Autonomous driving system for work vehicles

Abstract

This technology provides an autonomous driving system for agricultural vehicles that allows the vehicle to autonomously navigate along the furrows with high precision, regardless of changes in the shape of the furrows. [Solution] This is an autonomous driving system for an agricultural vehicle 10 that performs agricultural work while straddling ridges 5 in a field, and autonomously drives along the ridges 5. The system comprises an agricultural vehicle 10 having a pair of running parts 20 spaced apart to travel on running surfaces 7I and 7II adjacent to one ridge 5, a ridge-following LiDAR 50 installed on the agricultural vehicle 10 so as to include the running surfaces 7I and 7II adjacent to the ridge 5 being straddled by the agricultural vehicle 10 in its detection area, and a control device that estimates the centerline RCL of the ridge 5 being straddled by the agricultural vehicle 10 based on information about the two running surfaces 7I and 7II, and drives the agricultural vehicle 10 along the ridges 5 based on the error between the estimated centerline RCL of the ridge 5 and the centerline of the agricultural vehicle 10.

Description

Technical Field

[0001] The present invention relates to an autonomous driving system for a work vehicle, and more particularly to an autonomous driving system for a work vehicle that autonomously drives along a furrow.

Background Art

[0002] Conventionally, in order to save labor and improve efficiency in agricultural work, various autonomous driving systems have been proposed to automatically follow a furrow with an agricultural work vehicle that performs agricultural work while straddling the furrow in a field. Among such autonomous driving systems, for example, there is one that controls the positional relationship between the furrow and the planting unit based on the positional change of a pressure-sensitive furrow detection plate that contacts the furrow. However, in such an autonomous driving system, there is a problem that the furrow detection plate may collapse or shave the furrow, and the detection accuracy is low because it receives vibrations due to the contact between the furrow detection plate and the furrow.

[0003] Therefore, for example, in Patent Document 1, a traveling vehicle is equipped with a planting unit that can be adjusted to move left and right, and an ultrasonic furrow sensor that detects the distance to a transplanting furrow is provided in the planting unit. Based on the output signal of the furrow sensor, a transplanter has been proposed that controls the left and right movement of the planting unit to perform automatic following of the planting unit along the transplanting furrow.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] In the case of the above Patent Document 1, since an ultrasonic furrow sensor that does not contact the furrow is used, it is possible to suppress the furrow from collapsing or shaving.

[0006] However, in the case of Patent Document 1, the distance between the ridge and the slope is detected by a ridge sensor, but if the shape of the ridge is distorted, it becomes difficult to detect the distance to the ridge, so there is a problem that it has low robustness (the property of being less susceptible to disturbances) to changes in the shape of the ridge.

[0007] The present invention has been made in view of the above, and its objective is to provide a technology for an autonomous driving system for work vehicles that enables a work vehicle to autonomously drive along a furrow with high precision, without being affected by changes in the shape of the furrow. [Means for solving the problem]

[0008] To achieve the above objective, the autonomous driving system for a work vehicle according to the present invention estimates a point on the centerline of a furrow group from a point cloud consisting of multiple points on the driving surface.

[0009] Specifically, the present invention relates to an autonomous driving system for a work vehicle that autonomously drives along the ridges in a field having a group of ridges consisting of multiple rows and a driving surface formed between adjacent ridges, while performing work across the group of ridges.

[0010] Furthermore, this autonomous driving system comprises a distance sensor installed on the work vehicle, an estimation means for estimating the centerline of the furrow group that the work vehicle is straddling based on information detected by the distance sensor, and a control means for driving the work vehicle along the furrow group based on the error between the centerline of the furrow group estimated by the estimation means and the centerline of the work vehicle, and is configured to estimate a point on the centerline of the furrow group from a point cloud consisting of a plurality of points on the driving surface located on both sides of the furrow, and the distance sensor is installed diagonally downward in front of the work vehicle.

[0011] With this configuration, the estimation means estimates the centerline of the furrow group based on information about the running surface, where environmental and morphological changes are smaller than those of the furrows, rather than the furrows themselves, which are susceptible to significant changes in environment and shape due to the growth of planted crops. Therefore, even if the edges of the furrows are locally collapsed, the centerline of the furrow group can be estimated with high accuracy.

[0012] Furthermore, since the centerline of the furrow group is estimated from a point cloud consisting of multiple points on the running surface located on both sides of the furrow, the error is halved compared to estimating the centerline of the furrow group based on information from only one of the running surfaces, thus further improving the accuracy of estimating the centerline of the furrow group.

[0013] Furthermore, since the distance sensor is installed diagonally downward in front of the work vehicle, the error between the estimated center line of the furrow and the center line of the work vehicle can also reflect the posture of the work vehicle, thus enabling the work vehicle to reliably autonomously navigate along the furrow.

[0014] As described above, according to the present invention, since the centerline of the furrow group is estimated based on information about the travel surface where changes in environment and shape are relatively small, the agricultural vehicle can autonomously travel along the furrows with high accuracy, without being affected by changes in the shape of the furrows.

[0015] Furthermore, the point cloud is a point cloud on a flat surface that extends horizontally.

[0016] With this configuration, the estimation means estimates the centerline of the furrow group based on information about flat surfaces extending horizontally on the running surface where environmental and shape changes are smaller than those of the furrows. Therefore, even if the base of the furrows is locally collapsed, the centerline of the furrow group can be estimated with high accuracy.

[0017] Furthermore, when extracting the flat point cloud, the system is configured to obtain a point cloud consisting of multiple points on each slope in the flat sloping portion of the ridge, and a point cloud consisting of multiple points on the top surface of the ridge, and then extract only the flat point cloud from among them.

[0018] Further, the estimation means excludes point groups on the inclined normal planes on both sides of the ridge group from the point groups on the ridge group and on the two running surfaces on both adjacent sides of the ridge group, and excludes the point group on the top surface of the ridge group by extracting the point group included in a predetermined height range which is the height near the lower end of the ridge group, and among the point groups on the respective running surfaces which are the extracted point groups, the midpoint between the point closest to the ridge group among the point groups on one of the running surfaces and the point closest to the ridge group among the point groups on the other running surface is estimated as the point on the center line of the ridge group.

Effect of the Invention

[0019] As described above, according to the autonomous driving system of the agricultural work vehicle according to the present invention, the work vehicle can be autonomously driven along the ridge group with high accuracy without being affected by the shape change of the ridge group.

Brief Description of the Drawings

[0020] [Figure 1] It is a diagram schematically showing an autonomous driving system of an agricultural work vehicle according to an embodiment of the present invention. [Figure 2] It is a side view schematically showing an agricultural work vehicle. [Figure 3] It is a plan view schematically showing an agricultural work vehicle in a state where the crawler pitch is widened. [Figure 4] It is a plan view schematically showing an agricultural work vehicle in a state where the crawler pitch is narrowed. [Figure 5] It is a diagram schematically explaining a variable crawler pitch structure. [Figure 6] It is a perspective view schematically showing the installation position of LiDAR in an agricultural work vehicle. [Figure 7] It is a diagram schematically explaining the detection direction of the ridge-following LiDAR. [Figure 8] It is a diagram schematically explaining the detection area of the marker detection LiDAR. [Figure 9]A diagram schematically illustrating the detection area of a LiDAR for ridge following, where (a) of the figure is a side view and (b) of the figure is a top view. [Figure 10] A diagram schematically illustrating a method for estimating the center line of a ridge in an autonomous driving system. [Figure 11] A diagram schematically illustrating a method for estimating the center line of a ridge in an autonomous driving system. [Figure 12] A diagram schematically illustrating a method for estimating the center line of a ridge in an autonomous driving system. [Figure 13] A diagram schematically illustrating a method for estimating the center line of a ridge in an autonomous driving system. [Figure 14] A diagram schematically illustrating a method for estimating the center line of a ridge in an autonomous driving system. [Figure 15] A diagram schematically illustrating the relationship between the placement position of a LiDAR for ridge following and the detectable point cloud. [Figure 16] A diagram schematically illustrating the relationship between the detection direction of a LiDAR for ridge following when directed straight down and when directed obliquely downward in the front. [Figure 17] A diagram schematically illustrating a method for extracting a flat point cloud according to other embodiments. [Figure 18] A diagram schematically illustrating a conventional method for estimating the center line of a ridge.

Embodiments for Implementing the Invention

[0021] Hereinafter, embodiments for implementing the present invention will be described based on the drawings.

[0022] -Autonomous Driving System- FIG. 1 is a diagram schematically showing an autonomous driving system 1 of an agricultural work vehicle (work vehicle) 10 according to the present embodiment. This autonomous driving system 1 enables the agricultural work vehicle 10 to autonomously drive in a farm field 3 having a plurality of ridges 5 and a driving surface 7 formed between adjacent ridges 5, 5, and on a headland 9 extending in the arrangement direction of the ridges 5 on both outer sides in the longitudinal direction of the ridges 5, without requiring an operation by an operator or the like.

[0023] More specifically, this autonomous driving system 1 autonomously drives an agricultural vehicle 10 along the longitudinal direction of the ridges 5 in a field 3, performing agricultural work while straddling the ridges 5. It also allows the agricultural vehicle 10 to autonomously exit from the ridges 5 to the headlands 9, drive on the headlands 9, and enter the ridges 5 from the headlands 9. To enable such autonomous driving, the autonomous driving system 1 of this embodiment, as shown in Figure 1, includes markers 80 erected at both ends 5f in the longitudinal direction of the ridges 5, an agricultural vehicle 10, a ridge-following LiDAR 50, a marker detection LiDAR 60 (see Figure 3, etc.), and a control device 70 (see Figure 2).

[0024] <marker> The markers 80 are, for example, wooden stakes, and as shown in Figure 1, they are driven into the ground at both ends 5f in the longitudinal direction of every other furrow 5. There are no particular limitations on the material or shape of the markers 80. In this embodiment, the markers 80 are placed at every other furrow 5, but as will be described later, if the farm vehicle 10 crosses over the markers 80, the markers 80 may be placed at both ends 5f in the longitudinal direction of all furrows 5.

[0025] <Agricultural vehicles> Figure 2 is a schematic side view of the agricultural vehicle 10, Figure 3 is a schematic top view of the agricultural vehicle 10 with the crawler pitch widened, and Figure 4 is a schematic top view of the agricultural vehicle 10 with the crawler pitch narrowed. The agricultural vehicle 10 performs weeding work while straddling the ridges 5 in the field 3, and as shown in Figures 2 to 4, it is equipped with a pair of left and right running sections 20, a frame section 30, a frame lifting mechanism 36, a weeding attachment 40, an attachment lifting mechanism 45, a ridge-following LiDAR 50, a marker detection LiDAR 60, and a control device 70.

[0026] The left and right pairs of running sections 20 are spaced apart from each other in the width direction of the machine body so as to run on the running surfaces 7, 7 adjacent to each of the furrows 5. As shown in Figure 2, each running section 20 has a drive wheel 21, a driven wheel 22, two idler wheels 23, 24, a frame 25, a crawler belt 26, a running motor unit 27, and a running battery 29. The drive wheel 21, the driven wheel 22, and the two idler wheels 23, 24 are rotatably supported on the frame 25. The crawler belt 26 is wrapped around the drive wheel 21, the driven wheel 22, and the two idler wheels 23, 24 so as to be in contact with them. The running motor unit 27 is an integrated unit of motor and reduction gear and is mounted in a wheel-like manner on the drive wheel 21. The running battery 29 is located on the frame 25 above the running motor unit 27.

[0027] Each running unit 20 is configured to travel on the running surface 7 by the driving force of the drive wheels 21 (running motor unit 27), which are driven by power supplied from the running battery 29, and the crawler belt 26 rotates while being guided by the driven wheels 22 and two idler wheels 23 and 24. The forward (forward rotation), reverse (reverse rotation), and stopping of each running unit 20 are controlled by the control device 70.

[0028] As shown in Figures 3 and 4, the frame section 30 is formed in a substantially rectangular shape in plan view and is provided above the pair of running sections 20 so as to span across them. The frame section 30 has a pair of left and right vertical frame members 31 extending in the front-rear direction of the machine body, and first to fourth horizontal frame members 32, 33, 34, and 35 connecting the pair of left and right vertical frame members 31 in the width direction of the machine body. As shown in Figure 2, the pair of left and right vertical frame members 31 are connected to the frames 25 of the pair of left and right running sections 20 via a frame lifting mechanism 36. As a result, the crawler pitch can be changed by changing the width of the frame section 30.

[0029] Figure 5 is a schematic diagram illustrating the crawler pitch variable structure. As shown in Figure 5, the first to fourth transverse frame members 32, 33, 34, and 35 each have a pair of left and right outer pipe sections 32a, 33a, 34a, and 35a, and inner pipe sections 32b, 33b, 34b, and 35b that are fitted inside the pair of left and right outer pipe sections 32a, 33a, 34a, and 35a. Of the first to fourth transverse frame members 32, 33, 34, and 35, the second transverse frame member 33 is provided with a width variable screw section 33c.

[0030] In this frame section 30, when the width-adjustable screw section 33c is manually rotated in one direction, as shown by the black arrows in Figure 5, the left and right outer pipe sections 33a of the second horizontal frame member 33 move closer together, and the inner pipe section 33b fits within the left and right outer pipe sections 33a. Consequently, in the first, third, and fourth horizontal frame members 32, 34, and 35, the inner pipe sections 32b, 34b, and 35b also fit within the left and right outer pipe sections 32a, 34a, and 35a, and as shown in Figure 4, the distance between the left and right pairs of vertical frame members 31 in the width direction of the machine body becomes relatively narrower. Since the left and right pairs of vertical frame members 31 are connected to the left and right pairs of running sections 20 frames 25, rotating the width-adjustable screw section 33c in one direction causes the crawler pitch to become relatively narrower.

[0031] On the other hand, when the width-adjustable screw portion 33c is manually rotated in the other direction, as shown by the white arrows in Figure 5, the left and right outer pipe portions 33a of the second horizontal frame member 33 are separated, and the inner pipe portion 33b is exposed from the left and right outer pipe portions 33a. Consequently, in the first, third, and fourth horizontal frame members 32, 34, and 35, the inner pipe portions 32b, 34b, and 35b are exposed from the left and right outer pipe portions 32a, 34a, and 35a, respectively, as shown in Figure 3, the distance between the left and right pairs of vertical frame members 31 in the width direction of the machine body is relatively increased, and as a result, the crawler pitch is relatively increased.

[0032] As shown in Figure 2, the platform lifting mechanism 36 includes a first link arm 37, a second link arm 38, and an electric hydraulic cylinder 39. The first and second link arms 37 and 38 are connected in a roughly X shape, as shown in Figure 2, by the intermediate portions of the first link arm 37 and the second link arm 38 being pivotably connected to each other. The upper end of the first link arm 37 is rotatably and slidably connected to the vertical frame member 31, while its lower end is rotatably and non-slidably connected to the frame 25 of the running section 20. On the other hand, the upper end of the second link arm 38 is rotatably and non-slidably connected to the vertical frame member 31, while its lower end is rotatably and slidably connected to the frame 25 of the running section 20. The electric hydraulic cylinder 39 is fixed to the vertical frame member 31, and the tip of the rod 39a is connected to the upper end of the first link arm 37.

[0033] In the frame lifting mechanism 36 configured in this way, when the rod 39a of the electric hydraulic cylinder 39 extends, the upper end of the first link arm 37 and the upper end of the second link arm 38 move apart in the front-rear direction, and the upper end of the first link arm 37 and the lower end of the second link arm 38, and the lower end of the first link arm 37 and the upper end of the second link arm 38, move closer together in the vertical direction, causing the frame section 30 to descend. On the other hand, when the rod 39a of the electric hydraulic cylinder 39 retracts, the upper end of the first link arm 37 and the upper end of the second link arm 38 move closer together in the front-rear direction, and the upper end of the first link arm 37 and the lower end of the second link arm 38, and the lower end of the first link arm 37 and the upper end of the second link arm 38, move further apart in the vertical direction, causing the frame section 30 to rise.

[0034] The weeding attachment 40 is mounted to the farm vehicle 10 via an attachment lifting mechanism 45 so as to be able to move up and down, and is configured to remove weeds growing between the furrows 5 when towed by the farm vehicle 10. Specifically, the weeding attachment 40 has a cultivator 41 for weeding and a rake 43 for leveling the soil clumps removed from the cultivator 41.

[0035] <LiDAR for ridge tracing> Figure 6 is a schematic perspective view showing the installation positions of the LiDAR 50 and 60 on the agricultural vehicle 10. The furrow-following LiDAR (distance sensor) 50 is a 2D-LiDAR with an aperture angle of 270°, suitable for measurement and detection on a plane, and measures the distance to an object by measuring the time it takes for the laser beam to hit the object and reflect back. As shown in Figure 6, this furrow-following LiDAR 50 is attached to the center of the machine width direction at the front end of the mounting base 30 (first horizontal frame member 32) via the first to third brackets 51, 55, and 65.

[0036] More specifically, the first bracket 51 has a rectangular rear wall portion 52 that is bolted to the first horizontal frame member 32 of the frame portion 30 and extends downward, a rectangular horizontal wall portion 53 that extends forward from the lower end of the rear wall portion 52, and a rectangular front wall portion 54 that extends downward from the front end of the horizontal wall portion 53. The second bracket 55 has a rectangular support wall portion 56 that is parallel to the front wall portion 54, and a mounting wall portion 57 that extends forward from the left end of the support wall portion 56. Furthermore, the third bracket 65 has a rectangular support wall portion 66 that is parallel to the mounting wall portion 57, and a rectangular mounting wall portion 67 that extends to the right in the width direction of the machine from the upper end of the support wall portion 66. The support wall portion 56 has an arc-shaped elongated hole 56a that extends in the width direction of the machine body and curves downward, while the mounting wall portion 57 has an arc-shaped elongated hole 57a that extends in the front-rear direction and curves diagonally downward.

[0037] The second bracket 55 is attached to the first bracket 51 with a variable left-right angle by fastening its support wall portion 56 to the front wall portion 54 with a bolt 58 slidably inserted into an elongated hole 56a. The third bracket 65 is attached to the second bracket 55 with a variable front-back angle by fastening its support wall portion 66 to the mounting wall portion 57 with a bolt 59 slidably inserted into an elongated hole 57a. Thus, the ridge-following LiDAR 50 is attached to the upper surface of the mounting wall portion 67 of the third bracket 65. Therefore, the ridge-following LiDAR 50 is attached to the frame portion 30 with a variable left-right angle and a variable front-back angle via the first to third brackets 51, 55, and 65.

[0038] Figure 7 is a schematic diagram illustrating the detection direction of the ridge-following LiDAR 50. As shown in Figure 7, the ridge-following LiDAR 50 is mounted at the center of the machine width direction at the front end of the frame 30, with its front-to-back angle set so that the detection plane DP and the ground G form a 30° angle. Therefore, the ridge-following LiDAR 50 is capable of detecting the distance to the ridges 5 and the driving surface 7 in front of the agricultural vehicle 10. The information regarding the ridges 5 and the driving surface 7 detected by the ridge-following LiDAR 50 is input to the control device 70.

[0039] <LiDAR for marker detection> The marker detection LiDAR 60, like the furrow-following LiDAR 50, is a 2D-LiDAR with an aperture angle of 270°. As shown in Figure 6, this marker detection LiDAR 60 is attached to the front end of the right end in the width direction of the machine body of the frame section 30 via a lifting bracket 61.

[0040] More specifically, the lifting bracket 61 has a rectangular cylindrical support portion 62 attached by welding or the like to the front end of the right vertical frame member 31 of the frame portion 30, a vertically extending rectangular cylindrical support pipe 63 fitted and inserted inside the support portion 62, and a mounting plate 64 provided at the lower end of the support pipe 63. The marker detection LiDAR 60 is mounted on the upper surface of the mounting plate 64.

[0041] Multiple recesses 63a are formed on the left side of the support pipe 63 in the aircraft width direction, arranged at equal intervals in the vertical direction. Similarly, two recesses 62a are formed on the left side of the support part 62 in the aircraft width direction, arranged vertically at the same intervals as the recesses 63a. Therefore, when the support pipe 63 is fitted into the support part 62, the two recesses 62a of the support part 62 fit into the vertically arranged recesses 63a of the support pipe 63.

[0042] The recess 62a of the support part 62 is formed to be deep enough that when it fits into the recess 63a of the support pipe 63, it can support the weight of the support pipe 63, the mounting plate 64, and the marker detection LiDAR 60. However, if an operator moves the support pipe 63 up or down, the recess 63a of the support pipe 63 will allow it to be dislodged from the recess 63a of the support pipe 63. In other words, the height of the marker detection LiDAR 60 is variable relative to the base part 30 by moving the support pipe 63 up and down relative to the support part 62. As a result, even if the base part 30 is raised by the base lifting mechanism 36, for example, the height of the marker detection LiDAR 60 can be kept constant by pulling down the support pipe 63.

[0043] Figure 8 is a schematic diagram illustrating the detection area DA of the marker detection LiDAR 60. In this embodiment, the marker detection LiDAR 60 is positioned at the right end in the width direction of the front end of the agricultural vehicle 10, and its aperture angle is 270°. As shown in Figure 8, the area behind the marker detection LiDAR 60 and to the left in the width direction of the vehicle is a blind spot, while the remaining area is the detection area DA. Note that the detection area DA in Figure 8 is merely a schematic representation, and in reality, the marker detection LiDAR 60 is configured to detect objects at a much greater distance. Information about objects detected by the marker detection LiDAR 60 is input to the control device 70.

[0044] <Control device> As shown in Figure 2, the control device 70 is installed, for example, on the frame 30. The control device 70 controls the electric hydraulic cylinder 39 of the frame lifting mechanism 36 and the attachment lifting mechanism 45, and also controls the travel motor unit 27 of the travel unit 20 based on the detection results of the furrow-following LiDAR 50 and the marker detection LiDAR 60, so that the agricultural vehicle 10 can travel in the field 3 or travel and turn in the headland 9.

[0045] For example, if the control device 70 rotates both the left and right drive motor units 27 forward, the agricultural vehicle 10 will move forward, while if it rotates both the left and right drive motor units 27 backward, the agricultural vehicle 10 will move backward. Furthermore, if the control device 70 creates a difference in rotational speed between the left and right drive motor units 27, the agricultural vehicle 10 will turn in the direction of the slower-rotating drive motor unit 27. Moreover, if the control device 70 rotates one drive motor unit 27 forward and the other drive motor unit 27 backward, the agricultural vehicle 10 will perform a pivot turn.

[0046] -Autonomous driving- Next, we will describe the autonomous driving of the agricultural vehicle 10 in this embodiment.

[0047] As described above, the autonomous driving system 1 of this embodiment (1) allows the farm vehicle 10 to autonomously exit from the furrow 5 to the headland 9, drive on the headland 9, and enter the furrow 5 from the headland 9, and (2) allows the farm vehicle 10 to autonomously drive along the longitudinal direction of the furrow 5.

[0048] <(1) Regarding autonomous driving and autonomous turning on the headland> In this embodiment, when the farm vehicle 10 moves from the furrow 5 to the headland 9, travels on the headland 9, and then enters the furrow 5 from the headland 9, the farm vehicle 10 is driven and turned while simultaneously detecting multiple markers 80 using the marker detection LiDAR 60.

[0049] Specifically, when the farm vehicle 10, which is traveling across the furrows 5, approaches the headland 9, the marker detection LiDAR 60 detects multiple markers 80 that have been placed on the longitudinal ends 5f of the furrows 5, which are spaced alternately. Based on this detection result, the control device 70 acquires the end (end point) of the furrow 5 being worked on and causes the farm vehicle 10 to exit from the furrows 5 to the headland 9.

[0050] Here, if the turning center of the farm vehicle 10 is shifted to the left or right, the position of the centerline VCL of the farm vehicle 10 relative to the headland 9 after turning will be different depending on whether it turns to the right or to the left. For this reason, in this embodiment, when the farm vehicle 10 is turned at the headland 9, the control device 70 is configured to turn the farm vehicle 10 only in the direction in which the marker detection LiDAR 60 is located, that is, to the right.

[0051] Therefore, once exit to the headland 9 is complete, the marker detection LiDAR 60 detects multiple markers 80, and the control device 70 turns the farm vehicle 10 to the right (clockwise). However, since the turning direction is limited to the direction in which the marker detection LiDAR 60 is positioned, the multiple markers 80 do not enter the blind spot of the marker detection LiDAR 60, and thus the multiple markers 80 can continue to be detected even during the turn. In this way, since the marker detection LiDAR 60 continues to detect multiple markers 80 from exiting to the headland 9 until the completion of the turn, the attitude of the farm vehicle 10 relative to the furrow 5 can be continuously calculated, thereby improving the accuracy of the turning angle.

[0052] Then, when the farm vehicle 10, which is traveling forward or backward on the headland 9 after completing the turn, approaches the next furrow 5, the marker detection LiDAR 60 detects multiple markers 80 placed near the next furrow 5, and the control device 70 causes the farm vehicle 10 to turn right again. In this case as well, the marker detection LiDAR 60 continues to detect multiple markers 80, and after the turn is completed, based on this detection result, the control device 70 acquires the end (starting position) of the next furrow 5 and causes the farm vehicle 10 to enter the furrow 5 from the headland 9.

[0053] As described above, in the autonomous driving system 1 of this embodiment, it is possible to improve the turning angle accuracy with a single marker detection LiDAR 60 without using direction sensors or the like, so that the agricultural vehicle 10 can autonomously exit from and enter the furrow 5 with high accuracy while keeping costs down.

[0054] <(2) Regarding autonomous driving along the furrows> In order to autonomously drive the farm vehicle 10 along the furrow 5, it is necessary to obtain the reference centerline RCL of the furrow 5. Below, the estimation method for the centerline RCL of the furrow 5 in the autonomous driving system 1 of this embodiment will be described, but in order to make the present invention easier to understand, a conventional estimation method for the centerline RCL of the furrow 5 will be described first.

[0055] Figure 18 is a schematic diagram illustrating a conventional method for estimating the centerline RCL of a furrow 5. In the conventional estimation method, as shown in Figure 18, the distance sensor 150 installed on the farm vehicle detects one of the left and right slopes 5b of the furrow 5 between the running surface 7 on which a pair of left and right running parts 120 are running, that is, the furrow 5 straddled by the farm vehicle, and estimates the centerline RCL of the furrow 5 based on the detection result.

[0056] However, with these conventional estimation methods, it becomes difficult to detect the distance to the slope 5b when the shape of the ridge 5 is distorted (for example, when the base 5d of the slope 5b of the ridge 5 is distorted, as shown in Figure 18), and therefore the method has the problem of being less robust to changes in the shape of the ridge 5.

[0057] Moreover, conventional estimation methods estimate the centerline RCL of the furrow 5 based on information from one side of the slope 5b, which means they are susceptible to detection errors. Therefore, it is conceivable to increase the number of distance sensors 150 to acquire information from both sides of the furrow 5, but this would increase costs.

[0058] Therefore, in the autonomous driving system 1 according to this embodiment, the agricultural vehicle 10 is controlled based on information about the driving surfaces 7 adjacent to the furrow 5 that the agricultural vehicle 10 is straddling. Specifically, the control device 70 is configured to estimate the centerline RCL of the furrow 5 straddling the agricultural vehicle 10 based on information about the two driving surfaces 7 detected by the furrow-following LiDAR 50, and to make the agricultural vehicle 10 travel along the furrow 5 based on the error between the estimated centerline RCL of the furrow 5 and the centerline VCL of the agricultural vehicle 10.

[0059] Figure 9 schematically illustrates the detection area DA of the ridge-following LiDAR 50, where Figure (a) is a side view and Figure (b) is a top view. In Figure 9, arrow X indicates the front side in the longitudinal direction of the machine, arrow Y indicates the left side in the width direction of the machine, and arrow Z indicates the upper side in the vertical direction. As described above, the ridge-following LiDAR 50 is mounted at the center of the machine width direction at the front end of the frame 30 such that the detection plane DP and the ground G form a 30° angle. Therefore, as shown in Figure 9(a), it detects the distance to the ridges 5 and the running surface 7 in front of the farm vehicle 10. Consequently, as shown in Figure 9(b), the ridge-following LiDAR 50 includes the ridges 5I, 5II, and 5III in front of the farm vehicle 10, as well as the running surfaces 7I and 7II further forward than the ridges 5I, 5II, and 5III, in its detection area DA.

[0060] Figures 10 to 14 schematically illustrate the estimation method for the centerline RCL of furrow 5 in the autonomous driving system 1. When estimating the centerline RCL of furrow 5, the furrow-following LiDAR 50 detects furrow 5, which is straddled by the farm vehicle 10, and the adjacent driving surfaces 7I and 7II, as shown in Figure 10. More specifically, from the multiple points P detected by the furrow-following LiDAR 50, the control device 70 extracts a point group consisting of multiple points on the left driving surface 7I, a point group consisting of multiple points on the left slope 5b of furrow 5, a point group consisting of multiple points on the top surface 5a of furrow 5, a point group consisting of multiple points on the right slope 5c of furrow 5, and a point group consisting of multiple points on the right driving surface 7II, as shown by the dashed frame in Figure 10.

[0061] Next, the control device 70 extracts flat point clouds from these point clouds. More specifically, the control device 70 determines the normal of the plane on which each point P exists from the information around that point P, and extracts only the point clouds on flat planes where the normal is approximately perpendicular. Through this process, as shown in Figure 11, point clouds on the left slope 5b of the ridge 5 and on the right slope 5c of the ridge 5, where the normal is inclined, are excluded. As a result, as shown by the dashed box in Figure 11, the control device 70 extracts the point cloud on the left running surface 7I, the point cloud on the top surface 5a of the ridge 5, and the point cloud on the right running surface 7II.

[0062] Next, the control device 70 extracts point clouds that fall within a predetermined height range from the point clouds on the left running surface 7I, the point clouds on the top surface 5a of the ridge 5, and the point clouds on the right running surface 7II. Here, the "predetermined height range" can be, for example, the height near the bottom end of the ridge 5. Through this process, the point clouds on the top surface 5a of the ridge 5 are excluded, and as shown by the dashed frame in Figure 12, only the point clouds on the left running surface 7I and the point clouds on the right running surface 7II are clustered by the control device 70.

[0063] Next, as shown by the dashed frame in Figure 13, the control device 70 estimates the midpoint of the point PI closest to the furrow 5 in the point cloud on the left running surface 7I and the point PII closest to the furrow 5 in the point cloud on the right running surface 7II as the centerline RCL of the furrow 5 (more precisely, a point on the centerline RCL). Then, based on the error between the estimated centerline RCL of the furrow 5 and the centerline VCL of the farming vehicle 10, the control device 70 drives the farming vehicle 10 along the furrow 5.

[0064] For example, if the centerline VCL of the farm vehicle 10 is to the left of the estimated centerline RCL of the furrow 5, the control device 70 drives the travel motor unit 27 of the left travel unit 20 at a faster rotational speed than the travel motor unit 27 of the right travel unit 20 in order to steer the farm vehicle 10 to the right. Also, if the deviation between the centerline VCL of the farm vehicle 10 and the estimated centerline RCL of the furrow 5 is greater than a first predetermined amount, the control device 70 corrects the direction of travel of the farm vehicle 10 while reducing its speed. Furthermore, if the deviation between the centerline VCL of the farm vehicle 10 and the estimated centerline RCL of the furrow 5 is greater than a second predetermined amount (> first predetermined amount), the control device 70 corrects the direction of travel of the farm vehicle 10 by performing a pivot turn.

[0065] As described above, according to this embodiment, the control device 70 estimates the centerline RCL of the ridge 5 based on information about the adjacent running surfaces 7I and 7II, which are less susceptible to changes in environment and shape than the ridge 5 itself, due to the growth of planted crops, etc. Therefore, as shown in Figure 18, even if the base 5d of the ridge 5 is locally collapsed, the centerline RCL of the ridge 5 can be estimated with high accuracy.

[0066] Furthermore, since the centerline RCL of furrow 5 is estimated from a point cloud that satisfies the condition that the adjacent running surfaces 7I and 7II are included within a predetermined flat height range, even if there are some irregularities in the running surfaces 7I and 7II, the running surfaces 7I and 7II can be determined under the same conditions on both sides of furrow 5. Therefore, the centerline RCL of furrow 5 can be estimated with even greater accuracy while reflecting the shape of furrow 5.

[0067] Here, we will explain the case where the base of furrow 5 is locally collapsed. In the example shown in Figure 14, the left base of the original furrow 5 is point 5d1, so the midpoint between point 5d1 and the right base of furrow 5 (point 5d2) becomes the centerline RCL of furrow 5. If we were to estimate the centerline RCL of furrow 5 based on point 5d1', which is the closest point to furrow 5 in the point cloud on the left running surface 7I, the distance ΔE between point 5d1 and point 5d1' would directly become the error, resulting in a relatively large error ΔE.

[0068] In this embodiment, the control device 70 estimates the midpoint of the point 5d1' closest to the furrow 5 in the point cloud on the left running surface 7I and the point 5d2 closest to the furrow 5 in the point cloud on the right running surface 7II as a point on the centerline RCL' of the furrow 5. Therefore, even if the base of the furrow 5 is locally collapsed, the error (ΔE / 2) is halved, and the estimation accuracy of the centerline RCL' of the furrow 5 can be further improved.

[0069] Figure 15 schematically illustrates the relationship between the placement of the LiDAR 50 for ridge tracing and the detectable point cloud. In Figure 15, the region RI is defined as the area between two virtual extension planes VP1 and VP2, which are extensions of the slopes 5b and 5c on both sides of the ridge 5, at a height of VH or higher where they intersect, and between the two virtual extension planes VP1 and VP2. The region between two virtual extension planes VP1 and VP2, at a height of less than VH where they intersect, and between the two virtual extension planes VP1 and VP2, is defined as Region RIII. The region to the left of Region RI and Region RIII is defined as Region RII, and the region to the right of Region RI and Region RIII is defined as Region RIV.

[0070] As shown in region RIII of Figure 15, if the installation height of the ridge-following LiDAR 50C on the farm vehicle 10 is too low, the ridge-following LiDAR 50C can only detect the top surface 5a of the ridge 5, making it difficult to detect the adjacent travel surfaces 7I and 7II on either side of the ridge 5. Also, as shown in region RII of Figure 15, if the installation position of the ridge-following LiDAR 50B in the width direction of the farm vehicle 10 is too far to the left, the slope 5e of the ridge 5 will obstruct the view of the ridge-following LiDAR 50B, causing it to detect point 5d2' on the right travel surface 7II as the inner endpoint 5d2, resulting in an error ΔE between the actual center line RCL of the ridge 5 and the detected point, as shown in Figure 15. The same applies to region RIV of Figure 15. In these cases, it becomes difficult to accurately estimate the centerline RCL of the furrow 5, making it difficult to autonomously drive the farm vehicle 10 along the furrow 5 with high accuracy. In the worst case, it is conceivable that the farm vehicle 10 may come into contact with the furrow 5.

[0071] Therefore, as shown in Figure 12, it is preferable to install the LiDAR 50 for furrow tracing on the farm vehicle 10 at a height of VH or higher where two virtual extension surfaces VP1 and VP2, which are extensions of the slopes 5b and 5c on both sides of the furrow 5 straddled by the farm vehicle 10, intersect, and at a position in the width direction of the vehicle such that it is contained within the area between the two virtual extension surfaces VP1 and VP2, i.e., within the area RI. By positioning the LiDAR 50 for furrow tracing so that it is contained within the area RI, it is possible to always detect the running surfaces 7I and 7II adjacent to the furrow 5 using the LiDAR 50 for furrow tracing. In other words, as long as it is contained within the area RI, even if the position of the LiDAR 50 for furrow tracing shifts slightly to the left or right, as in the case of the LiDAR 50A for furrow tracing, it is possible to always detect the running surfaces 7I and 7II adjacent to the furrow 5.

[0072] In this embodiment, as described above, the LiDAR 50 for ridge tracing is mounted in the center of the machine width direction at the front end of the frame 30, and the installation height is variable by changing the height of the frame 30 with the frame lifting mechanism 36. As a result, the LiDAR 50 for ridge tracing can always be kept within the region RI, and thus it is possible to always detect the running surfaces 7I and 7II adjacent to the ridge 5.

[0073] Figure 16 schematically illustrates the relationship between the detection direction of the ridge-following LiDAR 50 when it is pointed directly downwards and when it is pointed diagonally downwards and forwards. In Figure 16, error ΔE0 is the deviation of the center line VCL of the farm vehicle 10 relative to the center line RCL of the ridge 5 when the detection direction of the ridge-following LiDAR 50 is pointed directly downwards, and error ΔE1 in Figure 16 is the deviation of the center line VCL of the farm vehicle 10 relative to the center line RCL of the ridge 5 when the detection direction of the ridge-following LiDAR 50 is pointed diagonally downwards and forwards.

[0074] Generally, if the centerline VCL of the farm vehicle 10 straddling the furrow 5II is tilted relative to the centerline RCL of the furrow 5II, the distance between the two centerlines is small directly beneath the farm vehicle 10 and increases as it moves away from the farm vehicle 10. Therefore, if the furrow-following LiDAR 50 is installed so that the running surfaces 7I and 7II directly beneath the farm vehicle 10 are included in the detection area, the attitude (direction) of the farm vehicle 10 is unlikely to be reflected in the error ΔE0 between the centerline RCL of the furrow 5 and the center of the farm vehicle 10.

[0075] Furthermore, as shown in Figure 16, if the posture of the farm vehicle 10 is taken into account, even though the center line VCL of the farm vehicle 10 is shifted to the left by an error ΔE1 relative to the center line RCL of the furrow 5, it may be detected that the center line VCL of the farm vehicle 10 is shifted to the right by an error ΔE0 relative to the center line RCL of the furrow 5 directly beneath the farm vehicle 10. In such a case, if the farm vehicle is steered to the left to make the error ΔE0 zero, it is conceivable that the right rear end of the farm vehicle 10 may come into contact with the slope of the furrow III on the right.

[0076] In this embodiment, as described above, the LiDAR 50 for ridge tracing is installed on the farm vehicle 10 facing diagonally downwards forward so that the front driving surface 7 of the farm vehicle 10 becomes the detection area. Therefore, the error ΔE1 between the estimated center line RCL of the ridge 5 and the center line VCL of the farm vehicle 10 can also reflect the attitude of the farm vehicle 10, thus enabling the farm vehicle 10 to reliably autonomously travel along the ridge 5.

[0077] (Other embodiments) The present invention is not limited to its embodiments and can be implemented in various other ways without departing from its spirit or main features.

[0078] In the above embodiment, the crawler pitch of the agricultural vehicle 10 was set to perform weeding work while straddling one row of furrows 5. However, the crawler pitch of the agricultural vehicle 10 may be set to perform weeding work while straddling two or more rows of furrows 5 (groups of furrows).

[0079] Furthermore, in the above embodiment, the normal vector was determined from the information around each point to extract a point cloud on a flat surface. However, this is not limited to this, and for example, as shown in Figure 17, a predetermined height range (H1, H2 in the example of Figure 17) may be set, and the point cloud included in this predetermined height range may be estimated to be a point cloud on a flat surface. In Figure 17, PI is the point closest to the furrow 5 on the left running surface 7I included in the height range H1, PII is the point closest to the furrow 5 on the right running surface 7II included in the height range H1, and PI' is the point closest to the furrow 5 on the height range H2(

[0080] ​In this case as well, an error ΔE may occur between A1 and A2 depending on how the height range is set. However, by estimating the centerline RCL of furrow 5 based on information about the two running surfaces 7I and 7II, the error is halved, thus improving the accuracy of estimating the centerline RCL of furrow 5.

[0081] Thus, the embodiments described above are merely illustrative in all respects and should not be interpreted restrictively. Furthermore, any modifications or changes that fall within the equivalent scope of the claims are all within the scope of the present invention. [Industrial applicability]

[0082] According to the present invention, agricultural vehicles can autonomously navigate along furrows with high precision, regardless of changes in the shape of the furrows, making it extremely beneficial for application to autonomous driving systems for agricultural vehicles. [Explanation of Symbols]

[0083] 1. Autonomous Driving System 3 Fields 5 ridges 5b Slope 5c slope 7 Running surface 10 Agricultural vehicles 20. Running section 50 LiDAR (distance sensor) for furrow tracing 70 Control device (estimation means) (control means) E error RCL Centerline RI area VCL-centric VP1 Virtual Extension Plane VP2 virtual extension plane P point

Claims

1. An autonomous driving system for a work vehicle that autonomously drives along the ridges in a field having a group of ridges consisting of multiple rows and a driving surface formed between adjacent ridges, while performing work across the group of ridges, A distance sensor installed on the aforementioned work vehicle, Estimation means for estimating the center line of the furrow group straddling the work vehicle based on information detected by the distance sensor, The system includes a control means for causing the work vehicle to travel along the furrows based on the error between the center line of the furrows estimated by the estimation means and the center line of the work vehicle, The system is configured to estimate a point on the centerline of the furrow group from a point cluster consisting of multiple points on the running surface located on both sides of the furrow, The distance sensor is characterized by being installed diagonally downward in front of the work vehicle. Autonomous driving system for work vehicles.

2. The point cloud is characterized by being a point cloud on a flat surface extending in the horizontal direction. An autonomous driving system for a work vehicle as described in claim 1.

3. When extracting the flat point cloud, Furthermore, the method is characterized by obtaining a point group consisting of multiple points on each slope in the flat sloping portion of the ridge, and a point group consisting of multiple points on the top surface of the ridge, and then extracting only the flat point group from among them. The autonomous driving system for a work vehicle according to claim 2.

4. The autonomous driving system for a work vehicle according to claim 1, characterized in that the estimation means is configured to exclude the point clouds on the slopes on both sides of the ridge group from the point clouds on the ridge group and the two adjacent running surfaces of the ridge group, and to exclude the point clouds on the top surface of the ridge group by extracting the point clouds that are included in a predetermined height range which is the height near the bottom end of the ridge group, and to estimate the midpoint of the point on the center line of the ridge group from the extracted point clouds on each running surface, which are the point clouds on each running surface, between the point closest to the ridge group in one of the point clouds on the running surface and the point closest to the ridge group in the point cloud on the other running surface.

Images