Autonomous driving method for vehicle and autonomous driving vehicle
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- SUGINO MACHINE
- Filing Date
- 2024-06-06
- Publication Date
- 2026-05-22
Smart Images
Figure 00000000_0000_ABST
Abstract
Description
[Technical Field]
[0001] The present invention relates to a method for autonomously driving a vehicle, and an autonomous vehicle. [Background technology]
[0002] A technology is well known in which a vehicle has multiple sensors and travels autonomously based on information acquired from the multiple sensors. An autonomous vehicle is known in which two sensors, a LiDAR and a camera, are fixed to the vehicle (Japanese Patent Laid-Open Publication No. 2022-145014, hereinafter referred to as Patent Document 1). The LiDAR is a two-dimensional LiDAR or a three-dimensional LiDAR. The vehicle travels autonomously while estimating its own position by acquiring the conditions of the driving space using the LiDAR, and the reliability of the own position is determined by acquiring the conditions of the road surface directly below the vehicle using a camera. Summary of the Invention [Problem to be solved by the invention]
[0003] The autonomous vehicle of Patent Document 1 can acquire road surface conditions and travel using 3D LiDAR. However, 3D LiDAR is more expensive than 2D LiDAR. On the other hand, when using 2D LiDAR, the road surface conditions cannot be grasped because there is a blind spot below the detection height of the 2D LiDAR.
[0004] The present invention provides an autonomous vehicle that is inexpensive and reduces the load on computing processing. [Means for solving the problem]
[0005] A first aspect of the present invention is Map data and destination data are stored, a first sensor fixed to a vehicle detects a space above a road surface in front of the vehicle and acquires upward data; estimating self-position data corresponding to the self-position of the vehicle based on the map data and the overhead data; determining a situation in the upper space based on the upper data; a second sensor fixed to the vehicle detects a blind spot area below the detection area of the first sensor and acquires distance image data that is the basis of a distance image as blind spot area data; The distance image has a matrix form in which a plurality of pixels are arranged vertically and horizontally, the range image data corresponds to a plurality of light receiving elements constituting the imaging element of the second sensor and is composed of a plurality of light receiving element data; the light receiving element data is a data set in which position data indicating a position of the light receiving element within the image capturing element is associated with distance data indicating a distance between the object detected by the second sensor and the light receiving element, a first window for selecting a portion of the pixels in the distance image in the form of a matrix arranged vertically and horizontally by selecting a portion of the light receiving element data of the distance image data is set so as to be configurable with respect to the distance image data; searching the distance data of the light receiving element data corresponding to each row for each set number of rows in the distance image, and extracting one of a maximum value and a minimum value as first peculiar data; placing the first window in the range image data so as to include the light receiving element data associated with the first peculiar data; determining a state of the blind spot area based on a correlation between the distance data included in all of the first windows arranged in the distance image data; generating route data for heading from the current location to the destination and control command data for controlling the vehicle based on the map data, the destination data, and the current location data; The method for autonomously driving a vehicle includes reflecting the results of determining the situation in the overhead space and the situation in the blind spot area in the route data and the control command data.
[0006] A second aspect of the present invention is Vehicles and a first sensor fixed to the vehicle in a state spaced apart from a road surface, the first sensor detecting a space ahead of the vehicle and above the road surface to acquire upward data; a second sensor fixed to the vehicle, facing a road surface ahead of the vehicle, for detecting a blind spot area below the detection area of the first sensor and acquiring distance image data that is the basis of a distance image as blind spot area data; a control device for controlling the vehicle; a destination input device for inputting destination data corresponding to the destination, The control device a storage unit that stores self-position data corresponding to the self-position of the vehicle, map data, the destination data, the overhead data, and the range image data; a self-location estimation unit that estimates the self-location data based on the map data and the overhead data; a route generation unit that generates route data corresponding to a route from the self-location to the destination; a control command generating unit that generates control command data for controlling the vehicle; an upper space determination unit that determines the status of the upper space based on the upper data; a blind spot area determination unit that determines the state of the blind spot area, the range image data is composed of a plurality of light receiving element data and corresponds to a plurality of light receiving elements that constitute the imaging element of the second sensor, the light receiving element data is a combination of position data indicating a position of the light receiving element within an image sensor and distance data indicating a distance between the object detected by the second sensor and the light receiving element, the blind spot area determination unit is configured to arrange a first window in the distance image data for selecting a portion of the light receiving element data of the distance image data so as to select a portion of the pixels of the distance image in the form of a matrix arranged vertically and horizontally by selecting a portion of the light receiving element data, searches for the distance data of the light receiving element data corresponding to each row for each number of rows set in the distance image and extracts one of the maximum and minimum values as first peculiar data, arranges the first window in the distance image data so as to include the light receiving element data associated with the first peculiar data, and determines the state of the blind spot area based on the correlation of the distance data included in all of the first windows arranged in the distance image data, the route generation unit generates the route data based on the map data, the destination data, and the vehicle's own position data, and reflects the determination results of the overhead space determination unit and the blind spot area determination unit in the route data; The control command generation unit is an autonomous vehicle that generates the control command data based on the route data and reflects the judgment results of the upper space judgment unit and the blind spot area judgment unit in the control command data.
[0007] The first sensor 11 may acquire the presence or absence of an object present in a detection range, the relative position of the object, and the shape of the object. The first sensor may be a two-dimensional Lidar (light detection and ranging) sensor. The cleaning device may be, for example, a jetting device or a dozer device. The jetting device sprays high-pressure water. The dozer device pushes out foreign objects on the road surface. [Effects of the Invention]
[0008] According to the present invention, it is possible to provide an autonomous vehicle that is inexpensive and has a small load on calculation processing during determination. [Brief explanation of the drawings]
[0009] [Figure 1] Autonomous vehicle side view [Figure 2] Plan view of an autonomous vehicle [Figure 3] Block diagram showing the configuration of an autonomous vehicle [Figure 4] FIG. 10 is an explanatory diagram showing an example of a distance image; [Figure 5] Flowchart of an autonomous vehicle driving method according to an embodiment [Figure 6] A flowchart that explains in detail a part of the process of FIG. 1. [Figure 7] An explanatory diagram showing the relationship between a distance image and distance data. [Figure 8] An explanatory diagram showing the relationship between a distance image and a dimensionless quantity [Figure 9] A perspective view showing an autonomous vehicle traveling inside a livestock barn. DETAILED DESCRIPTION OF THE INVENTION
[0010] As shown in FIGS. 1 and 2, an autonomous vehicle 100 according to the embodiment includes a vehicle 10, a first sensor 11, a second sensor 12, a destination input device 13 (see FIG. 3), and a control device 20.
[0011] The vehicle 10 is a crawler type vehicle and includes a vehicle body 10a, a pair of propulsion devices 10b, and a cleaning device 10c. The propulsion units 10b are provided on the left and right sides of the vehicle body 10a to propel the vehicle body 10a. The propulsion units 10b have left and right crawlers 10d and a rotation mechanism 10e. The rotation mechanism 10e is connected to the vehicle body 10a and the left and right crawlers 10d, and rotates the left and right crawlers 10d separately. The rotation mechanism 10e has two travel motors M1 (see FIG. 3). The travel motors M1 are vehicle drive units 14 that rotate the two crawlers 10d separately. The vehicle 10 moves straight by rotating the left and right crawlers 10d at the same speed and in the same direction. The vehicle 10 can change its direction of travel by rotating the left and right crawlers 10d at different speeds. The propulsion device 10b controls the two travel motors M1 to determine the speed and direction of travel of the vehicle 10.
[0012] The cleaning device 10c cleans the road surface. The cleaning device 10c is a dozer device 10c. The dozer device 10c includes an attachment 10f and an attachment drive mechanism (not shown). The attachment 10f is, for example, a bucket or a blade. The attachment drive mechanism is disposed inside the vehicle body 10a. The attachment drive mechanism controls the attachment 10f. The attachment drive mechanism can, for example, lower the attachment 10f to bring the attachment 10f into contact with the road surface, or raise the attachment 10f to move the attachment 10f away from the road surface. The attachment drive mechanism includes a cleaning motor M2 (see FIG. 3). The cleaning motor M2 is a cleaning drive unit 15.
[0013] The first sensor 11 is fixed to the upper part of the vehicle body 10a. The first sensor 11 is a two-dimensional Lidar sensor. When viewed from above, the first sensor 11 detects an upper space U in front of the vehicle 10. The first sensor 11 detects an angular range that is equal on both the left and right sides of the front. When viewed from above, the first sensor 11 has a wider detection angle range than the second sensor 12. The first sensor 11 acquires upper data D2 (see Figure 3). When the vehicle 10 is placed on a horizontal surface, the detection plane is a horizontal plane. The upper data D2 is data on the surroundings of the vehicle 10 on the detection plane. The first sensor 11 outputs the upper data D2 to the control device 20. The area below the detection plane is a blind spot for the first sensor 11.
[0014] The second sensor 12 is fixed to the front of the vehicle body 10a. The second sensor 12 is a depth sensor, for example, an infrared sensor. When viewed from above, the second sensor 12 detects a road surface area (blind spot area) R ahead of the vehicle 10. The road surface area R is evenly arranged on the left and right sides of the vehicle 10. As shown in FIG. 3, the detection range of the second sensor 12 is wider than the width of the vehicle 10. The second sensor 12 has a shorter detection distance than the first sensor 11. The second sensor 12 acquires road surface data (distance image data) D3 of the road surface area R. The road surface area R is an area ahead of the vehicle 10 and below the detection height of the first sensor 11. The road surface area R is near the road surface ahead of the vehicle 10. The second sensor 12 outputs the road surface data D3 to the control device 20.
[0015] The control device 20 is disposed inside the vehicle body 10a. As shown in Fig. 3, the control device 20 has a self-position estimation unit 30, a route generation unit 40, an overhead space determination unit 50, a road surface determination unit (blind spot area determination unit) 60, a control command generation unit 70, and a memory unit 80. The control device 20 may be integrated with the destination input device 13. The control device 20 is a computer that controls the vehicle 10.
[0016] Destination input device 13 is, for example, a touch panel. An operator inputs destination data D4 into destination input device 13. Destination data D4 indicates the destination to which autonomous vehicle 100 is heading. Destination input device 13 can display the input destination. Destination input device 13 also serves as an output device.
[0017] The storage unit 80 stores a program (not shown) for causing the vehicle 10 to travel autonomously. The program is, for example, ROS (Robot Operating System). The ROS has a SLAM (Simultaneous Localization and Mapping) function, a route generation function, an upper space determination function, a road surface area determination function, and a control command generation function. The SLAM function is a function that simultaneously creates an environmental map and estimates the vehicle's own position. The self-position estimation unit 30, the path generation unit 40, the overhead space determination unit 50, the road surface determination unit 60, and the control command generation unit 70 are formed in the control device 20 as functions of the ROS. The storage unit 80 stores map data D1, overhead data D2, road surface data D3, destination data D4, and self-position data D5.
[0018] The map data D1 may be input to the memory unit 80. The control device 20 may correct the map data D1 while driving. That is, the autonomous vehicle 100 is driven on the road surface at the planned driving location. As a result, the first sensor 11 acquires the overhead data D2 in time series. The overhead data D2 is sequentially output from the first sensor 11, and the self-position estimation unit 30 creates the map data D1. The map data D1 includes the road surface and the surrounding environment of the road surface. The map data D1 is output from the self-position estimation unit 30 and stored in the memory unit 80.
[0019] Destination data D4 is output from destination input device 13 and stored in memory unit 80. Autonomous driving begins when an operator inputs destination data D4 into destination input device 13. As autonomous vehicle 100 drives, overhead data D2 and road surface data D3 are stored in chronological order in memory unit 80.
[0020] The upper data D2 includes an angle and a distance. The angle is based on the front of the vehicle 10. The distance indicates the length from the first sensor to the object. A distance of 0 indicates that no object is present within the detection range. The position of the object relative to the vehicle 10 is determined based on the distance and angle. If an object that did not appear in the map data D1 is discovered during autonomous driving, the object is treated as a discovered object.
[0021] The road surface data D3 includes a plurality of light receiving element data. The total number of light receiving elements constituting the imaging element corresponds to the total number of light receiving element data. The light receiving element data includes position data and distance data. The position data indicates the position of the light receiving element within the imaging element. The distance data is provided in association with the position data and indicates the distance between the object detected by the second sensor 12 and the light receiving element.
[0022] Self-position estimation unit 30 receives the latest overhead data D2 from memory unit 80, compares the overhead data D2 with map data D1, and estimates the position of autonomous vehicle 100, as well as the direction of travel of autonomous vehicle 100. The self-position is the correlation between the actual position on the map and the direction of travel of autonomous vehicle 100. Self-position data D5 corresponding to the self-position is output from self-position estimation unit 30 and stored in memory unit 80. Self-position data D5 is updated as needed during autonomous travel.
[0023] Based on the upper data D2, the upper space determination unit 50 determines whether a found object exists and whether the found object is an obstacle. If an obstacle exists, the upper space determination unit 50 outputs the presence of the obstacle to the control command generation unit 70. An obstacle is a found object that is large enough that autonomous vehicle 100 cannot reach its destination as long as it travels on the road surface. Even if the found object is found, if autonomous vehicle 100 can travel on the road surface while avoiding the found object and reach its destination, the found object is not determined to be an obstacle.
[0024] Figure 4 shows an example of a distance image. The distance image represents the distance data of pixels using shades of color. The colors are assigned by settings and are unrelated to the road surface data D3. The distance image is displayed on an output device. The distance image has multiple pixels arranged in m rows and n columns, where m and n are positive integers.
[0025] FIG. 5 shows the correspondence between the distance image and distance data. Pixels P are indicated by circles. Distance data corresponding to each pixel P is indicated for each pixel P. In the distance image, pixels P are arranged in 12 rows and 21 columns. The road surface determination unit 60 searches for distance data in the light-receiving element data corresponding to each row, every three rows, and extracts the maximum value as first peculiar data D10 and the minimum value as second peculiar data D20. Unlike the example in FIG. 5, the minimum value may be extracted as the first peculiar data and the maximum value as the second peculiar data. The first peculiar pixel P1 corresponding to the first peculiar data D10 is indicated by a black circle. The second peculiar pixel P2 corresponding to the second peculiar data D20 is indicated by a circle with an X embedded in it. The first peculiar pixel P1 and the second peculiar pixel P corresponding to the second peculiar data D20 are distinguished from pixels P corresponding to other distance data by color and shape.
[0026] In FIG. 5, the ideal distance acquired by the light receiving element in each row is indicated on the left side. The ideal distance is distance data corresponding to pixel P when the second sensor 12 detects a horizontal road surface while the autonomous vehicle 100 is stationary, and is a value that does not include measurement error. The ideal distance increases as the distance increases from the second sensor 12. The ideal distance increases from pixel P in the lower row of the range image to pixel P in the upper row. The first peculiar data D10 is the maximum value among the distance data in the same row. The second peculiar data D20 is the minimum value among the distance data in the same row. For example, the distance data in the second row from the bottom (row 11) is 1078..., 1050..., 1098..., and 1110... (unit: mm). Therefore, the maximum value 1110... (column 15) is the first peculiar data D10, and the minimum value 1050... (column 7) is the second peculiar data D20.
[0027] 5, the road surface determination unit 60 selects light receiving element data at regular intervals (periods) to arrange a plurality of first windows W1 and a plurality of second windows W2 in the distance image. As a result, four first windows W1 and four second windows W2 are arranged in one distance image. The first window W1 and the second window W2 select a set of pixels P with s rows and t columns from the road surface data D3, which includes the first peculiar data D10 and the second peculiar data D20, respectively. s and t are positive integers. s (s=3 in FIG. 5) is smaller than m and smaller than half the value of m. t (t=4 in FIG. 5) is smaller than n and smaller than half the value of n. The distance image is large enough to have at least two first windows W1 arranged vertically and horizontally. s and t are preferably 3 or greater. The road surface determination unit 60 sets the first window W1 and the second window W2 for every three rows of the light receiving element data. As a result, the first window W1 is placed in the road surface data D3 so as to include the light receiving element data associated with the first peculiar data D10, and the second window W2 is placed in the road surface data D3 so as to include the light receiving element data associated with the second peculiar data D20.
[0028] As shown in Fig. 5, the first window W1 is placed in the road surface data D3 so that the first anomalous pixel P1 is located in the second row and second column of a set of pixels P arranged in 3 rows and 4 columns. When the first anomalous pixel P1 is located in the first column, the first window W1 may be placed so that the first anomalous pixel P1 is located in the second row and first column. When the first anomalous pixel P1 is located in the last column (the 21st column), the first window W1 may be placed so that the first anomalous pixel P1 is located in the second row and fourth column. When the first anomalous pixel P1 is located in the 20th column, the first window W1 may be placed so that the first anomalous pixel P1 is located in the second row and third column.
[0029] The arrangement method of the second window W2 is substantially the same as the arrangement method of the first window W1. As shown in FIG. 5, the second window W2 is arranged so that the second anomalous pixel P2 is located in the second row and second column of a group of pixels P arranged in 3 rows and 4 columns. When the second anomalous pixel P2 is located in the first column, the second window W2 may be arranged so that the second anomalous pixel P2 is located in the second row and first column. When the second anomalous pixel P2 is located in the last column (the 21st column), the second window W2 may be arranged so that the second anomalous pixel P2 is located in the second row and fourth column. When the second anomalous pixel P2 is located in the 20th column, the second window W2 may be arranged so that the second anomalous pixel P2 is located in the second row and third column.
[0030] The road surface determination unit 60 determines the condition of the area of the first window W1 for each first window W1 based on the correlation of the distance data included in the first window W1. By collecting the determination results for all first windows W1 arranged in the road surface data D3, the road surface determination unit 60 determines the condition of the road surface area R from the perspective of the road surface. The perspective of the road surface refers to, for example, whether the road surface is horizontal across its entire width, sloped, has steps, or has pits in part of its width. A slope is an uphill slope or a downhill slope. A step is a protruding step or a depression-like step. A pit is a hole or groove for collecting cleaning materials.
[0031] The road surface determination unit 60 determines the situation in the area of each second window W2 for each second window W2 based on the correlation of the distance data included in the second window W2. By collecting the determination results for all second windows W2 arranged in the road surface data D3, the road surface determination unit 60 determines the situation of the road surface area R from the perspective of whether an obstacle or cleaning object exists on the road surface.
[0032] The distance data included in windows W1 and W2 may be used directly in the calculation process for making the determination. The distance data may include fluctuations due to shaking of autonomous vehicle 100 during autonomous driving.
[0033] The distance data is made dimensionless. A normalization constant is used for the normalization. For example, the normalization constant is a reference distance, which is the distance that would be used as a reference when the second sensor 12 acquires the road surface data D3 under ideal conditions. There may be one normalization constant for one piece of road surface data D3, or there may be multiple normalization constants. When there are multiple normalization constants, for example, the normalization constants are different values for all the first windows. The normalization constants may be the same value for the first window W1 and the second window W2 located in the same row, or they may be different values. Preferably, the number of digits of the dimensionless quantity is 2 to 4. The dimensionless quantity may include a decimal point. The road surface judgment unit 60 calculates a first dimensionless quantity and a second dimensionless quantity for the distance data included in the windows W1 and W2 using equations (1) and (2). L1=L0 / C (formula 1) L2=L0 / C (formula 2) where: L1: First dimensionless quantity L2: Second dimensionless quantity L0: Distance data C: Normalization constant
[0034] FIG. 6 shows the relationship between the distance image and the dimensionless quantity. In FIG. 6, the first dimensionless quantity is plotted for the first window W1, and the second dimensionless quantity is plotted for the second window W2. Within all of the first windows W1, the first dimensionless quantity increases by 0.1 from the bottom row to the top row. This indicates that the road surface is horizontal. Within all of the second windows W2, the second dimensionless quantity increases by 0.1 from the bottom row to the top row. This indicates that there are no obstacles or cleaning objects on the road surface.
[0035] Tables 1 and 2 show examples of road surface region R determination based on the first and second dimensionless quantities. The first row shows the determination results, and the third and subsequent rows show examples of distance data and dimensionless quantities. The normalization constant C is set to 1000 in the case where there is one normalization constant for one road surface data D3. [Table 1] For example, if the first dimensionless quantity included in the first window W1 increases by a set value proportionally with the distance of the row of pixels P from the second sensor 12, the road surface determination unit 60 determines that the road surface in the area of the first window W1 is a horizontal plane. In Table 1, the first dimensionless quantity increases by 0.1 in the area of the entire first window W1. In this case, the road surface determination unit 60 determines that the entire road surface is a horizontal plane.
[0036] If the first dimensionless quantity included in the first window W1 increases proportionally with the distance of the row of pixels P from the second sensor 12, and the increase is greater than that in the case of a horizontal surface, the road surface determination unit 60 determines that the road surface in the area of the first window W1 is a downward slope. In Table 1, the first dimensionless quantity increases by 0.2 in all areas of the first window W1. In this case, the road surface determination unit 60 determines that the entire road surface is a downward slope.
[0037] If the first dimensionless quantity included in the first window W1 has a value significantly larger than that of a horizontal plane relative to the value in the row immediately below, the road surface judgment unit 60 determines that a depression-like step exists in the area of the first window W1. Also, if the value in the row immediately below is significantly smaller than that of a horizontal plane relative to the value in the row immediately below, the road surface judgment unit 60 determines that a protrusion-like step exists in the area of the first window W1. In Table 1, the value increases proportionally by 0.1 for each row of pixel P away from the second sensor 12 up to the fifth row from the bottom, then increases significantly to 1.8 in the sixth row from the bottom, increases by 0.1 from the sixth row to the eighth row, then decreases significantly to 1.8 in the ninth row from the bottom, and then increases proportionally by 0.1. In this case, the road surface judgment unit 60 determines that a depression-like step exists in the area of the second-lowest first window W1 and that a protrusion-like step exists in the area of the third-lowest first window W1.
[0038] If the first dimensionless quantity included in the first window W1 takes a value that is extremely larger in multiple first windows W1 that are successively arranged vertically than in the case of a horizontal surface, the road surface judgment unit 60 judges that a pit exists on the road surface for the road surface region R. In Table 1, the first dimensionless quantity in the fifth to ninth rows from the bottom has values of 6.0 to 6.8 that are extremely larger than the value in the case of a horizontal surface. For example, if there are four or more consecutive rows of values that are 3.0 or more larger than the value in the case of a horizontal surface, the road surface judgment unit 60 judges that a pit exists on the road surface.
[0039] [Table 2]
[0040] If the second dimensionless quantity included in the second window W2 takes a value that is significantly smaller in multiple second windows W2 that are successively arranged vertically than in the case of a horizontal plane, the road surface judgment unit 60 judges that an obstacle exists in the road surface region R. In Table 2, in the fourth to tenth rows from the bottom, the second dimensionless quantity is 1.0 to 0.9, which is significantly smaller than the first dimensionless quantity of 1.3 to 1.9 in the case of a horizontal plane. For example, if there are four or more consecutive rows of values that are 0.3 or more smaller than in the case of a horizontal plane, the road surface judgment unit 60 judges that an obstacle exists in the road surface region R.
[0041] Among the second dimensionless quantities included in the second window W2, if only one second dimensionless quantity included in the second window W2 has, for example, two identical values, the road surface determination unit 60 determines that a cleaning object is present in the road surface region R. A cleaning object is, for example, a foreign object that has fallen onto the road surface. In Table 2, the second dimensionless quantities in the fifth and sixth rows from the bottom have the same value, 1.4. In this case, the road surface determination unit 60 determines that a cleaning object is present in the road surface region R.
[0042] In all other cases, the road surface determination unit 60 determines that no obstacles or cleaning objects exist on the road surface of the road surface area R. If an obstacle exists, the road surface determination unit 60 outputs the presence of the obstacle as a determination result to the control command generation unit 70. In all other cases, except when an obstacle exists, the road surface determination unit 60 outputs the condition of the road surface area R as a determination result to the path generation unit 40.
[0043] Before autonomous driving, the route generation unit 40 generates route data corresponding to a route from the vehicle's own position to the destination based on the destination data D4, map data D1, and vehicle's own position data D5. During autonomous driving, the route generation unit 40 reflects the determination results of the overhead space determination unit 50 and the road surface determination unit 60 in the route data. When a discovered object is present on the route, the route generation unit 40 generates new route data to reach the destination while avoiding the discovered object, and outputs the new route data to the control command generation unit 70. When a cleaning object or a pit is present, the route generation unit 40 generates new route data to reach the destination via the cleaning object or pit, and outputs the new route data to the control command generation unit 70.
[0044] The control command generation unit 70 generates control command data for controlling the autonomous vehicle 100. When the control command generation unit 70 receives input indicating the presence of an obstacle, it outputs a stop command to the traveling motor M1 as control command data. When the control command generation unit 70 receives input of route data, it outputs a traveling command to the traveling motor M1 as control command data, and when the autonomous vehicle reaches a cleaning area or pit, it outputs a cleaning command to the cleaning motor M2 as control command data. When the route data indicates the presence of a slope or step, it outputs a traveling command to the traveling motor M1 to rotate at a speed different from the speed at which the autonomous vehicle travels on a horizontal surface.
[0045] As shown in FIG. 7, control device 20 controls autonomous vehicle 100 in the following procedure. 1) Before autonomous driving, map data D1 is stored in advance in storage unit 80 (step S1). Specifically, before autonomous driving, autonomous vehicle 100 is driven in the planned driving location. As a result, first sensor 11 acquires overhead data D2, and self-position estimation unit 30 generates map data D1, and finally generates self-position data D5 corresponding to the self-position where the vehicle stopped. Map data D1 and self-position data D5 are output from self-position estimation unit 30 to storage unit 80 and stored in storage unit 80. Note that map data for the planned driving location may be stored in storage unit 80 in advance before driving in the planned driving location. The registered map data D1 may be corrected based on data from driving.
[0046] 2) An operator inputs destination data D4 into the destination input device 13 (step S2). The destination data D4 is output from the destination input device 13 to the storage unit 80 and stored in the storage unit 80. 3) When the autonomous vehicle 100 is stopped, in other words, before autonomous driving, the route generation unit 40 generates route data for heading from the self-position before autonomous driving to the destination based on the destination data D4, the self-position data D5, and the map data D1 (step S3).
[0047] 4) The first sensor 11 acquires the upper data D2 and outputs the upper data D2 to the upper space determination unit 50 and the self-position estimation unit 30 (step S4). 5) The self-position estimation unit 30 generates self-position data D5 based on the map data D1 and the overhead data D2, and outputs the self-position data D5 to the route generation unit 40 (step S5). 6) The upper space determination unit 50 determines whether or not an obstacle is present (step S6). If no obstacle is present, the upper space determination unit 50 outputs the upper data D2 to the path generation unit 40 as the situation of the upper space U. If an obstacle is present, the upper space determination unit 50 outputs the position of the obstacle to the control command generation unit 70. 7) If an obstacle is present, the control command generator 70 outputs a stop command to the traveling motor M1 (step S7), and the control device 20 ends the control.
[0048] 8) The second sensor 12 acquires road surface data D3 and outputs the road surface data D3 to the road surface determination unit 60 (step S8). 9) The road surface determination unit 60 determines the road surface conditions (step S9). If no obstacle is present, the road surface determination unit 60 outputs the road surface data D3 to the path generation unit 40 as the conditions of the road surface area R. If an obstacle is present, the road surface determination unit 60 outputs the position of the obstacle to the control command generation unit 70. 10) If an obstacle is present, the control command generator 70 outputs a stop command to the traveling motor M1 (step 10), and the control device 20 ends the control.
[0049] 11) Route generation unit 40 reflects the latest self-position data D5, overhead data D2, and road surface data D3 in the route data (step S11). That is, route generation unit 40 generates route data based on the latest self-position data D5, destination data D4, map data D1, overhead data D2, and road surface data D3. The route data is data for directing autonomous vehicle 100 to the destination or cleaning site when the latest self-position data D5 and destination data D4 indicate different positions, and is data for stopping autonomous vehicle 100 when the latest self-position data D5 and destination data D4 indicate the same position. The route data is output to control command generation unit 70.
[0050] 12) The control command generator 70 outputs a travel command to the travel motor M1 or a cleaning command to the cleaning motor M2 based on the route data (step S12). 13) More specifically, as shown in Fig. 8, the control command generator 70 outputs a driving command to the driving motor M1 (step S13). The driving command includes not only a command to rotate the driving motor M1 but also a command to stop it. 14) Next, the control command generator 70 determines whether or not a cleaning area exists based on the route data (step S14). 15) If a cleaning site exists, the control command generator 70 determines whether or not the cleaning site has been reached based on the route data until the cleaning site is reached (step S15). 16) When the cleaning area is reached, the control command generator 70 outputs a cleaning command to the cleaning motor M2 (step S16). 17) The control command generator 70 determines whether or not cleaning has been completed until the cleaning is completed (step S17). 18) When cleaning is completed or when there is no cleaning area, the control command generation unit 70 determines whether the destination has been reached as shown in Fig. 1 (step S18). If the destination has been reached, the control device 20 ends the control, and if the destination has not been reached, the process returns to 12).
[0051] The planned travel location is, for example, the inside of a livestock barn. The inside of the livestock barn has a livestock stall unit 91U and a floor surface 95, as shown in FIG. Floor surface 95 has an area where livestock stall unit 91U is installed and an area where autonomous vehicle 100 travels. The area where autonomous vehicle 100 travels is the road surface. The road surface is, for example, a straight line in the front-to-rear direction. The livestock unit 91U has a plurality of livestock pens 91. The livestock pens 91 house livestock. The livestock pens 91 are shaped like a rectangular parallelepiped. The livestock pen units 91U are arranged such that the livestock pens 91 are stacked vertically and adjacent to each other in the front-to-rear direction. For example, droppings 92 fall from the livestock pens 91 onto the road surface as cleaning material. The first sensor 11 detects the livestock pen units 91U. The autonomous vehicle 100 can autonomously travel along the livestock pen units 91U while maintaining a distance from the livestock pen units 91U.
[0052] In autonomous vehicle 100, first sensor 11 detects an upper space U, and second sensor 12 detects a road surface area R. Autonomous vehicle 100 determines the condition of road surface area R based on the correlation between the distance data included in all first windows W1 and the correlation between the distance data included in second window W2. Therefore, the load of the calculation process during the determination corresponds to the total number of distance data included in all first windows W1 and the total number of distance data included in all second windows W2. The smaller the sum of the total number of distance data included in all first windows W1 and the total number of distance data included in all second windows W2 is compared to the total number of distance data included in road surface data D3, the smaller the load of the calculation process during the determination and the faster the calculation process. Autonomous vehicle 100 determines the condition of road surface region R based on the correlation of the first dimensionless quantity included in all first windows W1 and the correlation of the second dimensionless quantity included in second window W2, which reduces measurement errors compared to determining the condition of road surface region R based on the correlation of distance data. The first dimensionless quantity and the second dimensionless quantity have fewer digits than distance data, which reduces the load on calculation processing.
[0053] Because the first sensor is a two-dimensional lidar sensor, autonomous vehicle 100 is less expensive than a three-dimensional lidar sensor.
[0054] For example, in the autonomous vehicle disclosed in Utility Model Registration No. 2546531, multiple sensors other than distance sensors and two distance sensors are fixed to the vehicle. The two distance sensors detect the distance to a detectable object installed on the side of the traveling space, and another sensor other than the distance sensors detects the presence or absence of another detectable object installed on the side of the traveling space, thereby allowing the vehicle to travel autonomously.
[0055] Road conditions are not limited to horizontal surfaces, but may include slopes and steps. It is desirable for an autonomous vehicle to drive while reflecting road conditions. However, the autonomous vehicle disclosed in Utility Model Registration No. 2546531 does not reflect road conditions.
[0056] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the gist of the present invention, and all technical matters included in the technical ideas described in the claims are subject to the present invention. The above-described embodiments are preferred examples, but a person skilled in the art can realize various alternatives, modifications, variations, or improvements from the contents disclosed in this specification, and these are included in the technical scope described in the appended claims. [Explanation of symbols]
[0057] 10 vehicles 11 First Sensor 12 Second Sensor 13 Destination input device 14 Vehicle drive unit 15 Cleaning drive unit 20 Control device 30 Self-position estimation part 40 Route generation unit 50 Upper space determination section 60 Road surface determination section 70 Control command generation unit 80 Storage section D1 map data D2 Upward Data D3 Range image data D4 Destination Data D5 Self-location data P pixel U upper space R road area W1 First Window W2 Second Window
Claims
1. Map data and destination data are stored. A first sensor fixed to the vehicle detects the space in front of the vehicle and above the road surface and acquires upward data. Based on the map data and the overhead data, self-position data corresponding to the vehicle's own position is estimated. Based on the aforementioned upper data, the condition of the upper space is determined. A second sensor fixed to the vehicle detects a blind spot area below the detection area of the first sensor and acquires distance image data, which will be the basis for the distance image, as blind spot area data. The aforementioned distance image is in the form of a matrix in which multiple pixels are arranged vertically and horizontally. The distance image data corresponds to a plurality of light-receiving elements that constitute the image sensor of the second sensor and is composed of a plurality of light-receiving element data. The aforementioned light-receiving element data is a combination of position data indicating the position of the light-receiving element within the image sensor and distance data indicating the distance between the detected object detected by the second sensor and the light-receiving element. A first window for selecting some of the pixels in the distance image data in the form of a matrix arranged vertically and horizontally is set to be placeable on the distance image data by selecting some of the light-receiving element data of the distance image data. For each row set for the distance image, the distance data of the light-receiving element data corresponding to each row is searched and one of the maximum and minimum values is extracted as the first singular data. The first window is positioned in the distance image data to include the light-receiving element data associated with the first singular data, Based on the correlation of the distance data contained in all of the first windows arranged in the distance image data, the status of the blind spot area is determined. Based on the map data, destination data, and self-position data, route data for traveling from the self-position to the destination and control command data for controlling the vehicle are generated. An autonomous driving method for a vehicle, which reflects the determination results of the conditions in the overhead space and the determination results of the conditions in the blind spot area into the route data and the control command data.
2. A second window for selecting some of the pixels in the distance image data in the form of a matrix arranged vertically and horizontally is set to be placeable on the distance image data by selecting some of the light-receiving element data from a portion of the distance image data. For each set number of rows in the distance image, the distance data of the light-receiving element corresponding to each row is searched, and the other of the maximum and minimum values is extracted as the second singular data. The second window is positioned in the distance image data to include the photodetector data associated with the second singular data, An autonomous driving method for a vehicle according to claim 1, wherein the status of the blind spot area is determined based on the correlation of the distance data contained in all of the second windows arranged in the distance image data and the correlation of the distance data contained in all of the first windows arranged in the distance image data.
3. A normalization constant is set to non-dimensionalize the distance data to a number of digits smaller than the distance data. The distance data contained in all of the first windows arranged in the distance image data is made dimensionless based on the normalization constant to obtain a first dimensionless quantity. An autonomous driving method for a vehicle according to claim 1 or 2, wherein the condition of the blind spot region is determined based on the first dimensionless quantity.
4. The distance data contained in all of the second windows arranged in the distance image data is made dimensionless based on the normalization constant to obtain a second dimensionless quantity with fewer digits than the distance data. The state of the blind spot region is determined based on the second dimensionless quantity and the first dimensionless quantity. The autonomous driving method for the vehicle described in claim 3.
5. The autonomous driving method for a vehicle according to claim 1 or 2, wherein the determination of the conditions of the blind spot area is whether or not the road surface is a horizontal plane, a slope, or a step.
6. The autonomous driving method for a vehicle according to claim 1 or 2, wherein the location in which the vehicle travels is inside a livestock barn.
7. Vehicles and, A first sensor is fixed to the vehicle at a distance from the road surface and detects the space in front of the vehicle and above the road surface to acquire upward data. A second sensor is fixed to the vehicle and, facing the road surface located in front of the vehicle, detects a blind spot area below the detection area of the first sensor and acquires distance image data, which will be the basis of the distance image, as blind spot area data. A control device for controlling the vehicle, It includes a destination input device for inputting destination data corresponding to a destination, The control device is A storage unit that stores self-position data corresponding to the vehicle's own position, map data, destination data, upward data, and distance image data, A self-position estimation unit that estimates the self-position data based on the map data and the above data, A route generation unit that generates route data corresponding to the route from its own position to the destination. 、 A control command generation unit that generates control command data for controlling the vehicle, An upper space determination unit that determines the state of the upper space based on the aforementioned upper data, It includes a blind spot area determination unit that determines the status of the blind spot area, The distance image data consists of data from multiple photodetectors and corresponds to multiple photodetectors that constitute the image sensor of the second sensor. The aforementioned light-receiving element data is a combination of position data indicating the position of the light-receiving element within the image sensor and distance data indicating the distance between the detected object detected by the second sensor and the light-receiving element. The blind spot determination unit is configured such that a first window can be placed on the distance image data to select some of the pixels in the distance image in the form of a matrix arranged vertically and horizontally by selecting some of the light-receiving element data from a portion of the distance image data, and for each number of rows set on the distance image, it searches for the distance data of the light-receiving element data corresponding to each row and extracts one of the maximum and minimum values as a first singular data, and places the first window on the distance image data to include the light-receiving element data associated with the first singular data, and determines the status of the blind spot area based on the correlation of the distance data included in all of the first windows placed on the distance image data. The route generation unit generates the route data based on the map data, the destination data, and the self-position data, and reflects the determination result of the upper space determination unit and the determination result of the blind spot area determination unit into the route data. The control command generation unit generates the control command data based on the route data, and reflects the determination result of the upper space determination unit and the determination result of the blind spot area determination unit in the control command data, in an autonomous vehicle.