

[0026] (2) When using lane numbers The second example of the corrected position acquisition method is a method of calculating the corrected position using map information in which road information is registered and the lane number of the vehicle that is acquired from a camera image. Here, the map information includes at least one of the number of lanes, center lines, and road width.

[0027] 4A is a diagram for explaining an example of processing (using lane numbers) by the corrected position acquisition unit 102 according to the embodiment. The second corrected position acquisition method includes the following four steps.

[0028] First, the corrected position acquisition unit 102 corrects the first absolute position 400 based on the GPS value to an absolute position 402 on the lane center line 401 by route matching.

[0029] Next, the corrected position acquisition unit 102 obtains the absolute position 402 of the left edge of the lane (left edge of the driving lane). Specifically, the corrected position acquisition unit 102 obtains the distance from the road centerline to the left edge of the lane from the number of lanes included in the map and the road width. The corrected position acquisition unit 102 then calculates the point on a line that passes through the absolute position 402 on the road centerline 401 and is perpendicular to the road centerline 401, which is the distance from the absolute position 402 to the left edge of the lane, as the absolute position 402 of the left edge of the lane.

[0030] As for the road width, a value included in the map information may be used, or a generally defined road width value may be used.

[0031] Next, the corrected position acquisition unit 102 determines which lane from the left the vehicle is traveling in from the camera image at the time the first absolute position 400 based on the GPS value was acquired, and acquires the lane number indicating the identified lane.

[0032] FIG. 4B is a diagram illustrating an example of a camera image according to an embodiment. In the example of the camera image in FIG. 4B, the corrected position acquisition unit 102 acquires lane number 2, which indicates the second lane. For example, the corrected position acquisition unit 102 may receive an operation input indicating the lane number from a user via an input device. Alternatively, for example, the corrected position acquisition unit 102 may acquire the lane number from the camera image using image processing that employs lane detection. Alternatively, for example, the corrected position acquisition unit 102 may acquire the lane number from the camera image using a deep neural network that answers questions about the camera image.

[0033] 4A , the corrected position acquisition unit 102 calculates a final corrected position 404 of the first absolute position 400 from the lane width and the lane number. Specifically, the corrected position acquisition unit 102 calculates the distance from the absolute position 403 of the left edge of the lane to the corrected position 404 by multiplying the lane width by the lane number. The corrected position acquisition unit 102 then determines, as the corrected position 404 of the first absolute position 400, a point on a line that passes through the absolute position 403 of the left edge of the lane and is perpendicular to the road centerline 401 and that is a distance of the lane width by the lane number from the absolute position 403 of the left edge of the lane. Here, the lane width may be a value included in the map information, or a commonly defined value for the road width may be used.

[0034] The above-described method for obtaining the corrected position is an example, and the position of the right edge of the lane and the lane number from the right edge may also be used.

[0035] (3) When using an image associated with an absolute location on a map A third example of a method for obtaining a corrected position is a method for calculating a corrected position by associating a camera image with an image that is associated with an absolute position on a map.

[0036] In the embodiment, the corrected position acquisition unit 102 refers to a database that stores an image and the position on a map of a place indicated by the image. The database may be provided in the information processing device 100, or may be provided in a server device that communicates with the information processing device 100. The image stored in the database may be a camera image captured in the past by an in-vehicle camera or the like, or may be a satellite image centered on a position on a map registered in the database.

[0037] First, the corrected position acquisition unit 102 uses a first absolute position based on a GPS value and a preset distance threshold to extract images of registered locations in a database that are within the distance threshold from the first absolute position. Then, the corrected position acquisition unit 102 selects from the database an image that is most similar to the camera image when the first absolute position was acquired (the similarity with the camera image is equal to or greater than a second threshold), and calculates a relative position from the selected image. For example, this series of processes may be realized using a deep neural network or the like. Also, for example, the process of selecting highly similar images may be realized using a deep neural network, and the process of calculating a relative position from the selected image may be realized using a geometric calculation using feature point matching.

[0038] The correction position acquisition unit 102 calculates the correction position of the first absolute position by adding the relative position between the first absolute position and the absolute position of the selected image to the absolute position of the selected image registered in the database.

[0039] (4) Other methods for obtaining corrected position As another example of a method for acquiring the corrected position, a position acquired from a beacon installed on a road may be used as the corrected position of the first absolute position based on the GPS value. Specifically, the corrected position acquisition unit 102 acquires the corrected position of the first absolute position through road-to-vehicle communication between a beacon transmitter (access point device) installed in the environment and a beacon receiver installed in the vehicle. For example, the road-to-vehicle communication is wireless LAN (Local Area Network) communication or Bluetooth communication.

[0040] As another example of a method for acquiring the corrected position, an operation input indicating the corrected position of the first absolute position may be received from the user via an input device. Specifically, the operation input indicating the corrected position of the first absolute position may be received from the user based on clues such as a camera image when the first absolute position was acquired and common sense such as driving on the center of the lane.

[0041] Although various corrected position acquisition methods (1) to (4) for the first absolute position have been described, the corrected position used to calculate the second absolute position is not limited to that acquired by one method. That is, at least one corrected position acquired by the above-described corrected position acquisition methods (1) to (4) may be used.

[0042] 1, the relative position acquisition unit 103 acquires at least one of the time-series relative position and orientation from information other than GPS. Here, the relative position and orientation are acquired from, for example, the amount of change indicating the position and orientation of the vehicle in adjacent frames of a camera image.

[0043] Other information besides GPS can be obtained from devices such as camera images, wheel speed, gyro sensors, or LiDAR sensors. For example, the relative position and orientation of the vehicle can be determined from camera images or LiDAR using SLAM technology.

[0044] The interval calculation unit 104 calculates an interval with small error fluctuation, which indicates fluctuation of the error between the first absolute position and the relative position. Two calculation methods will be described as examples of a method for calculating an interval with small error fluctuation.

[0045] (1) How to compare trajectories The first method for calculating the section with low error variation is to compare the trajectory of the relative position with the trajectory of the first absolute position, and calculate the section with consecutive times when the position error or orientation error of the two trajectories is below a threshold as the section with low error variation.

[0046] 5 is a diagram illustrating an example of processing by the interval calculation unit 104 according to the embodiment. The interval calculation unit 104 calculates an interval in which the error between the trajectory of the relative position and the trajectory of the first absolute position is equal to or smaller than a threshold value.

[0047] The relative position trajectory is a trajectory obtained by overlapping the relative positions in the time series in the time direction. The first absolute position trajectory is a collection of first absolute positions at multiple times. In the trajectory comparison process, both the first absolute positions and the relative positions may be used, or either one of them may be used.

[0048] When comparing two trajectories, the section calculation unit 104 first divides the trajectories into predetermined moving distances or time intervals.

[0049] Next, the interval calculation unit 104 aligns the divided trajectories. In aligning, a transformation (rotation, translation, or scale) is calculated that minimizes the distance between trajectories at the same time while maintaining the shape of the trajectories. By performing the calculated transformation on one of the trajectories, the two trajectories can be compared in the same coordinate system. The interval calculation unit 104 compares the trajectories after transformation, and calculates an interval in which the position error or orientation error is equal to or less than a predetermined first threshold as an interval with little error variation.

[0050] The section calculation unit 104 performs the above process for each of the divided sections of the trajectory. In the example of Fig. 5, sections 501 and 502 of the divided trajectory are calculated as sections with little error variation. As shown in Fig. 5, sections with a high degree of agreement between the position of the SLAM trajectory and the position of the GPS trajectory are calculated as sections with little error variation.

[0051] (2) Methods using relative position changes, GPS accuracy, etc. The second method of calculating sections with low error variation is to calculate sections where the change in at least one of the speed and direction of a moving object is below the third threshold, sections where the GPS accuracy is above the fourth threshold, or sections where the change in the number of positioning satellites is below the fifth threshold as sections with low error variation (sections on a map where the error variation is below the first threshold).

[0052] GPS has a characteristic that the GPS error does not fluctuate significantly (the GPS error is below a threshold) when the structural environment is constant, the vehicle behavior change (for example, speed change or right / left turns) is below a threshold, and the combination of satellites used for positioning remains the same. Utilizing this characteristic, a section with a high degree of match is determined. Specifically, the section calculation unit 104 calculates the amount of change in at least one of the speed and direction of the moving object from the relative positions calculated at multiple times, and calculates a section where the amount of change is below a predetermined third threshold as a section with little error variation.

[0053] Furthermore, for example, since accuracy can be obtained from GPS, a section where accuracy is equal to or greater than a predetermined fourth threshold may be calculated as a section where error fluctuation is small.

[0054] Furthermore, for example, since the number of positioning satellites can be obtained from GPS, a section in which the combination or number of satellites used for positioning remains unchanged may be calculated as a section with little error variation.

[0055] Two methods for calculating intervals with small error fluctuations have been introduced above, but one or more of the intervals obtained by the above calculation methods can be used. Also, intervals obtained by multiple calculation methods can be ANDed or ORed.

[0056] 1, the second absolute position calculation unit 105 includes a position calculation unit 106 and a loss calculation unit 107. The loss calculation unit 107 includes a constant differential loss calculation unit 108, a corrected position loss calculation unit 109, and a relative change loss calculation unit 110.

[0057] The second absolute position calculation unit 105 calculates a second absolute position including at least one of the own position and orientation on the map at the acquisition time of the corrected position and at a time other than the acquisition time, using the first absolute position, the relative position, the corrected position for one or more times, and the section with small error fluctuation.

[0058] 6 is a diagram illustrating an example of the functional configuration of the second absolute position calculation unit 105 according to the embodiment. The second absolute position calculation unit 105 according to the embodiment includes a position calculation unit 106 and a loss calculation unit 107.

[0059] The loss calculation unit 107 calculates the loss used to calculate the absolute position using the first absolute position, the relative position, the corrected position for one or more times, and the section with small error fluctuations (the section with a high degree of coincidence between the SLAM trajectory position and the GPS trajectory position).The position calculation unit 106 then calculates a second absolute position where the loss approaches a minimum.

[0060] Specifically, the loss calculation unit 107 includes a constant differential loss calculation unit 108 , a corrected position loss calculation unit 109 , and a relative change loss calculation unit 110 .

[0061] The constant difference loss calculation unit 108 calculates the loss used in the process of making the difference between the second absolute position and the first absolute position approach a constant for each section where the error fluctuation is small.

[0062] The corrected position loss calculation unit 109 calculates a loss used in the process of bringing the second absolute position at the time when the above-mentioned corrected position is acquired closer to the above-mentioned corrected position.

[0063] The relative change loss calculation unit 110 calculates the loss used in the process of bringing the time series change in the second absolute position closer to the time series change in the relative position described above.

[0064] It is sufficient that at least one of the constant differential loss calculation unit 108, the corrected position loss calculation unit 109, and the relative change loss calculation unit 110 is provided.

[0065] Each loss will be specifically described below.

[0066] (1) Constant differential loss The constant difference loss calculation unit 108 receives as input the above-mentioned first absolute position and the above-mentioned section with small error fluctuations, and calculates a constant difference loss to be used in position calculation for making the difference between the first absolute position and the estimated second absolute position approach a constant for each section.

[0067] The constant difference loss is a loss that makes the deviation between the GPS trajectory and the estimated trajectory constant by taking advantage of the fact that the GPS error is steady (constant) when the structural environment is constant, there is little change in the vehicle's behavior (change in the vehicle's behavior is below a threshold), and the combination of satellites used for positioning does not change.

[0068] FIG. 7 is a diagram illustrating the constant error loss according to an embodiment. Conventionally, trajectories are connected between absolute positions given by the above-described corrected positions by optimization using relative change loss or the like. However, drift error occurs in the relative trajectory due to accumulation over time, which can cause the trajectory to make small or large turns when turning, resulting in a large error in the estimated absolute position. Adding this constant error loss is expected to steady the deviation of the estimated trajectory 503 (trajectory of the estimated second absolute position) from the GPS trajectory, thereby preventing the trajectory from making small or large turns when turning.

[0069] Furthermore, while real-time processing cannot utilize information from the point onward where the GPS error characteristics change, the embodiment assumes offline processing, and information from the point onward where the GPS error characteristics change can also be utilized. Therefore, by applying this constant error loss to multiple sections with small error fluctuations (sections 501 and 502 in the example of FIG. 7 where the degree of match between the SLAM trajectory and the GPS trajectory is high), it becomes unnecessary to acquire corrected positions at the timing when the error characteristics change, and only a small number of GPS corrected positions (one or more) will suffice.

[0070] This constant error loss is not used for all frames and GPS tracks, but only for sections where the error fluctuations are small. When the GPS error becomes unstable due to environmental changes, or when the GPS value is unstable, the constant differential loss has a negative effect, so it is only used for sections limited to those where the error fluctuations are small.

[0071] Specifically, the constant differential loss E bias is the residual e bias and variance Σ bias is defined by the following formula (1).

[0072]

number

[0073] Here, S is the number of intervals with small error fluctuations. bias is the position component of the estimated second absolute position (coordinate t on the xz plane (ground surface)) in the section where the error fluctuation is small. x and t z ) and the GPS value is designed to have a small error variance. First, the constant difference loss calculation unit 108 calculates a straight line (a s x+b s y+c s = 0) is calculated using the least squares method. s , b s and c s is a fixed variable.

[0074] Then, the constant difference loss calculation unit 108 calculates the position components (t g,x ,t g,z ), the distance from each line is calculated, and the bias error is defined as in the following equations (2) to (4).

[0075]

number

[0076]

number

[0077]

number

[0078] Here, G is the number of GPS values ​​in one section with small error fluctuation. The constant difference loss calculation unit 108 calculates one error e g Calculate the error e g is the distance between the point indicated by the position component of the second absolute position and the line passing through each GPS value (point on the xz plane (ground)) (the distance d g ) and G d g is the difference from the average.

[0079] This error e g By setting this as the loss of position estimation, the error between the second absolute position and the GPS value within the interval can be made constant (the variance becomes smaller).

[0080] (2) Corrected position loss The corrected position loss calculation unit 109 calculates the corrected position loss used in the process of moving the second absolute position at the time when the corrected position was acquired closer to the corrected position. The error indicated by this corrected position loss is expected to have the effect of correcting errors in the GPS value. In the embodiment, by introducing a constant differential loss, the number of corrected positions can be small (one or more), and the timing of acquiring the corrected position is arbitrary and independent of the environment.

[0081] Specifically, the corrected position loss E anchor is the residual error e at each correction position n anchor and variance Σ anchor It is defined in the following format using and .

[0082]

number

[0083] where N is the number of correction positions obtained. n anchor is the corrected position x at time t t and the position component t of the estimated second absolute position t =(t x ,t y ,t z ) and is defined by the following equation (6):

[0084]

number

[0085] As described above, there are various methods for acquiring the corrected position. While some acquisition methods can stably obtain a highly accurate position, others have characteristics such as being prone to errors in the traveling direction of the vehicle or in the left-right direction. The corrected position loss calculation unit 109 may adjust the weight of the corrected position loss for the traveling direction of the moving body or the weight of the corrected position loss for the left-right direction of the moving body depending on the method for calculating the corrected position.

[0086] For example, (1) when landmarks (such as stop lines and pedestrian crossings) are used, errors in the left and right directions are likely to occur when detected from an image, since errors in the left and right directions are likely to occur. Also, for example, (2) when lane numbers are used, errors in the traveling direction are likely to occur because corrections are made in the left and right directions based on the position obtained by route matching. Therefore, the corrected position loss calculation unit 109 calculates the weight of loss (Σ anchor The adjustment method may be based on a design by the user, or may be adjusted during optimization by the position calculation unit 106.

[0087] (3) Relative Change Loss The relative change loss calculation unit 110 calculates the relative change loss used in the process of approximating the time series change of the second absolute position to the time series change of the relative position. This relative change loss is expected to have the effect of smoothly connecting the values ​​of the GPS corrected position, which are highly accurate but only a few points, using at least one of the relative position and orientation estimated by a camera, LiDAR, or the like, which are dense in time series.

[0088] Specifically, the relative change loss E rel is the residual e rel , variance Σ rel is defined by the following equation (7).

[0089]

number

[0090] residual e rel In explaining the above, first, the position component and the orientation component of the estimated second absolute position are expressed as T t and the change in position component and change in orientation component between frames in the above relative position are defined as ΔO t-1→t where T t and ΔO t-1→t is an element of the rigid transformation SE(3) in three-dimensional space.

[0091] residual e rel is calculated by converting the transformation representing each position and orientation from the rigid transformation SE(3) to the similarity transformation Sim(3). t element τ of the similarity transformation Sim(3) t and convert it to the relative position ΔO t-1→t The element ο of the similarity transformation Sim(3) t Convert to.

[0092] Sim(3) is a 7-dimensional [φ x ,φ y ,φ z ,t x ,t y ,t z,s] and can be converted to matrix form as τ = [sR|t]. When converting parameters from SE(3) to Sim(3), a scale is added to the estimated parameters. The residual is defined as follows:

[0093]

number

[0094] The relative change loss calculation unit 110 calculates the relative position ΔO between adjacent frames. t-1→t is the constraint (fixed) condition, and the residual e rel The second absolute position τ at which t-1 ,τ t where ξ(·) is the transformation from the matrix representation of the relative position and orientation to the seven-dimensional parameters of the similarity transformation Sim(3). b is the bias for adjusting the scale parameter.

[0095] In the embodiment, the loss is calculated by converting the absolute position and orientation into a similarity transformation, but the loss may be calculated by directly converting the absolute position and orientation into a rigid body transformation.

[0096] The position calculation unit 106 receives the above-mentioned losses as input and calculates the second absolute position. Specifically, the losses are added together and the second absolute position (at least one of the position component and the orientation component) that minimizes the added losses is calculated. The second absolute position is calculated using an algorithm such as the Levenberg-Marquardt method, which is a method for solving a nonlinear least-squares problem, or the Gauss-Newton method.

[0097] The position calculation unit 106 may calculate the second absolute position using some of the above-mentioned losses as input. For example, the position calculation unit 106 may calculate the second absolute position so that the constant difference loss for making the difference between the first absolute position and the second absolute position on the map closer to a constant for each section with small error fluctuations described above becomes smaller.

[0098] For example, the position calculation unit 106 may calculate the second absolute position so that the corrected position loss is smaller, further based on the corrected position loss for bringing the second absolute position at at least one time closer to the corrected position of the first absolute position obtained at at least one time.

[0099] For example, the position calculation unit 106 may calculate the second absolute position including at least one of the position and orientation of the moving body, further based on a relative change loss for making the time series change of the second absolute position closer to the time series change of the relative position, so that the relative change loss is smaller.

[0100] [Examples of information processing methods] 8 is a flowchart showing an example of an information processing method according to an embodiment. First, the first absolute position acquisition unit 101 acquires a first absolute position indicating the self-position of a mobile object on a map based on a GPS (step S1).

[0101] Next, the corrected position acquisition unit 102 corrects the first absolute position for one or more times to acquire a corrected position of the first absolute position (step S2).

[0102] Next, the relative position acquisition unit 103 acquires the time-series relative position from information other than the GPS (step S3).

[0103] Next, the interval calculation unit 104 calculates at least one interval with small error fluctuations that indicate fluctuations in the error between the first absolute position and the relative position, using the calculation method described above (step S4).

[0104] Next, the second absolute position calculation unit 105 calculates the second absolute position so that, for example, the difference between the first absolute position and the second absolute position on the map approaches a constant for each section calculated in step S4, a constant difference loss is reduced (step S5).

[0105] As described above, in the information processing device 100 according to the embodiment, the section calculation unit 104 calculates an error variation between the position indicated by the first absolute position and the position indicated by the relative position, based on the first absolute position of the mobile object based on GPS and the relative position of the mobile object based on information other than GPS, and calculates at least one section on the map in which the error variation is equal to or less than a first threshold. Then, the second absolute position calculation unit 105 calculates the second absolute position so as to reduce a constant difference loss for making the difference between the first absolute position and the second absolute position on the map closer to a constant for each section.

[0106] As a result, the information processing device 100 of the embodiment can calculate the absolute position of a moving object on a map with higher accuracy. Specifically, since the processing of the information processing device 100 of the embodiment is offline processing, unlike online real-time processing, it is possible to use information after the point at which the GPS error characteristics change. Therefore, by applying a constant differential loss to multiple sections with small error fluctuations, it becomes unnecessary to acquire a corrected position at the timing when the error characteristics change, and only a small number of GPS corrected positions (one or more) are required. Furthermore, the timing at which the corrected position is acquired is arbitrary and does not depend on the environment.

[0107] Furthermore, when landmarks are used to obtain the corrected position, the number of landmarks to be installed can be reduced. Furthermore, when the corrected position is input by the user, fewer corrected positions need to be input than before, which reduces the user's workload.

[0108] Furthermore, when using camera images, LiDAR, wheel speeds, etc. to calculate the relative position, drift errors in the relative position due to time-series accumulation can also be reduced at the same time.

[0109] Finally, an example of the hardware configuration of the information processing apparatus 100 according to the embodiment will be described.

[0110] [Example of hardware configuration] 9 is a diagram illustrating an example of a hardware configuration of the information processing device 100 according to the embodiment. The information processing device 100 according to the embodiment includes a processor 201, a main storage device 202, an auxiliary storage device 203, a display device 204, an input device 205, and a communication device 206. The processor 201, the main storage device 202, the auxiliary storage device 203, the display device 204, the input device 205, and the communication device 206 are connected via a bus 210.

[0111] Note that some of the above configurations may not be provided in the information processing device 100. For example, if the information processing device 100 can use the input function and display function of an external device, the information processing device 100 may not be provided with the display device 204 and the input device 205.

[0112] The processor 201 executes a program read from the auxiliary storage device 203 to the main storage device 202. The main storage device 202 is a memory such as a ROM and a RAM. The auxiliary storage device 203 is a HDD, a memory card, or the like.

[0113] The display device 204 is, for example, a liquid crystal display. The input device 205 is an interface for operating the information processing device 100. The display device 204 and the input device 205 may be realized by a touch panel or the like having a display function and an input function. The communication device 206 is an interface for communicating with other devices.

[0114] For example, the program executed by the information processing device 100 is provided as a computer program product in the form of a file in an installable or executable format, recorded on a computer-readable storage medium such as a memory card, hard disk, CD-RW, CD-ROM, CD-R, DVD-RAM, and DVD-R.

[0115] Furthermore, for example, the program executed by the information processing device 100 may be stored on a computer connected to a network such as the Internet, and may be provided by being downloaded via the network.

[0116] Furthermore, for example, the program executed by the information processing device 100 may be provided via a network such as the Internet without being downloaded. Specifically, the information processing may be performed by a so-called ASP (Application Service Provider) type service that realizes processing functions by issuing an execution instruction and obtaining the results from a server computer without transferring the program.

[0117] Furthermore, for example, the program for the information processing device 100 may be provided by being pre-installed in a ROM or the like.

[0118] The program executed by the information processing device 100 has a modular configuration that includes functions that can be realized by the program among the above-mentioned functional configurations. As for each function, the processor 201 reads the program from a storage medium and executes it, and the above-mentioned functional blocks are loaded onto the main storage device 202, in terms of actual hardware. In other words, the above-mentioned functional blocks are generated on the main storage device 202.

[0119] Note that some or all of the above-described functions may be realized by hardware such as an integrated circuit (IC) rather than by software.

[0120] Furthermore, each function may be realized using a plurality of processors 201, in which case each processor 201 may realize one of the functions, or may realize two or more of the functions.

[0121] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be embodied in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included within the scope and spirit of the invention, and are also included in the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0122] 100 Information processing device 101 First absolute position acquisition unit 102 Correction position acquisition unit 103 Relative position acquisition unit 104 Section calculation unit 105 Second absolute position calculation unit 106 Position calculation section 107 Loss calculation section 108 Constant difference loss calculation section 109 Correction position loss calculation unit 110 Relative change loss calculation unit 201 processor 202 Main storage 203 Auxiliary storage device 204 Display device 205 Input Device 206 Communication Equipment 210 Bus

Claims

1. Computer, a section calculation unit that calculates, from a first absolute position of a moving object based on a GPS (Global Positioning System) and a relative position of the moving object based on information other than the GPS, an error variation that indicates a variation in error between a position indicated by the first absolute position and a position indicated by the relative position, and calculates at least one section on a map in which the error variation is equal to or less than a first threshold value; a second absolute position calculation unit that calculates the second absolute position so that a constant difference loss for making a difference between the first absolute position and the second absolute position on the map approach a constant for each section becomes smaller; A program to function as a

2. the second absolute position calculation unit calculates the second absolute position further based on a corrected position loss for making the second absolute position at the at least one time closer to a corrected position of the first absolute position obtained at the at least one time, so that the corrected position loss becomes smaller. The program according to claim 1.

3. The computer a corrected position acquisition unit that calculates the corrected position by associating the landmarks on the camera image with the landmarks on the map based on absolute positions of the landmarks registered on the map and a camera image taken by a camera of the moving body; The program according to claim 2, further functioning as:

4. The computer a corrected position acquisition unit that calculates the corrected position from map information in which road information including at least one of the number of lanes, lane center lines, and road width is registered, a lane number identified from a camera image captured by a camera of the moving body, and a lane center line of the road obtained by route matching the first absolute position; The program according to claim 2, further functioning as:

5. The lane number identified from the camera image is identified by an operation input indicating the lane number from a user, an image recognition process that recognizes lanes from the camera image, or a deep neural network that answers questions about the camera image. The program according to claim 4.

6. The computer a corrected position acquisition unit that compares a camera image captured by a camera of the mobile body with an associated image associated with an absolute position on the map, and calculates the corrected position based on the absolute position on the map associated with the associated image whose similarity to the camera image is equal to or greater than a second threshold; The program according to claim 2, further functioning as:

7. The computer a corrected position acquisition unit that calculates the corrected position based on a position acquired using a beacon installed on a road or an operation input from a user indicating a position; The program according to claim 2, further functioning as:

8. the second absolute position calculation unit adjusts a weight of the corrected position loss with respect to a traveling direction of the moving body or a weight of the corrected position loss with respect to a left-right direction of the moving body according to a method for calculating the corrected position. The program according to any one of claims 3 to 7.

9. the second absolute position calculation unit calculates the second absolute position further based on a relative change loss for making a time-series change in the second absolute position closer to a time-series change in the relative position so that the relative change loss is smaller; the second absolute position includes at least one of a position and an orientation of the moving body; The program according to claim 1.

10. The computer a relative position acquisition unit that calculates the relative position from at least one of a camera image captured by a camera of the moving body, a wheel speed of the moving body, a gyro sensor of the moving body, and a LiDAR (Light Detection and Ranging) sensor of the moving body; The program according to claim 9, further functioning as:

11. the section calculation unit compares the trajectory of the relative position with the trajectory of the first absolute position; the error variation is at least one of an error variation between a position of the trajectory of the relative position and a position of the trajectory of the first absolute position, and an error variation between a direction of the trajectory of the relative position and a direction of the trajectory of the first absolute position. The program according to claim 10.

12. The interval calculation unit calculating a change in at least one of the velocity and the orientation of the moving object from the relative positions calculated at a plurality of times, and calculating a section where the change is equal to or less than a third threshold as a section on the map where the error variation is equal to or less than the first threshold; The program according to claim 10.

13. The interval calculation unit A section where the accuracy of the GPS is equal to or greater than a fourth threshold, or a section where the change in the number of GPS positioning satellites is equal to or less than a fifth threshold, is calculated as a section on the map where the error variation is equal to or less than the first threshold. The program according to claim 10.

14. the first absolute position is an absolute position indicated by the GPS value, or an absolute position obtained by correcting the GPS value using road information including at least one of the number of lanes, lane centerlines, and road width; The program according to any one of claims 1 to 7.

15. a section calculation unit that calculates, from a first absolute position of a moving object based on a GPS (Global Positioning System) and a relative position of the moving object based on information other than the GPS, an error variation that indicates a variation in error between a position indicated by the first absolute position and a position indicated by the relative position, and calculates at least one section on a map in which the error variation is equal to or less than a first threshold value; a second absolute position calculation unit that calculates the second absolute position so that a constant difference loss for making a difference between the first absolute position and the second absolute position on the map approach a constant for each section becomes smaller; and An information processing device comprising:

16. an information processing device calculating, from a first absolute position of a moving object based on a GPS (Global Positioning System) and a relative position of the moving object based on information other than the GPS, an error variation indicating a variation in error between a position indicated by the first absolute position and a position indicated by the relative position, and calculating at least one section on a map where the error variation is equal to or less than a first threshold; a step of calculating the second absolute position by the information processing device so that a constant difference loss for making a difference between the first absolute position and the second absolute position on the map approach a constant for each section becomes smaller; An information processing method including:

Images