JP7344743B2 - Occupancy map creation method and occupancy map creation device - Google Patents

Description

The present invention relates to an occupancy map creation method and an occupancy map creation device.

In a system that provides driving support for a vehicle, an occupancy map is created that indicates the occupancy status of obstacles on the road surface in order to avoid obstacles such as structures and moving objects that exist on the road surface on which the vehicle is traveling. As occupancy maps, grid maps are known that constitute grids by dividing the road surface at predetermined intervals and indicate whether or not obstacles occupy the grids.

For example, according to the image generation device shown in Patent Document 1, the positions of a plurality of measurement points set in the surrounding environment of a vehicle using a distance sensor and distance measurement data that record the distance to the measurement points are used. A technique for generating an occupancy map is disclosed. According to this image generation device, a distance sensor acquires a plurality of distance measurement data indicating the direction of the obstacle and the distance to the obstacle, and by superimposing these distance measurement data, the occupancy of obstacles around the vehicle is determined. An occupancy map indicating the area is generated.

Japanese Patent Application Publication No. 2005-326944

According to the technique disclosed in Patent Document 1, an occupancy map is generated by superimposing a plurality of distance measurement data in a predetermined period. Therefore, especially when there is a moving object moving at low speed on the road surface, some of the superimposed ranging data may indicate an area where the moving object existed in the past. be. In such a case, there is a risk that the area will be marked as occupied by an obstacle, even though there is currently no moving object in the area and there is a low possibility that the area will interfere with vehicle travel.

The present invention has been made to solve the above problems, and provides an occupancy map creation method and an occupancy map creation device that improves the accuracy of determining the occupancy state of an obstacle. .

The occupancy map creation method of the present invention is a method of creating an occupancy map that shows the occupancy status of obstacles on the road surface around a vehicle. In the occupancy map creation method, the occupancy map creation device acquires the positions of a plurality of measurement points set around the vehicle as ranging data at a predetermined cycle, and acquires the positions of a plurality of measurement points set around the vehicle as distance measurement data from the current time to the first hour before. An occupancy map showing the occupied and unoccupied parts of obstacles on the road surface is generated using a plurality of distance measurement data during a period of Using distance measurement data, an unoccupied map showing the unoccupied parts of obstacles on the road surface is generated, and of the occupied parts shown on the occupancy map, the areas included in the unoccupied parts shown on the unoccupied map are Change to occupied part.

According to the occupancy map creation method of the present invention, it is possible to improve the accuracy of determining the occupancy state of an obstacle.

FIG. 1 is a schematic configuration diagram of a driving support system that executes the occupancy map creation method according to the first embodiment. FIG. 2 is an explanatory diagram showing acquisition control of ranging data. FIG. 3 is an explanatory diagram of a method for generating a grid map from ranging data. FIG. 4 is a diagram showing an example of the final grid map. FIG. 5 is a flowchart of final grid map generation control. FIG. 6 is a diagram illustrating an example of a region for which a final grid map is generated. FIG. 7 is a diagram showing a short-term occupancy grid map and a long-term occupancy grid map. FIG. 8 is a diagram showing a short-term unoccupied grid map. FIG. 9 is a diagram showing the final grid map. FIG. 10 is a flowchart of final grid map generation control according to the second embodiment. FIG. 11 is a schematic configuration diagram of a driving support system according to a third embodiment. FIG. 12 is a flowchart of final grid map generation control. FIG. 13 is a schematic configuration diagram of a driving support system according to a fourth embodiment. FIG. 14 is a flowchart of final grid map generation control.

DESCRIPTION OF THE PREFERRED EMBODIMENTS An occupancy map creation method and an occupancy map creation device according to embodiments of the present invention will be described below with reference to the drawings.

(First embodiment)
FIG. 1 is a schematic configuration diagram of a driving support system that executes the occupancy map creation method according to the first embodiment.

The driving support system 100 is a system installed in a vehicle for the purpose of assisting a driver in driving the vehicle. The driving support system 100 includes an odometry measurement section 11, a distance measurement sensor 12, a distance measurement data storage section 13, short-term accumulated distance measurement data 14, long-term accumulated distance measurement data 15, an occupancy grid map generation section 16, A short-term occupancy grid map 17, a long-term occupancy grid map 18, an unoccupied grid map generation section 19, a short-term unoccupied grid map 20, a long-term unoccupied grid map 21, an evaluation section 22, and an output section 23. have

Among these components, a distance measurement data storage section 13, short-term accumulated distance measurement data 14, long-term accumulated distance measurement data 15, occupancy grid map generation section 16, short-term occupancy grid map 17, long-term occupancy grid map 18, and unoccupied grid map The generation unit 19, the short-term unoccupied grid map 20, the long-term unoccupied grid map 21, the evaluation unit 22, and the output unit 23 are integrally configured as the grid map creation device 10. In the grid map creation device 10, the ranging data storage section 13 to the output section 23 may be configured in the same microcomputer or may be configured in different microcomputers. Further, the odometry measurement unit 11 and the distance measurement sensor 12 are provided as sensors separate from the grid map creation device 10.

Below, details of each configuration of the driving support system 100 will be explained.

When the odometry measurement unit 11 acquires odometry information regarding a running vehicle, the odometry measurement unit 11 outputs the odometry information to the ranging data storage unit 13 . Odometry information is vehicle travel information and is used to create a travel trajectory and the like. The odometry information includes, for example, the turning angle and speed of the vehicle.

The distance measurement sensor 12 is a sensor that measures the distance and direction from the vehicle of objects around the vehicle, that is, the object position with respect to the vehicle, and outputs the acquired distance measurement data to the distance measurement data storage section 13. The distance measuring sensor 12 is not limited to a laser radar (LiDAR), but may be a stereo camera or the like. The type of distance measuring sensor 12 is not limited as long as it can measure the relative distance and direction to the vehicle to a road surface on which obstacles such as structures and moving objects are provided. Therefore, the position as ranging data is indicated in a vehicle coordinate system centered on the vehicle.

The distance measurement data storage section 13 receives input of vehicle odometry information acquired by the odometry measurement section 11 and distance measurement data acquired by the distance measurement sensor 12. Then, the distance measurement data storage unit 13 converts the distance measurement data observed in the vehicle coordinate system (that is, the object position in the vehicle coordinate system) into the object position in the odometry coordinate system in accordance with these inputs.

The distance measurement data storage unit 13 creates short-term accumulated distance measurement data 14 and long-term accumulated distance measurement data 15 by accumulating the distance measurement data in the odometry coordinate system for a predetermined period. The long-term accumulated distance measurement data 15 records distance measurement data for a period from the present to a predetermined first time period. In the short-term accumulated distance measurement data 14, distance measurement data for a period from the present time to a predetermined first time period is recorded. Note that the second time is shorter than the first time. Note that the short-term accumulated distance measurement data 14 and the long-term accumulated distance measurement data 15 are represented by an absolute coordinate system centered on the position of the vehicle at the time when the accumulation of the distance measurement data starts.

The occupancy grid map generation unit 16 creates a short-term occupancy grid map 17 in a predetermined two-dimensional area by superimposing the short-term accumulated distance measurement data 14. At the same time, the occupancy grid map generation unit 16 creates a long-term occupancy grid map 18 by superimposing the accumulated distance measurement data on the long-term accumulated distance measurement data 15. Then, the occupancy grid map generation section 16 transmits the created short-term occupancy grid map 17 and long-term occupancy grid map 18 to the unoccupied grid map generation section 19.

Note that the occupancy grid maps (short-term occupancy grid map 17 and long-term occupancy grid map 18) generated by the occupancy grid map generation unit 16 are created by dividing a two-dimensional predetermined area around the vehicle into a lattice (grid) shape. This is a discrete representation of the presence of objects in the grid. That is, the occupancy grid map is a map in which the presence of obstacles in a predetermined area is expressed as a set of grids on a road surface where obstacles such as structures and moving objects are present. In this occupied grid map, each grid is shown as an occupied grid occupied by an obstacle or an unoccupied grid not occupied by an obstacle.

As occupancy grid maps, there are known occupancy grid maps that show discrete values in an XY orthogonal coordinate system and discrete values in an r-θ polar coordinate system. In this embodiment, the occupancy grid map generated by the occupancy grid map generation unit 16 is expressed in an XY orthogonal coordinate system. Note that details of this occupancy grid map will be explained later using FIG. 4.

The unoccupied grid map generation unit 19 performs ray tracing processing, which will be described later, on each of the short-term occupied grid map 17 and the long-term occupied grid map 18 that are input, thereby identifying grids that can be determined to be unoccupied grids. Then, the unoccupied grid map generation section 19 generates a short-term unoccupied grid map 20 and a long-term unoccupied grid map 21 indicating unoccupied grids, and outputs them to the evaluation section 22 . Note that details of the ray tracing process will be explained in step S17 of FIG. 5, which will be described later.

The long-term occupancy grid map 18 generated by the occupancy grid map generation section 16 and the short-term unoccupied grid map 20 generated by the unoccupied grid map generation section 19 are input to the evaluation section 22 . The evaluation unit 22 evaluates, among the occupied grids in the long-term occupied grid map 18, the grids shown as unoccupied grids in the short-term unoccupied grid map 20 as unoccupied grids.

Finally, in the long-term occupancy grid map 18, the evaluation unit 22 outputs, as a final grid map, a grid that has been evaluated as an unoccupied grid among the occupied grids in the long-term occupancy grid map 18 and has been changed to an unoccupied grid. output to section 23. In this way, in the long-term occupancy grid map 18, the occupancy grid caused by the past presence of a moving object can be deleted, so that the accuracy of determining the occupancy grid in the final grid map can be improved.

The output unit 23 outputs the grid map generated by the evaluation unit 22 to an automatic driving system or the like outside the driving support system 100. For example, an automated driving system uses the received grid map to assist the vehicle in avoiding obstacles.

Note that the long-term unoccupied grid map 21 can be generated by the same process as the short-term unoccupied grid map 20, but is not used for the processing in the evaluation unit 22. Therefore, generation of the long-term unoccupied grid map 21 may be omitted.

Here, an outline of the acquisition control of distance measurement data by the distance measurement sensor 12 will be explained.

FIG. 2 is an explanatory diagram showing acquisition control of distance measurement data by the distance measurement sensor 12. This figure shows a situation in which the vehicle 200 travels in a three-dimensional space having x, y, and z axes. Note that the traveling direction corresponds to the x-axis, the vehicle width direction corresponds to the y-axis, and the height direction corresponds to the z-axis.

The distance measurement sensor 12 sets a plurality of measurement points in a predetermined rectangular area on the xy plane at a predetermined period, and measures distances to these measurement points. In this figure, the distance is measured at four timings: time t-1, t, t+1, and t+2. The distance measurement sensor 12 then generates distance measurement data using this measurement result. The distance measurement data indicates the presence or absence of an object at an arbitrary point in a three-dimensional predetermined area in a relative coordinate system with vehicle 200 as a reference. Therefore, by using the distance measurement data, it is possible to know the height of the road surface including obstacles such as structures and moving objects.

According to this figure, the distance measurement data at time t-1 includes height information of a cylindrical obstacle on the rear left when the traveling direction is set as the front. The distance measurement data at time t includes height information of a rectangular parallelepiped obstacle on the left front. Further, the distance measurement data at time t+1 includes height information of a cylindrical obstacle in the right front. In this way, the distance measurement data indicates the position in the xy plane and the height in the z direction of the structure on the road surface.

Note that the coordinate system of the distance measurement data is converted by the distance measurement data storage section 13. Specifically, the distance measurement data storage unit 13 uses odometry information input from the odometry measurement unit 11 to determine the position of the vehicle 200 at each measurement timing (t-1 to t+2) using odometry coordinates. . Then, the distance measurement data storage unit 13 converts the distance measurement data acquired by the distance measurement sensor 12 into absolute coordinates based on a predetermined location indicated by odometry coordinates from a relative coordinate system based on the vehicle 200. Perform the conversion to the system.

Finally, the ranging data storage unit 13 uses the plurality of ranging data converted into the absolute coordinate system to generate short-term accumulated ranging data 14 and long-term accumulated ranging data 15 in which ranging data is accumulated. do. Then, the occupancy grid map generation unit 16 generates a short-term occupancy grid map 17 and a long-term occupancy grid map 18 using the short-term accumulated distance measurement data 14 and the long-term accumulated distance measurement data 15, respectively.

Below, a method for generating occupancy grid maps (short-term occupancy grid map 17 and long-term occupancy grid map 18) from ranging data will be explained using FIG. 3, and using FIG. An example of the final grid map will be explained.

FIG. 3 is an explanatory diagram of a method for generating an occupancy grid map from ranging data.

As described above, the distance measurement sensor 12 sets a plurality of measurement points x _j in a predetermined area around the distance measurement sensor 12, obtains distances to these measurement points x _j , and obtains distance measurement data. seek. The distance measurement data has a normal distribution of covariance Σ with the measurement point x _j as the average (center of distribution) at each arbitrary point in a predetermined three-dimensional area, corresponding to one measurement point x _j . shown. Note that the covariance Σ corresponds to the probability of existence of an object. That is, for one measurement point x _j , a normal distribution (Gaussian distribution) of the covariance Σ in the three-dimensional region is obtained, which has a peak at the measurement point x _j . This means that each of the measurement points x _j defines a normal distribution with a peak at itself, and influences the distribution value of the covariance Σ at any point in the three-dimensional space.

Therefore, using the normal distribution of the covariance Σ corresponding to all measurement points x _j , the covariance in the z direction (height direction) _is calculated for each grid x Calculate the sum of variance Σ. Thereby, it is possible to obtain an occupancy value corresponding to the height in the z direction of the structure existing in the grid x _i in the xy plane. In addition, in calculating the occupancy value, the distance r _j from the ranging sensor 12 to the measurement point x _j and the height h _j from the xy plane at the measurement point x _j are weighted to the distribution value of the covariance Σ. Riding.

The farther the distance r _j from the distance measurement sensor 12 to the measurement point x _j becomes, the sparser the distribution of the measurement points x _j becomes. Therefore, by weighting the distribution value of the covariance Σ by multiplying it by the distance r _j to the measurement point x _j , it is possible to increase the influence of the normal distribution of the measurement point x _j on the obtained occupancy value. Thereby, it is possible to compensate for the sparse distribution of measurement points x _j .

A laser reflected from the road surface is used to measure the measurement point x _j . However, when measuring a measurement point x _j at a certain height, the distribution value may be smaller than the correct distribution value depending on the reflection characteristics of the road surface. Therefore, by weighting the distribution value of the covariance Σ by multiplying it by the height h _j of the measurement point x _j , it is possible to compensate for the inaccuracy of the normal distribution caused by the height of the measurement point x _j .

Therefore, the occupancy value C _i in any grid x _i can be expressed as follows. _{However ,} N (.; Moreover, x _i indicates the position of the grid in relative coordinates with respect to the vehicle on the xy plane. i is an index indicating the relative coordinate position.

In addition, in the occupancy grid map, the occupancy value C _i in each grid is subjected to a binarization process to determine whether it is larger than a certain threshold value, and it is determined whether the obstacle in each grid is occupied or not. Status is indicated. Note that the occupancy grid map may be expressed as continuous values of occupancy values C _i without performing the binarization process.

The representation format of the distance measurement data is not limited to the point cloud format in which measurement points that measure the distance from the distance measurement sensor 12 as described above are distributed in a three-dimensional space. The distance distribution from the distance measurement sensor 12 to the measurement point may be expressed as a distance image projected onto a two-dimensional plane, such as an output result from a stereo camera.

FIG. 4 is a diagram showing an example of the final grid map that is finally generated. For example, a two-dimensional rectangular area with a predetermined length in the front and rear in the driving direction and left and right in the vehicle width direction is formed with the vehicle 200 at the center, and this rectangular area is divided into a grid to form a grid. Note that the occupancy grid maps (short-term occupancy grid map 17 and long-term occupancy grid map 18) and unoccupied grid maps (short-term unoccupied grid map 20 and long-term unoccupied grid map 21, respectively) are also created in similar shapes. Ru.

According to this figure, the rectangular area is a range from 80 m forward to 10 m rearward (+80 m to -10 m) and 20 m left and right (-20 m to +20 m) with the vehicle 200 at the center. Further, in the rectangular area, a grid of 10 to 20 cm square is formed. Since such a two-dimensional occupancy grid map has a rectangular shape with a long front end in the traveling direction, the occupancy state of obstacles in front of the vehicle can be more clearly indicated.

The outline of the method for creating such a final grid map is as follows.

First, in the final grid map, a short-term occupancy grid map 17 and a long-term occupancy grid map 18 are generated, and then a short-term unoccupied grid map 20 is generated using the short-term occupancy grid map 17.

Here, the long-term occupancy grid map 18 uses a plurality of distance measurement data accumulated over a long period of time. Therefore, even if distance measurement data cannot be acquired temporarily, other distance measurement data can be used in a complementary manner, so that it is possible to improve the accuracy of determining the occupancy state of obstacles in each grid.

However, if distance measurement data that has been accumulated over a long period of time is used, even though the area occupied by the moving object 63 in the past, which is indicated by relatively old distance measurement data, does not currently interfere with the travel of the vehicle 200. Therefore, there is a risk that the grid will be determined to be an occupied grid. Therefore, among the occupied grids in the long-term occupied grid map 18, the grids indicated as unoccupied grids in the short-term unoccupied grid map 20, which is generated using relatively new distance measurement data, are considered to be unoccupied by obstacles. Decide and change to an unoccupied grid.

In this way, even if the moving object is shown as an occupied grid in the long-term occupied grid map 18, if it is shown as an unoccupied grid in the short-term unoccupied grid map 20, the moving object Therefore, it can be determined that the grid is an unoccupied grid because it does not interfere with the current running of the vehicle 200. In this way, it is possible to improve the accuracy of determining the occupied area. The details of the final grid map generation control will be described below using FIG. 5.

FIG. 5 is a flowchart of final grid map generation control by the driving support system 100. The grid map creation device 10 of the driving support system 100 is programmed to repeatedly perform a series of grid map generation control processes shown in this figure at predetermined timing. In the following, for clarity of explanation, the blocks that implement predetermined functions in the grid map creation device 10 in FIG. 1 will be described as performing respective processes. Note that the grid map creation device 10 may perform a series of processes.

Evaluation grid map generation control consists of two main parts: distance measurement data superimposition control, which is collectively shown as step S10 and consists of steps S11 to S14, and evaluation control, which is collectively shown as step S15, and consists of steps S16 to S18. including processing.

First, the distance measurement data superimposition control in step S10 will be explained.

In step S11, the distance measurement data storage section 13 receives the distance measurement data acquired by the distance measurement sensor 12. As described above, the distance measurement data indicates the occupancy value C _i of an obstacle in the grid x _i in a predetermined area around the vehicle 200 in the vehicle coordinate system.

In step S12, the distance measurement data storage section 13 acquires distance measurement data from the distance measurement sensor 12 and odometry information from the odometry measurement section 11. Then, the distance measurement data storage unit 13 converts the distance measurement data from the vehicle coordinate system to the absolute coordinate system using the odometry information.

As described above, since the distance measurement data is shown with the position of the vehicle 200 as the zero point, it is a relative coordinate system in which the zero point changes depending on the acquisition timing of the distance measurement data. The odometry coordinates are, for example, an absolute coordinate system whose zero point is the position of the vehicle 200 at a specific timing such as when the vehicle starts traveling. Therefore, by using odometry information, it is possible to convert the acquired ranging data from a relative coordinate system to an absolute coordinate system.

In step S13, the distance measurement data storage unit 13 further records the distance measurement data converted into odometry coordinates into a time buffer.

Here, the time buffer is a memory area that records a plurality of elements of a predetermined length, and is generally referred to as an array. In this example, the time buffer is configured to be able to record a predetermined number of ranging data. Further, in this time buffer, distance measurement data acquired at a certain timing is recorded as one element constituting the array of the time buffer, and distance measurement data for a predetermined period is recorded in the entire time buffer.

More specifically, when the distance measurement data storage unit 13 receives new distance measurement data from the distance measurement sensor 12, it adds the new distance measurement data as the final element of the time buffer, and also adds the new distance measurement data as the final element of the time buffer. Delete the oldest ranging data that is an element. By doing so, the time buffer stores the distance measurement data of a predetermined length of time from the present to the past since the number of stored distance measurement data is held constant. A memory structure in which new elements are added and the oldest element is deleted is called a ring buffer.

In this embodiment, two independent time buffers are provided: a short-term time buffer and a long-term time buffer. For example, it is assumed that the short-term time buffer can store 5 pieces (5 steps) of distance measurement data, and the long-term time buffer can store 50 pieces (50 steps) of distance measurement data.

The amount of distance measurement data accumulated by the distance measurement data storage section 13, that is, the number of elements in the time buffer array may be changed depending on the vehicle speed. As an example, the number of elements in the time buffer array may be changed to a greater extent as the vehicle speed becomes lower. By doing so, it is possible to prevent the distance measurement data from being deleted from the time buffer before the vehicle passes through the driving environment for which the grid map is to be generated.

As another example, the number of elements in the time buffer may be decreased as the vehicle speed increases. By doing this, if the vehicle has already passed the driving environment, unnecessary distance measurement data indicating the passing driving environment can be easily deleted, which reduces the calculation load in occupancy grid map generation control and It is possible to reduce the amount of memory used to store distance data.

Furthermore, while the vehicle is stopped, the accumulation of distance measurement data by the distance measurement data storage section 13 may be interrupted. If a moving object 63 enters the distance measurement range while the vehicle is stopped, a new area becomes a blind spot for the distance measurement sensor 12. By interrupting the accumulation of distance measurement data in such cases, it is possible to prevent the distance measurement data indicating the occupancy status of obstacles in the blind spot area from being erased due to overwriting of the distance measurement data while the vehicle is stopped. . Note that instead of suspending the accumulation of distance measurement data by the distance measurement data storage section 13, the acquisition of distance measurement data by the distance measurement sensor 12 may be stopped.

In step S14, the distance measurement data storage unit 13 superimposes the distance measurement data in the absolute coordinate system recorded in each element of the array for each of the short-term time buffer and the long-term time buffer. The ranging data storage unit 13 uses a short-term time buffer to generate short-term accumulated ranging data 14 based on ranging data from a relatively short time in the past to the present, and also uses a long-term time buffer to generate short-term accumulated ranging data 14 based on ranging data from a relatively short time in the past to the present. , generates long-term accumulated distance measurement data 15 based on distance measurement data from a relatively long time in the past to the present.

Here, since the ranging data stored in the time buffers (short-term time buffer and long-term time buffer) have been converted into odometry coordinates in step S13, their zero points are equal. Therefore, by superimposing the distance measurement data recorded in the time buffer along with the coordinates, the accumulated distance measurement data (short-term accumulated distance measurement data 14) and long-term accumulated distance measurement data 15). In this way, since a plurality of distance measurement data are used, it is possible to generate high-density accumulated distance measurement data that cannot be obtained from distance measurement data obtained by one measurement.

Next, the evaluation control in step S15 will be explained.

In step S16, the occupancy grid map generation unit 16 projects the normal distribution of the covariance Σ in the three-dimensional space shown in the short-term accumulated distance measurement data 14 onto the road surface, and Calculate the occupancy value C _i at x _i . Then, the occupancy grid map generation unit 16 generates a two-dimensional short-term occupancy grid map 17 by performing a binarization process on the occupancy value C _i .

Similarly, the occupancy grid map generation unit 16 generates a long-term occupancy grid map 18 using the long-term accumulated ranging data 15. In the short-term occupied grid map 17 and the long-term occupied grid map 18, each grid is shown as an occupied grid or an unoccupied grid depending on whether or not an obstacle is occupied.

In step S17, the unoccupied grid map generation unit 19 performs ray tracing processing on each of the short-term unoccupied grid map 17 and long-term unoccupied grid map 18 detected in step S16, and An occupancy grid map 21 is generated.

In addition, in the ray tracing process, in the occupancy grid map (short-term occupancy grid map 17 and long-term occupancy grid map 18), in a 360° area in the circumferential direction centered on the distance measurement sensor 12, the distance measurement sensor 12 is used as the starting point. A ray (line of sight), which is a virtual optical path, is sequentially rotated and projected. Then, the unoccupied grid map generation unit 19 determines the presence or absence of an occupied grid in each of the ray projection paths.

When an occupied grid exists on the ray projection path, there is no obstacle between the ranging sensor 12 and the occupied grid, so the grid between the ranging sensor 12 and the occupied grid is It is determined as an unoccupied grid with no occupied obstacles. Note that a grid located on the side opposite to the distance measurement sensor 12 from the occupied grid is determined to be an unknown grid because the distance measurement sensor 12 becomes blind due to an obstacle.

In this way, by performing ray tracing processing on the occupancy grid maps (short-term occupancy grid map 17 and long-term occupancy grid map 18), the areas of unoccupied grids are determined. Then, the unoccupied grid map generation unit 19 generates unoccupied grid maps (short-term unoccupied grid map 20 and long-term unoccupied grid map 21) including the unoccupied grids and unknown grids determined in this way.

In step S18, the evaluation unit 22 evaluates the occupancy state in the long-term occupancy grid map 18 for each grid using the unoccupancy state in the short-term unoccupancy grid map 20. Specifically, the evaluation unit 22 determines that among the occupied grids in the long-term occupied grid map 18, the grids that are unoccupied in the short-term unoccupied grid map 20 are unoccupied grids. Then, the evaluation unit 22 outputs the long-term occupied grid map 18 in which the grid is changed to an unoccupied grid to the output unit 23 as a final grid map.

As a result, among the occupied grids of the long-term occupied grid map 18, the occupied grid where the moving object 63 existed in the past is changed to an unoccupied grid. Therefore, it is possible to suppress the influence of the past occupied area of the moving object 63 shown in relatively old ranging data in the long-term occupancy grid map 18, so that the occupancy grid can be determined more accurately and the final grid map can be generated. I can do it.

A final grid map is generated through such final grid map generation control. Below, details of the final grid map generation control will be explained using a specific example with reference to FIGS. 6 to 9.

FIG. 6 is a diagram illustrating an example of a region for which a final grid map is generated.

The vehicle 200 is equipped with a distance sensor 12 and travels upward in the figure in a travelable area 62 surrounded by obstacles 61. Further, it is assumed that a moving object 63 is moving from the center of the figure to the upper right in front of the vehicle 200 in the travelable area 62. In this figure, distance measurement data is acquired eight times while the moving object 63 is moving, and in this figure, eight moving objects 63 are shown that are acquired by each distance measurement data. In addition, in the following, these eight consecutive distance measurement data will be referred to as 1st to 8th distance measurement data in chronological order, and each of the moving objects 63 shown in these 1st to 8th distance measurement data will be referred to as 1st to 8th distance measurement data. , are denoted as 1st to 8th moving objects 63.

FIG. 7 is a diagram showing a short-term occupancy grid map 17 and a long-term occupancy grid map 18 generated by the occupancy grid map generation unit 16. In this figure, a short-term occupancy grid map 17 is shown in FIG. 7(a), and a long-term occupancy grid map 18 is shown in FIG. 7(b).

In these figures, grids occupied by obstacles are shown as occupied grids with upward hatching to the right, and other grids are shown as unoccupied grids with no hatching. Further, in these figures, a vehicle 200 is shown for convenience of explanation.

To generate the short-term occupancy grid map 17, comparatively short-term distance measurement data is used, and in detail, the seventh and eighth distance measurement data among the eight distance measurement data are used. Note that the moving objects 63 shown in the first to sixth distance measurement data are shown with dotted lines for explanation. In this way, the short-term occupancy grid map 17 generated using the 7th and 8th ranging data includes, in addition to the wall surface 61A of the obstacle 61, the wall surface of the 7th and 8th two moving objects 63 in the movement trajectory. 63A is shown as the occupancy grid. On the other hand, areas other than the wall surfaces 61A and 63A are shown as unoccupied grids.

On the other hand, to generate the long-term occupancy grid map 18, relatively long-term distance measurement data, that is, all eight distance measurement data are used. In the long-term occupancy grid map 18, similarly to the short-term occupancy grid map 17, the areas of the wall surfaces 61A and 63A are occupied grids, and the areas other than the wall surfaces 61A and 63A are unoccupied grids.

FIG. 8 is a diagram showing a short-term unoccupied grid map 20 and a long-term unoccupied grid map 21 obtained by ray tracing processing. 8(a) shows a short-term unoccupied grid map 20, and FIG. 8(b) shows a long-term unoccupied grid map 21.

In these figures, occupied grids occupied by obstacles are hatched upward to the right, and unoccupied grids not occupied by obstacles are hatched with dots. Note that areas without hatching are unknown grids in which the occupation state of obstacles is unknown.

As shown in FIG. 8A, the unoccupied grid map generation unit 19 virtually generates a 360° two-dimensional area centered on the distance measurement sensor 12 in the short-term occupancy grid map 17. 12, a ray 81 is projected while rotating.

If there are occupied grids such as the wall surfaces 61A and 63A at the projection destination of the ray 81, it is determined that the area between the ranging sensor 12 and these occupied grids is an unoccupied grid 82. On the other hand, in the projection destination of the ray 81, an area on the opposite side of the distance measurement sensor 12 with respect to the wall surfaces 61A and 63A is determined to be an unknown grid 83.

By performing the ray tracing process in this manner, a short-term unoccupied grid map 20 is generated in which an unoccupied grid 82 in which no obstacle exists, an unknown grid 83 in which the existence of an obstacle is unknown, and the like are shown. Note that the short-term unoccupied grid map 20 also records the occupied grids shown in the short-term occupied grid map 17.

Similarly, as shown in FIG. 8B, a long-term unoccupied grid map 21 is generated by performing ray tracing processing on the long-term occupied grid map 18. In the long-term occupancy grid map 18, the unknown grids 83, which are blind spots on the wall surface 63A of the moving object 63, are distributed over a wider range than the unknown grids 83 in the short-term occupancy grid map 17 shown in FIG. 8(a). Become.

FIG. 9 is a diagram showing the final grid map obtained through the evaluation process of step S18. In the final grid map, occupied grids are hatched upward to the right, and unoccupied grids are hatched with dots. Note that areas without hatching are unknown grids.

Of the occupancy grids in which obstacles exist, which are hatched upward to the right in the long-term occupancy grid map 18 in FIG. Areas of unoccupied grids are evaluated as unoccupied grids. Specifically, the portion where the six wall surfaces 63A of the first to sixth moving objects 63 are present is an occupied grid in the long-term occupied grid map 18, but is an unoccupied grid in the short-term unoccupied grid map 20. Therefore, it is recorded as an unoccupied grid.

By doing so, occupied grids caused by the past occupancy state of the moving object 63 in the long-term occupancy grid map 18, such as the six wall surfaces 63A of the first to sixth moving objects 63, are determined as unoccupied grids. be done. Therefore, it is currently possible to accurately determine unoccupied grids that do not interfere with the travel of vehicle 200, and therefore it is possible to generate a grid map that more appropriately indicates the occupancy state of obstacles.

According to the first embodiment, the following effects can be obtained.

According to the occupancy grid map creation method of the first embodiment, ranging data is acquired on the road surface around the vehicle 200 at predetermined intervals. The distance measurement data shows the positions of a plurality of measurement points set on the road surface and the distances to these measurement points. Thereafter, a long-term occupancy grid map 18 is generated using a plurality of ranging data from a first time in the past to the present, and a long-term occupancy grid map 18 is generated from a second time shorter than the first time in the past to the present. A short-term unoccupied grid map 20 is generated using a plurality of distance-measured occupancy data during the period. Then, among the occupied areas in the long-term occupied grid map 18, the areas that become unoccupied grids in the short-term unoccupied grid map 20 are changed to unoccupied grid areas.

In such a configuration, the long-term occupancy grid map 18 uses a plurality of distance measurement data accumulated over a long period of time. Therefore, even if distance measurement data cannot be acquired temporarily, other distance measurement data can be used in a complementary manner, making it possible to improve the accuracy of determining the occupancy state of obstacles in each grid.

However, if distance measurement data that has been accumulated over a long period of time is used, even though the area occupied by the moving object 63 in the past, which is indicated by relatively old distance measurement data, does not currently interfere with the travel of the vehicle 200. Therefore, there is a risk that the grid will be determined to be an occupied grid. Therefore, among the occupied grids in the long-term occupied grid map 18, the grids indicated as unoccupied grids in the short-term unoccupied grid map 20, which is generated using relatively new distance measurement data, are considered to be unoccupied by obstacles. Determine and record as an unoccupied grid.

In this way, by evaluating the long-term occupancy grid map 18 using the short-term unoccupied grid map 20, the present running of the vehicle 200 can be made as if the moving object 63 were on the grid occupied in the past. A grid that does not interfere can be determined as an unoccupied grid. In this way, it is possible to improve the accuracy of determining the occupancy state of the obstacle shown in the occupancy grid map.

According to the occupancy grid map creation method of the first embodiment, as the vehicle speed becomes lower, the amount of data accumulated in the accumulated ranging data (short-term accumulated ranging data 14 and long-term accumulated ranging data 15) increases. The accumulation time (first time and second time) of ranging data used to generate a grid map is lengthened. When the vehicle speed is low, the range of the driving environment indicated by the distance measurement data accumulated in the accumulated distance measurement data per unit time becomes narrow. Therefore, by increasing the storage time, it is possible to prevent the distance measurement data from being deleted from the time buffer before the vehicle 200 passes through the driving environment for which the grid map is to be generated.

On the other hand, as the vehicle speed increases, the time for accumulating the accumulated ranging data is shortened. When the vehicle speed is high, the range of the driving environment indicated by the distance measurement data accumulated in the accumulated distance measurement data per unit time becomes wide. Therefore, by shortening the accumulation time, if the vehicle 200 has already passed the driving environment, unnecessary ranging data indicating the passed driving environment can be easily deleted. Therefore, it is possible to reduce the calculation load in occupancy grid map generation control and the amount of memory used to store distance measurement data.

According to the occupancy grid map creation method of the first embodiment, the accumulation of distance measurement data by the distance measurement data storage section 13 may be interrupted while the vehicle is stopped. If a moving object 63 enters the distance measurement range while the vehicle is stopped, a new area becomes a blind spot for the distance measurement sensor 12. Therefore, by interrupting the accumulation of distance measurement data while the vehicle is stopped, it is possible to prevent the distance measurement data indicating the occupancy status of obstacles in the blind spot area from being overwritten and erased by the distance measurement data while the vehicle is stopped. I can do it.

(Second embodiment)
In the first embodiment, after the ranging data showing a normal distribution of covariance in the three-dimensional space are superimposed in the three-dimensional space (S14), the normal distribution of the covariance Σ shown in the superimposed ranging data is A two-dimensional occupancy grid map was generated by projecting the distribution values in the height direction onto the road surface (S16). In this embodiment, as another method for creating an occupancy grid map, a two-dimensional map is created by projecting the normal distribution of the covariance Σ onto the road surface using the distance measurement data. An example will be described in which an occupancy grid map is generated by generating subgrid maps of , and then superimposing these subgrid maps on a two-dimensional plane.

FIG. 10 is a flowchart of evaluation grid map generation control according to the second embodiment.

According to this figure, compared to the grid map generation control of the first embodiment shown in FIG. 5, steps S21 and S22 are added in place of steps S14 and S16.

In step S21, the distance measurement data storage unit 13 projects the distribution value of the normal distribution of the covariance Σ in the height direction onto the road surface for one piece of distance measurement data recorded in the short-term buffer. Then, the distance measurement data storage unit 13 calculates an occupancy value C _i which is the sum total in the z-axis direction in each grid x _i , and performs a binarization process on the occupancy value C _i , thereby converting the short-term sub-grid Generate a map.

In this way, the distance measurement data storage unit 13 generates, for each of the distance measurement data recorded in the short-term time buffer, a short-term subgrid map indicating the occupation state of obstacles in the area indicated by the distance measurement data. Similarly, the ranging data storage unit 13 generates a long-term subgrid map using each of the ranging data recorded in the long-term time buffer.

In step S22, the occupancy grid map generation unit 16 superimposes the occupancy states shown in each of the short-term sub-grid maps generated by the ranging data storage unit 13 for each grid on a two-dimensional plane, thereby generating short-term occupancy. A grid map 17 is generated.

Here, in each of the short-term sub-grid maps, the occupancy state shown for the same grid is different, and one short-term sub-grid map may be an occupied grid, and another short-term sub-grid map may be an unoccupied grid. Therefore, the occupancy status of obstacles in a certain grid is determined by the ratio of the number of short-term sub-grid maps in which the grid is shown as an occupied part of the total number of short-term sub-grid maps used to generate the short-term occupancy grid map 17. exceeds a predetermined threshold (first ratio), it is determined to be an occupied portion. In this way, the occupancy state of obstacles for each grid is determined.

Similarly, the occupancy grid map generation unit 16 generates the long-term occupancy grid map 18 by superimposing the occupancy states shown in each of the long-term sub-grid maps for each grid. The occupancy status of an obstacle in a certain grid is determined by the ratio of the number of long-term sub-grid maps in which the grid is shown as an occupied part of the total number of long-term sub-grid maps used to generate the long-term occupancy grid map 18. (second ratio) exceeds a predetermined threshold value, it is determined that the area is occupied, and the occupancy state of the obstacle for each grid is determined.

In this way, the short-term occupancy grid map 17 and the long-term occupancy grid map 18 are generated using each of the plurality of short-term sub-grid maps and long-term sub-grid maps.

Here, when the short-term occupancy grid map 17 and the long-term occupancy grid map 18 are generated, the threshold of the ratio of sub-grid maps used to determine that a certain grid is an occupancy grid is the generation of the short-term occupancy grid map 17. The threshold value (first ratio) used in is smaller than the threshold value (second ratio) used in generating the long-term occupancy grid map 18.

Therefore, in the short-term occupancy grid map 17, each grid is more easily determined to be an occupancy grid than in the long-term occupancy grid map 18. Therefore, in the short-term unoccupied grid map 20 generated based on the short-term occupied grid map 17, the number of unoccupied grids is relatively small. This is because grids indicated as unoccupied grids in the sub-grid map are indicated as unoccupied grids at a higher rate, and the accuracy of determining unoccupied grids can be improved.

In this way, the evaluation unit 22 changes the grids that are determined to be unoccupied with high accuracy in the short-term unoccupied grid map 20 out of the occupied grids shown in the long-term occupied grid map 18 to unoccupied grids. Since the information is recorded, erroneous determination of the occupancy state can be suppressed.

According to the method for creating an occupancy grid map of the second embodiment, in the generation of the long-term occupancy grid map 18, the long-term sub-grid maps exceeding the first proportion of all the long-term sub-grid maps are occupied by obstacles. The area indicated as occupied is recorded as the occupied area.

On the other hand, in generating the short-term occupancy grid map 17, the grids that are shown to be occupied by obstacles in the short-term sub-grid maps exceeding the second proportion, which is smaller than the first proportion, among all the short-term sub-grid maps, are Record as occupancy grid. Therefore, in the short-term occupancy grid map 17, since the determination threshold is relatively low compared to the long-term occupancy grid map 18, the number of occupied grids is relatively large, while the number of unoccupied grids is relatively small.

This means that in the short-term occupancy grid map 17, grids that are indicated as unoccupied grids at a relatively high rate among all the short-term sub-grid maps are indicated as unoccupied grids. Therefore, in the short-term unoccupied grid map 20 generated based on the short-term occupied grid map 17, the accuracy of determining unoccupied grids can be improved.

As a result, the evaluation unit 22 changes the grid determined to be an unoccupied grid detected with high accuracy among the occupied grids shown in the long-term occupied grid map 18 to an unoccupied grid and records it. It is possible to further improve the accuracy of determining the occupancy state.

(Third embodiment)
In the first and second embodiments, the evaluation unit 22 erases the occupied grid due to the presence of the moving object 63, but the present invention is not limited thereto. In the third embodiment, an example will be described in which a sensor is further provided to detect position information of a moving object moving at high speed, and an occupancy grid indicated by the detected position information is erased.

FIG. 11 is a schematic configuration diagram of a driving support system 100 according to the third embodiment.

According to this figure, compared to the driving support system 100 of the first embodiment shown in FIG. Based on the position information, distance measurement data filtered by the distance measurement data filter section 31 is input.

The moving object detection unit 30 detects position information of a moving object that is present around the vehicle and moves at high speed, and outputs the position information of the detected moving object to the ranging data filter unit 31. For example, the moving object detection unit 30 is an image processing unit that includes a camera, and detects the position of a moving object that moves at high speed by analyzing captured time-series images.

The distance measurement data obtained by the distance measurement sensor 12 and the position information of the moving object obtained by the moving object detection section 30 are input to the distance measurement data filter section 31 . The distance measurement data filter section 31 erases the distance measurement data in the area where the moving object detected by the moving object detection section 30 is present from the distance measurement data acquired by the distance measurement sensor 12, and uses the erased distance measurement data. , and output to the distance measurement data storage section 13.

FIG. 12 is a flowchart of occupancy grid map generation control according to this embodiment.

According to this figure, compared to the occupancy grid map generation control of the first embodiment shown in FIG. 5, the process of step S30 is added after step S11.

In step S30, the distance measurement data filter section 31 correlates the distance measurement data acquired from the distance measurement sensor 12 with the position of the moving object detected by the moving object detection section 30. A moving object temporarily occupies the road surface, but may move to other areas while the vehicle is moving. Therefore, the distance measurement data filter section 31 deletes the distance measurement data of the position where the moving object detected by the moving object detection section 30 exists, based on the result of the association. Then, the distance measurement data from which the position of the moving object has been deleted is output to the distance measurement data storage section 13.

According to the third embodiment, the following effects can be obtained.

According to the method for creating an occupancy grid map of the third embodiment, the distance measurement data filter section 31 is used to calculate the movement detected by the moving object detection section 30 with respect to the distance measurement data obtained from the distance measurement sensor 12. Data indicating objects is removed in advance.

The moving object detection unit 30 detects a moving object by determining whether the object is a moving object based on changes in the position of the object acquired at predetermined intervals. Therefore, the moving object detection unit 30 has high accuracy in detecting a moving object that moves at a relatively high speed, but has low accuracy in detecting a moving object that moves at a relatively low speed, such as when the object starts moving.

Here, objects moving at high speed are less likely to stay around the vehicle and are less likely to interfere with the running of the vehicle. Therefore, by deleting high-speed moving objects detected by the moving object detection unit 30 from the distance measurement data in advance, it is possible to reduce the calculation load in the process of generating the occupancy grid map, and to reduce the storage space associated with the accumulation of the distance measurement data. This makes it possible to reduce the amount of water used. On the other hand, as in the first and second embodiments, objects moving at low speed can be evaluated using the short-term unoccupied grid map against the long-term occupied grid map, so that they do not obstruct the vehicle's progress. An occupied grid can be evaluated as an unoccupied grid.

In this way, by using the moving object detection section 30 in combination, it is possible to improve the accuracy of determining the occupancy grid while suppressing the resources necessary for processing moving objects at high speed.

(Fourth embodiment)
In the third embodiment, an example has been described in which the moving object detection unit 30 is used to erase an occupied grid caused by the presence of a moving object, but the present invention is not limited to this. In the fourth embodiment, an example will be described in which an occupied grid caused by the presence of a moving object is erased using another method.

FIG. 13 is a schematic configuration diagram of a driving support system 100 according to the fourth embodiment.

According to this figure, compared with the driving support system 100 of the second embodiment shown in FIG. A filter section 40 is provided.

The accumulated distance measurement data filter section 40 uses the detected position of the moving object acquired by the moving object detection section 30 to eliminate grids caused by the moving object in the short-term accumulated distance measurement data 14 and the long-term accumulated distance measurement data 15. I do.

FIG. 14 is a flowchart of occupancy grid map generation control according to this embodiment.

According to this figure, compared to the occupancy grid map control of the second embodiment shown in FIG. 12, the process of step S40 is further added after step S13.

In step S40, the accumulated distance measurement data filter section 40 compares the detected position of the moving object acquired by the moving object detection section 30 with the grid in the short-term accumulated distance measurement data 14 and the long-term accumulated distance measurement data 15. . Then, the accumulated distance measurement data filter section 40 erases the grid information at the detection position of the moving object acquired by the moving object detection section 30 in the short-term accumulated distance measurement data 14 and the long-term accumulated distance measurement data 15. In this way, filtering processing can be performed on the short-term accumulated distance measurement data 14 and the long-term accumulated distance measurement data 15.

Even in this case, since the distance measurement data corresponding to the moving object can be removed in advance from the distance measurement data acquired by the distance measurement sensor 12, the calculation load in the generation process of the occupancy grid map can be reduced, and the distance measurement data The amount of memory used for storing data can be reduced.

According to the fourth embodiment, the following effects can be obtained.

For example, at a certain time, the moving object detection section 30 is temporarily unable to detect a moving object, and the distance measurement data filter section 31 is unable to remove data indicating a moving object, so that distance measurement data corresponding to the moving object remains. Sometimes it happens. Even in such a case, when the moving object is successfully detected at another time, the distance measurement data corresponding to the moving object can be removed from the accumulated distance measurement data.

In this way, based on the detection result of a moving object by the moving object detection unit 30 at a certain time, it is possible to delete the distance measurement data corresponding to the moving object in the distance measurement data at another time, so that the occupancy grid map It is possible to further reduce the calculation load in the generation process and the amount of storage space used in accumulating distance measurement data.

Although the embodiments of the present invention have been described above, the above embodiments merely show a part of the application examples of the present invention, and are not intended to limit the technical scope of the present invention to the specific configurations of the above embodiments. do not have. Furthermore, the above embodiments can be combined as appropriate.

10 Grid map creation device 11 Odometry measurement section 12 Distance measurement sensor 13 Distance measurement data storage section 14 Short-term accumulated distance measurement data 15 Long-term accumulated distance measurement data 16 Occupancy grid map generation section 17 Short-term occupancy grid map 18 Long-term occupancy grid map 19 Unoccupied Grid map generation section 20 Short-term unoccupied grid map 21 Long-term unoccupied grid map 22 Evaluation section 23 Output section 30 Moving object detection section 31 Distance measurement data filter section 40 Accumulated distance measurement data filter section 61 Obstacle 63 Moving object 81 Ray 82 Non-occupied Occupied grid 83 Unknown grid 100 Driving support system 200 Vehicle

Claims (7)

A method for creating an occupancy map showing the occupancy status of obstacles on a road surface around a vehicle, the method comprising:
The occupancy map creation device is
Acquire the positions of a plurality of measurement points set around the vehicle at a predetermined period as distance measurement data,
Generating an occupancy map showing occupied and unoccupied parts of the obstacle on the road surface using the plurality of distance measurement data for a period from now until a first time ago;
Generating an unoccupied map showing an unoccupied portion of the obstacle on the road surface using the plurality of distance measurement data for a period from now to a second time shorter than the first time;
A method for creating an occupancy map, comprising changing an area included in the unoccupied portion shown in the unoccupied map out of the occupied portion shown in the occupancy map to the unoccupied portion. The method for creating an occupancy map according to claim 1,
In generating the occupancy map, the occupancy status of the obstacle at a predetermined point on the occupancy map is such that the predetermined point is the occupied portion of the total distance measurement data used for generating the occupancy map. If the ratio occupied by the number of distance measurement data exceeds a first ratio, it is determined that the part is occupied,
In the generation of the unoccupied map, the occupancy status of the obstacle at a predetermined point on the unoccupied map is determined by determining whether the predetermined point is occupied among the total distance measurement data used to generate the unoccupied map. When the ratio of the number of distance measurement data that is a part exceeds a second ratio smaller than the first ratio, it is determined to be the occupied part, and based on the determined occupied part, the A method for creating an occupancy map in which unoccupied areas are defined. The method for creating an occupancy map according to claim 1 or 2, further comprising:
detecting the position of a moving object moving on the road surface;
A method for creating an occupancy map, comprising erasing occupancy information of the obstacle indicated at the position of the detected moving object in the distance measurement data. The method for creating an occupancy map according to claim 3, further comprising:
The occupancy information indicated at the detected position of the moving object in the plurality of distance measurement data used to generate the occupancy map and the plurality of distance measurement data used to generate the non-occupancy map. How to create an occupancy map to erase. The method for creating an occupancy map according to any one of claims 1 to 4,
The method for creating an occupancy map, wherein the faster the speed of the vehicle is, the shorter the first time and the second time are. The method for creating an occupancy map according to any one of claims 1 to 5,
A method for creating an occupancy map, in which the occupancy map and the non-occupancy map are generated by excluding the ranging data acquired during a period when the vehicle is stopped. A device for creating an occupancy map showing the occupancy status of obstacles on a road surface around a vehicle,
A distance measurement data acquisition unit that measures distances to a plurality of measurement points set around the vehicle at a predetermined period, and acquires the position of the measurement points and the distance to the measurement points as distance measurement data. and,
an occupancy map generation unit that generates an occupancy map indicating occupied and unoccupied portions of the obstacle on the road surface using the plurality of distance measurement data from the past from a first time to the present;
unoccupied, generating an unoccupied map showing an unoccupied portion of the obstacle on the road surface using the plurality of distance measurement data from the past to the present, which is a second time shorter than the first time; A map generation unit,
An occupancy map creation device comprising: a changing unit that changes an area included in the unoccupied portion shown in the unoccupied map out of the occupied portion shown in the occupancy map to the unoccupied portion.

Images