JP7318892B2 - self-driving forklift - Google Patents

Description

The present invention is an automatic operation forklift that automatically recognizes the position of a pallet or the like placed at a predetermined position in a vast facility such as a distribution warehouse and facilitates the transportation of cargo or the like placed on the pallet. It is about.

Conventionally, in vast facilities such as distribution warehouses, products are loaded on pallets and stored in designated locations, and forklifts are used to load and unload products while they are still on pallets. Move pallets. As automation of various kinds of work in the logistics industry progresses in recent years, automation of unloading work by forklifts is expected, and various automatically operated forklifts have been proposed or developed.

Automatic recognition of the shape of the opening into which the pallet and its forks are to be inserted is important for an automatically operating forklift. However, the pallets may be placed, for example, outdoors in strong sunlight or in deep dark places in warehouses, and the pallets themselves are made of plastic material or wood, so automatic recognition is not easy.

For example, automatic pallet recognition using a three-dimensional image such as a TOF camera is not suitable for detecting a black object or a surface exposed to sunlight, and the pallet detection accuracy is greatly reduced. Further, for example, the automatic recognition of a palette based on an image captured by a camera is susceptible to the texture of the object surface, ambient light, and the like, and the object cannot be recognized when the lighting is dark. Therefore, it is required to reliably recognize the pallet even in such an environment.

On the other hand, automatic pallet recognition using a two-dimensional lidar is effective because it is less susceptible to the above-described textures and ambient light, but cannot acquire three-dimensional environmental information.

According to the forklift having such a configuration, the height position of the laser sensor is adjusted by moving the fork up and down, so that the two-dimensional point cloud obtained by scanning the laser sensor and the height position are combined to obtain a three-dimensional point cloud. It is designed to recognize loads or pallets.

On the other hand, in Patent Document 1, a two-dimensional sensor, which is a laser sensor attached to a fork, is scanned in a first direction, and the fork is raised and lowered to detect an object existing in a space set ahead. To specify the position and orientation of the cargo or pallet to be picked up by measuring the three-dimensional distance data to the vehicle, and move the vehicle to the unloading position calculated based on the specified location and orientation of the cargo or pallet. A forklift is disclosed that is adapted to generate trajectory data of .

Japanese Patent No. 6469506

By the way, in the forklift according to Patent Document 1, the laser sensor is directly attached to the fork, and the height position of the laser sensor is adjusted by moving the fork up and down. There is a problem that the sensor is damaged or the detection accuracy is lowered. In addition, it is necessary to raise and lower the forks in order to recognize the load or pallet by the laser sensor, and there is a risk of getting caught due to the raising and lowering of the forks. Furthermore, when a load or pallet is loaded with a fork, the field of view of the laser sensor is occupied by the load or pallet, so the laser sensor cannot be used. , or signs and the like provided on the floor, shelves, walls, etc., in the distribution warehouse.

In view of the above points, the present invention provides an automatic operation forklift that can accurately recognize an object in front with a simple configuration and can recognize an object in front even when a load or a pallet is loaded on the fork. intended to provide.

In order to achieve the above object, the present invention provides a vehicle body, a running section for running and steering the vehicle body, a fork attached to the front part of the vehicle body and supported by a mast so as to be vertically movable, and a fork that can be lifted up and down. an automatic driving forklift comprising: a fork driving unit that allows the vehicle to move, a position sensor, and a control unit that controls the traveling unit and the fork driving unit based on traveling data while referring to the current position from the position sensor, the sensor further comprising: A 2D lidar with an object detection unit consisting of a part and a shape recognition part, a sensor part attached to the vehicle body, and arranged so that the horizontal scanning direction is inclined downward toward the front, and the horizontal scanning of the 2D lidar. and a moving means for moving in a direction different from the direction of , the moving means being fixed to the vehicle body separately from the movable fork, and a direct acting actuator for vertically translating the two-dimensional rider. and a height sensor that detects the height position of the 2D lidar, and the shape recognition unit adds the height position detected by the height sensor to the 2D point cloud acquired by the horizontal scanning of the 2D lidar. to acquire a 3D point cloud, select the 3D point cloud belonging to the vertical plane, detect the vertical plane from a slice region of a certain height, and recognize the shape and position of the object located in front Consists of configuration. The linear actuator may be fixed to the side of the mast outside the fork.

The shape recognition unit recognizes the shape and position of the opening of the pallet by extracting the left and right column portions of the opening of the pallet to be conveyed by the fork from the three-dimensional point cloud belonging to the vertical plane .
Based on the relative positional relationship between the recognized shape and position of the pallet and the opening and the current position detected by the position sensor, the control unit creates travel data up to the pick-up position where the fork directly faces the opening of the pallet. The traveling unit is controlled based on this traveling data. It is preferable that the shape recognition unit recognizes the shape and position of the front floor surface and the markers installed on the floor surface from the three-dimensional point cloud.

To provide an extremely excellent automatic driving forklift capable of accurately recognizing an object in front with a simple configuration and recognizing an object in front even when a load or a pallet is mounted on the fork according to the present invention. can be done.

1 is a schematic diagram showing the overall configuration of an embodiment of an automatically operating forklift according to the present invention; FIG. FIG. 2 is a block diagram showing the configuration of the automatic operation forklift in FIG. 1; 1. (A) is a schematic perspective view of the laser navigator shown in FIG. 1, (B) is a schematic plan view showing a laser beam irradiation state, and (C) is a schematic plan view showing a position detection state in a distribution warehouse. FIG. 3 is a schematic diagram showing a basic route of travel data created by a control unit of an automatically operating forklift; The control unit shows an example of creating travel data for approaching the pickup position, (A) is the relative positional relationship between the current position of the automatic operation forklift and the pickup position, and (B) is based on forward movement only. FIG. 4 is a schematic diagram showing a travel route based on travel data; The control unit shows another example of creation of travel data for approaching the pickup position, (A) is the relative positional relationship between the current position of the automatic operation forklift and the pickup position, and (B) is backward and forward. FIG. 4 is a schematic diagram showing a travel route based on travel data in which forward movement is combined; Fig. 2 shows the arrangement of a sensor unit of an object detection unit, where (A) is a schematic perspective view and (B) is a side view; FIG. 4 is a schematic diagram showing measurement states at different height positions for recognition of the pallet and its opening by the object detection unit; (A) to (F) are graphs showing 2D point clouds acquired by a 2D LIDAR at 6 different heights with successively changing heights. 10 is a graph showing a three-dimensional point cloud calculated by adding a height position to the two-dimensional point cloud at each height position in FIG. 9; FIG. 11 is a schematic diagram showing a slice region for recognizing the shape of the pallet and its opening from the three-dimensional point cloud of FIG. 10; FIG. 12 is a schematic diagram showing a palette extracted from each slice region of FIG. 11 and the shape of its opening; FIG. 4 is a schematic diagram showing measurement states at different height positions for recognition of a floor surface and markers by an object detection unit; FIG. 2 is a flow chart sequentially showing the cargo handling work of the automatically operated forklift of FIG. 1 ; FIG. 5 is a flow chart sequentially showing an operation of recognizing the shape and position of the pallet and its opening by the object detection unit;

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below based on embodiments shown in the drawings.
1 and 2 show the overall configuration of an embodiment of an automatically operated forklift. The automatic operation forklift 10 is configured, for example, as a reach type forklift, and includes a vehicle body 11, a traveling unit 12 that automatically moves the vehicle body 11 back and forth and automatically steers, and a fork 13 provided in the front part of the vehicle body 11. , a fork drive unit 14 that moves the fork 13 up and down, a detection unit 15 that acquires various information, a control unit 16 provided inside the vehicle body 11, and an object detection unit 20.

The vehicle body 11 to the control unit 16 described above have the same configuration as that of a conventional automatic operation forklift. That is, the vehicle body 11 has a substantially rectangular parallelepiped outer shape, and is provided with projecting portions 11a extending forward from both left and right sides. has a rear wheel 11c. A rear wheel 11c is supported as a steering wheel so as to be able to swing about a vertical axis.

The traveling unit 12 rotates the front wheels 11b or the rear wheels 11c to cause the automatically operated forklift 10 to travel forward and backward, and steers the rear wheels 11c to turn left and right. The fork 13 is vertically movably supported along a pair of masts 13 a fixedly arranged to extend vertically in the front portion of the vehicle body 11 , and can be vertically moved by the fork driving section 14 . The detector 15 includes a laser navigator serving as a position sensor 15a provided above the vehicle body 11, and an obstacle sensor 15b and a bumper sensor 15c provided below the vehicle body 11. FIG.

FIG. 3(A) shows a schematic perspective view of the position sensor 15a in the automatic operation forklift 10, and (B) and (C) show the laser beam irradiation state and the position detection state in the distribution warehouse, respectively. The position sensor 15a is composed of a laser scanner 15d and a light receiving unit 15e, and emits a laser beam from the laser scanner 15d, and is arranged on a wall surface of a facility 15f such as a distribution warehouse as shown in FIG. 3(C). The current position and direction in the distribution warehouse 15f are determined by collecting the laser beams reflected by the plurality of reflecting portions 15g by the light receiving portion 15e. Thus, the laser navigator 15a functions as a highly accurate position sensor.

The obstacle sensor 15b is composed of, for example, a two-dimensional lidar, detects an obstacle behind the vehicle body 11 of the automatically operated forklift 10, and sends out a detection signal. The bumper sensor 15c detects the obstacle and sends out a detection signal when it comes into contact with the obstacle. Further, the automatic operation forklift 10 is provided with an emergency stop button 17a, a voice/pattern lamp 17b, and a blinker 17c as a safety section 17 in order to further enhance safety.

The basic route 16c of the travel data 16b created by the control unit 16 will be described with reference to FIG. The control unit 16 creates travel data 16b based on the pickup position and the unloading position input by the operation unit 16a (see FIG. 1) provided on the outer surface of the vehicle body, and controls the travel unit 12 based on the travel data 16b. to control. That is, the control unit 16 causes the automatically operating forklift 10 to travel along a basic route 16c (described later) set, for example, in a distribution warehouse, thereby moving the automatically operating forklift 10 to the vicinity of the designated pickup position. , After the pallet 30 is loaded on the fork 13 at the loading position, the pallet 30 is similarly moved along the basic path 16c to the vicinity of the designated unloading position, and the pallet 30 is unloaded from the fork 13 at the unloading position. . At that time, based on detection signals from the laser navigator 15a, the obstacle sensor 15b, and the bumper sensor 15c of the detection unit 15, the control unit 16 detects obstacles, people, or other automatic objects along the travel route based on the travel data 16b during travel. Drive along the travel path by slowing down, stopping or detouring as appropriate to avoid contact with the operating forklift.

The control unit 16 has the map data in the facility 15f described above, and also registers the positions of the basic route 16c and the marker 16d shown in FIG. Furthermore, the control unit 16 is sequentially updated when the position of the marker 16d is changed, such as newly established or abolished, and always has the latest data. The basic path 16c is a path for moving within the facility 15f as shown in FIG. Alternatively, the marker 16d indicating the unloading position is accessed by branching from the basic route 16c.

When the control unit 16 arrives near the pick-up position along the basic route 16c based on the travel data 16b, the control unit 16 detects the shape and position of the pallet 30 and its opening 31 recognized by the object detection unit 20, which will be described later, and The current position detected by the laser navigator 15a of the detector 15 and the relative positional relationship are calculated. Based on this relative positional relationship, the control unit 16 creates traveling data 16b up to the pickup position where the tip of the fork 13 faces the opening 31 of the pallet 30 straight. to control. As a result, the tip of the fork 13 of the automatically operated forklift 10 directly faces the opening 31 of the pallet 30 arranged at the unloading position.

By the way, in response to the relative positional relationship between the current position of the automatically operated forklift 10 and the unloading position, if the forklift truck 10 moves forward as it is as shown in FIG. If it is not possible to face the vehicle straight, as shown in FIG. On the other hand, as shown in FIG. 6(A), when the angle with respect to the pallet 30 is greatly deviated and it is not possible to face the opening 31 of the pallet 30 at the unloading position in a straight line even if it turns around only by moving forward. 6(B), the control unit 16 creates travel data 16b such that the vehicle once reverses and then reverses the steering to move forward.

Further, when the control unit 16 arrives near the unloading position along the basic route 16c based on the travel data 16b, the control unit 16 detects the shape and position of the floor surface marker 16d similarly sent from the object detection unit 20. A relative positional relationship with the current position detected by the laser navigator 15a of the unit 15 is calculated. Based on this relative positional relationship, the control unit 16 creates traveling data 16b up to the unloading position where the fork 13 is aligned on the marker 16d, and controls the traveling unit 12 based on this traveling data 16b. As a result, the fork 13 of the automatically operated forklift 10 is aligned with the marker 16d on the floor.

Here, cargo handling work is performed as follows. That is, the automatic operation forklift 10 advances from a state in which the forks 13 face the opening 31 of the pallet 30 straight at the unloading position, and the fork 13 is completely inserted into the opening 31 of the pallet 30. Fork 13 is raised. Thus, the pallet 30 is mounted on the fork 13 and loaded. In the case of unloading, the automatic operation forklift 10 is at the unloading position, and the fork 13 and the mounted pallet 30 are positioned above the marker 16d on the floor until the bottom surface of the pallet 30 contacts the floor. is lowered and retreated, the fork 13 is completely pulled out from the opening 31 of the pallet 30, and unloading is completed.

As described above, the control unit 16 can automatically operate the recognized pallet 30 and the shape and position of the opening 31 of the pallet 30 from the current position detected by the position sensor 15a to the pickup position according to the traveling data 16b. can. Therefore, the fork 13 can enter the opening 31 of the pallet 30 straight by the control unit 16 controlling the traveling unit 12 to move forward from the state of moving to the loading position.

The sensor unit 22 acquires a three-dimensional point cloud on the floor near the marker 16d indicating the place where the pallet 30 is placed, and the shape recognition unit 24 extracts the boundary line of the marker 16d based on the difference in reflectance of the marker 16d. to determine a match with the shape of the known marker 16d to recognize the shape and position of the marker 16d. As a result, the control unit 16 recognizes the shape and position of the floor surface and the marker 16d with the sensor unit 22 in a state where the load or pallet 30 is mounted on the fork 13, thereby automatically driving to the position of the designated marker 16d. , and the cargo or pallet 30 can be unloaded at the location indicated by the marker 16d.

Further, the object detection unit 20 includes a sensor unit 22 and a shape recognition unit 24, as shown in FIG. The shape recognition unit 24 recognizes the shape and position of an object positioned ahead from the three-dimensional point group acquired by the sensor unit 22, as will be described later.

According to the automatic operation forklift 10 of the present invention, the sensor unit 22 of the object detection unit 20 acquires the front three-dimensional point group while the automatic operation forklift 10 is positioned near the unloading position, and the shape recognition unit 24 recognizes the shape and position of an object located in front based on the distance data of individual pixels from the 3D point cloud. Thereby, the control unit 16 can grasp the relative position between the vehicle body 11 and the object from the shape and position of the forward object and the current position detected by the position sensor 15a.

Since the sensor section 22 of the object detection section 20 is attached to the vehicle body 11 of the automatic operation forklift 10, the following advantages are obtained compared to the case where the sensor section 22 is attached to the fork 13. That is, since the sensor section 22 is directly attached to the vehicle body 11 instead of the fork 13, the sensor section 22 can detect an object or the like in front even when the load or the pallet 30 is mounted on the fork 13. . Therefore, it is possible to recognize not only the pallet 30 placed at the pickup position, but also the floor surface, the markers 16d installed on the floor surface, or other objects such as the walls and shelves of the warehouse.

In addition, since it is not necessary to move the fork 13 up and down or swing it when the sensor unit 22 recognizes an object in front, the risk of being caught by the fork movement is eliminated. Furthermore, while the forks 13 are susceptible to impact and vibration during cargo handling work of the automatically operated forklift 10, by attaching the sensor unit 22 to the vehicle body 11, damage to the sensor unit 22 due to the impact and vibration of the forks 13 can be avoided.

Next, some configuration examples of the object detection unit 20 will be described.
(Configuration example 1 of object detection unit)
As a configuration example 1 of the object detection unit 20, the sensor unit 22 is composed of a two-dimensional lidar 22a and a moving means 21 for moving in a direction different from the horizontal scanning direction of the two-dimensional lidar 22a. In this case, the shape recognition unit 24 acquires the height position corresponding to the movement amount of the moving means 21, adds the height position to the two-dimensional point group acquired by the horizontal scanning of the two-dimensional lidar 22a, and obtains the three-dimensional The original point cloud can be calculated.

In configuration example 1, the two-dimensional lidar 22a is used as the sensor unit 22, and by moving the two-dimensional lidar 22a in a direction different from the horizontal scanning of the two-dimensional lidar 22a by the moving means 21, for example, in the vertical direction, the two-dimensional point by the two-dimensional lidar 22a is detected. In addition to acquiring the group, the height position corresponding to the amount of movement of the moving means 21 is acquired, and the height position is added to the two-dimensional point group to calculate the three-dimensional point group. As a result, a highly accurate three-dimensional point group can be acquired using the two-dimensional lidar 22a with high detection accuracy, so that the shape and position of an object in front can be recognized with high accuracy.

Therefore, compared to the case of recognizing an object using a 3D image such as a TOF camera, by using the 2D lidar 22a as the sensor unit 22, accuracy does not decrease even when detecting a black object or a surface exposed to sunlight. In addition, it is less likely to be affected by the texture of the surface of the object, ambient light, etc., and can reliably recognize the object even when the lighting is dark.

(Configuration example 2 of object detection unit)
A configuration example 2 of the object detection unit 20 is a configuration in which the sensor unit 22 is a two-dimensional rider 22a, and the moving means 21 in the height direction of the two-dimensional rider 22a is a linear actuator 21a. As shown in detail in FIG. 7, the object detection unit 20 is mounted on one of a pair of vertically extending masts 13a fixedly arranged on the front part of the vehicle body 11 of the automatic operation forklift 10. A linear motion actuator 21a, a two-dimensional lidar 22a, a height sensor 23, and a shape recognition section 24 are attached so as to extend in the vertical direction.

The linear motion actuator 21a has a longitudinally extending fixed portion 21b fixed to the side surface of the mast 13a, and a movable portion 21c moves vertically along the main body 21b. The linear motion actuator 21a is used to recognize the shape and position of the pallet 30 or the floor surface marker 16d by the control unit 16 when the automatic operation forklift 10 arrives near the loading position or the unloading position, for example. It operates in conjunction with the two-dimensional lidar 22a as follows.

The two-dimensional lidar 22a is attached to the movable portion 21c of the linear motion actuator 21a, and has a scanning plane 22b extending horizontally as shown in FIG. O is set downward toward the front. The lidar described above is a sensor that is also called a laser radar, and it is a sensor that performs light detection and ranging (LIDAR) or laser imaging detection and ranging (Laser Imaging Detection and Ranging). be. A two-dimensional lidar is used here.

Specifically, for example, Tim581 manufactured by SICK is used as the two-dimensional lidar 22a. The specifications of this two-dimensional lidar 22a are a horizontal aperture angle of 270 degrees, a scanning frequency of 15 Hz, an angular resolution of 0.33 degrees, and a detection distance (reflectance of 10%) of 8 m. The mounting height and inclination angle of the two-dimensional rider 22a to the mast 13a are set, as shown in FIG. 7, so that the horizontal scanning plane 22b does not hit the fork 13 and can cover the front floor surface. As a result, the two-dimensional lidar 22a outputs two-dimensional data scanned in the horizontal direction as a two-dimensional point group 22c.

The height sensor 23 is configured using a direct acting actuator 21a. The linear motion actuator 21a is driven by a stepping motor, and the absolute rotation angle of the motor can be obtained by an encoder built into the stepping motor. The height position of the two-dimensional rider 22a is detected by converting the absolute rotation angle of the motor into the height position of the linear actuator 21a based on the known gear ratio used in the linear actuator 21a. The height sensor 23 detects the upward movement distance from the lower end of the movement range of the movable portion 21c in the main body 21b of the direct acting actuator 21a, and outputs the movement distance as the height position h. The height sensor 23 only needs to be able to detect the height positions of the movable portion 21c and the two-dimensional rider 22a, and may be another type of height sensor such as an air pressure sensor.

Although not shown in FIG. 1, the shape recognition unit 24 is provided adjacent to the two-dimensional lidar 22a or inside the vehicle body 11, and is highly sensitive to the two-dimensional point group 22c detected by the horizontal scanning of the two-dimensional lidar 22a. The height position h detected by the height sensor 23 is added and treated as integrated data.

As shown in FIG. 8, by the linear motion actuator 21a, the height position h of the movable part 21c and the two-dimensional rider 22a is continuously set to h1, h2, h3, . Acquisition of the two-dimensional point group 22c by the two-dimensional lidar 22a is repeatedly performed on the pallet 30 placed in front while changing the direction. FIG. 8 shows an example of acquiring a two-dimensional point cloud 22c at three different height positions h1, h2, and h3. For each height position h1 to h3, an upper limit is set for the moving speed of the movable portion 21c of the linear actuator 21a in order to maintain a certain degree of measurement density.

FIGS. 9A to 9F show examples of the two-dimensional point cloud 22c acquired at six different heights by sequentially changing the height, and when the height changes, the acquired two-dimensional point cloud 22c changes. I know that

In this way, the shape recognition unit 24 constructs the two-dimensional point group 22c for each height position h as one three-dimensional point group 25 as shown in FIG. Here, 10 to 15 layers of scan data are obtained with one palette 30 . That is, the height positions of the two-dimensional lidar 22a are scan data obtained at 10 to 15 different heights.

According to the above configuration, while scanning in the horizontal direction with the two-dimensional rider 22a, the two-dimensional rider 22a is vertically translated by the linear motion actuator 21a to change the height position. A three-dimensional point group 25 can be obtained by adding a height position to the dimensional point group 22c.

Subsequently, the shape recognition unit 24 recognizes the shape and position of the pallet 30 and its opening 31 by the calculation method of steps 1 to 7 below.
In step 1, the shape recognition unit 24 sets a three-dimensional target area for the three-dimensional point group 25 constructed from the two-dimensional point group 22c acquired at a predetermined height described in FIG. Select and extract points and exclude points outside the region of interest. Thereby, the shape recognition unit 24 creates a three-dimensional point group A within the target area.
In step 2, the shape recognition unit 24 calculates the normal direction of each point with respect to the three-dimensional point group A, selects points belonging to a vertical plane with an appropriate threshold value, and uses the selected points to determine the point group B to build.
In step 3, the shape recognition unit 24 defines a plurality of slice regions 26a, 26b, . Each is selected and distributed as a set C of point cloud.

In step 4, the shape recognition unit 24 detects a plane perpendicular to each point cloud of the point cloud set C by an appropriate method such as RANSAC, and the distance from the plane is greater than a set threshold. Exclude points.
In step 5, each point group in the point group set C is clustered using the distance as a condition, and the area of each cluster obtained is compared with the area of the circumscribing rectangle of the cluster, as shown in FIG. A sufficiently close cluster 27a is extracted, and clusters 27b, 27c, and 27d close in size to the left and right column portions of the opening 31 of the pallet 30 are extracted. Here, among the extracted clusters 27a to 27d, a set of three horizontally adjacent clusters 27b, 27c, and 27d is called a triplet. The shape recognition unit 24 extracts triplets in which the mutual distance between the three clusters is within a certain range from all combinations of triplets, and sets them as a set D.

At step 6, the shape recognition section 24 repeats the above-described steps 4 and 5 for all the slices 26a, 26b, . . . included in the set C described above. As a result, the shape recognition unit 24 compares all the obtained triplets with the initial point group A, extracts only those with pointsless gaps between the three clusters included in the triplet, Let this gap be set E.
In step 7, the shape recognition unit 24 groups the gaps included in the set E that are close to each other into one group, and after grouping the gaps in the group into one, sets the center position as the position of the unloading hole. Output. Here, if there are a plurality of groups, the shape recognition unit 24 outputs the position of the unloading hole corresponding to one pallet 30 for each group. In this manner, the shape recognition unit 24 recognizes the shape and position of the pallet 30 and its opening 31 located in front and outputs them to the control unit 16 .

Further, when the shape recognition unit 24 recognizes the shape and position of the marker 16d provided on the floor, as shown in FIG. While changing the position h to h1, h2, and h3, the two-dimensional lidar 22a acquires the two-dimensional point group 22c for the marker 16d placed in front repeatedly. As a result, in the same manner as the shape and recognition of the pallet 30, the two-dimensional point group 22c at each height position is integrated to acquire one three-dimensional point group 25, and further the three-dimensional point group A, By obtaining point cloud B, point cloud set C and set D and extracting clusters for marker 16d, the shape and position of marker 16d are recognized.

Next, the operation of the automatic operation forklift 10 according to the present invention will be described according to the flowchart of FIG.
In step ST1, when a loading position and an unloading position are input on the operation panel 16a of the control unit 16, the control unit 16 moves from the current position to the loading position and from the loading position to the unloading position in step ST2. to create running data 16b. As a result, the control unit 16 causes the traveling unit 12 to travel along the basic route 16c to the vicinity of the pickup position by controlling the traveling unit 12 in step ST3.

At step ST4, when the automatically operated forklift 10 arrives near the unloading position, the control unit 16 causes the object detection unit 20 to recognize the shape and position of the pallet 30 placed at the unloading position and its opening 31. , and in step ST5, the object detection unit 20 performs recognition of the shapes and positions of the pallet 30 and its opening 31 according to the flowchart of FIG.

As a result, the control unit 16 controls the fork 13 based on the relative positional relationship between the shape and position of the pallet 30 and its opening 31 recognized by the object detection unit 20 in step ST6 and the current position detected by the position sensor 15a. creates traveling data 16b up to a pickup position that directly faces the opening 31 of the pallet 30, and in step ST7, based on this traveling data 16b, the traveling unit 12 is controlled to move the automatically operating forklift 10 to the pickup position. move up to Subsequently, in step ST8, the control unit 16 controls the traveling unit 12 to slowly move the fork 13 forward while inserting it into the opening 31 of the pallet 30, and then controls the fork driving unit 14 to move the fork 13. The pallet 30 is mounted on the fork 13 by raising it.

Next, in step ST9, the control section 16 controls the traveling section 12 based on the traveling data 16b that has already been created, and causes the traveling section 12 to travel along the basic route 16c from the pickup position to the vicinity of the unloading position. When the automatic operation forklift 10 arrives near the unloading position, in step ST10, the control unit 16 instructs the object detection unit 20 to start recognizing the shape and position of the marker 16d provided at the unloading position. In ST11, the object detection unit 20 performs recognition of the shape and position of the marker 16d.

In step ST12, the control unit 16 unloads the fork 13 aligned on the marker 16d based on the relative positional relationship between the shape and position of the marker 16d recognized by the object detection unit 20 and the current position detected by the position sensor 15a. Travel data 16b up to the starting position is created. In step ST13, based on the created travel data 16b, the traveling unit 12 is controlled to move the automatically operated forklift 10 to the unloading position.
Subsequently, the control unit 16 controls the fork driving unit 14 in step ST14 to slowly lower the forks 13 until the bottom surface of the pallet 30 touches the floor surface, and then controls the traveling unit 12 in step ST15. Then, the fork 13 is completely pulled out from the opening 31 of the pallet 30 to unload.
Thus, the cargo handling work by the automatically operated forklift 10 is completed.

The operation of recognizing the shape and position of the pallet 30 and its opening 31 by the object detection unit 20 is performed as follows according to the flow chart of FIG.
First, in step ST21, when the control unit 16 instructs the object detection unit 20 to start recognition work, in step ST22, the linear motion actuator 21a and the two-dimensional rider 22a of the object detection unit 20 are interlocked, and different The two-dimensional lidar 22a performs horizontal scanning at height positions h1, h2, and h3 to obtain a two-dimensional point group 22c, and the height sensor 23 obtains the height position h at that time. Then, in step ST23, after correcting these two-dimensional point groups 22c by the shape recognition unit 24, one three-dimensional point group 25 is formed by combining them with the height position h.

Thereafter, in step ST24, the shape recognition unit 24 excludes points outside the target region to create a three-dimensional point group A, and in step ST25 selects points belonging to a vertical plane to construct a point group B. At step ST26, points belonging to a plurality of slice regions 26a, 26b, . Thereafter, in step ST27, a plane perpendicular to each point group in the point group set C is detected, and clustering is performed by excluding points far from this plane to extract clusters.

At step ST28, the shape recognizing unit 24 extracts, from among a set of three horizontally adjacent clusters (triplets), those clusters whose mutual distance is within a certain range, and sets them as a set D, and at step ST29. Set E is extracted only for those with pointless gaps between the three clusters contained in the triplet. As a result, in step ST30, the shape recognition unit 24 groups the gaps included in the set E that are close to each other into one group. The shape and position of the pallet 30 and its opening 31 located in front are recognized and output to the control unit 16 .
In this way, the task of recognizing the shape and position of the pallet 30 and its opening 31 is completed.

According to the above configuration, the shape recognition unit 24 can recognize, for example, the pallet 30 placed at the unloading position and its opening 31 from the three-dimensional point group 25 acquired by the sensor unit 22. The shape and position of the pallet 30 can be reliably recognized even if the pallet 30 is not placed exactly on the table.

The present invention can be embodied in various forms without departing from its gist. For example, the direct-acting actuator 21a, the two-dimensional rider 22a, and the height sensor 23 of the object detection unit 20 may be attached to a support member separately provided on the vehicle body 11 without being attached to the mast 13a. In this case, even in the automatic operation forklift 10 having a structure in which the mast 13a swings back and forth to swing the fork 13 up and down, the object detection unit 20 can be fixedly arranged with respect to the vehicle body 11. be.

The automatic operation forklift 10 according to the present invention includes a swinging means for swinging the moving means 21 in a direction different from the horizontal scanning of the two-dimensional rider 22a, that is, upward or downward, and an angle detection for detecting the swinging angle of the swinging means. means. In this configuration, the shape recognition unit 24 calculates the height position of the detection target based on the swing angle detected by the angle detection means, and calculates the height position of the detection target in the two-dimensional point group 22c by the two-dimensional lidar. is added to calculate the three-dimensional point cloud 25 .
According to this configuration, while scanning in the horizontal direction with the two-dimensional lidar 22a, the two-dimensional lidar is vertically swung by the swinging means to change the height position of the detection target, thereby enabling two-dimensional A three-dimensional point group 25 can be obtained by adding a height position to the dimensional point group 22c. In this case, since the swing means swings the two-dimensional rider 22a in the vertical direction, the sensor section 22 can be made compact and simple.

Instead of the above-described embodiment, the sensor unit 22 of the object detection unit 20 is a three-dimensional lidar arranged on the vehicle body 11, and the shape recognition unit 24 detects an object located in front of the three-dimensional point group 25 detected by the three-dimensional lidar. may be recognized. With this configuration, the 3D point cloud 25 can be obtained directly by the 3D rider. , the sensor section 22 can be configured more simply and compactly.

10 Autonomous Operation Forklift 11 Vehicle Body 11a Overhang 11b Front Wheel 11c Rear Wheel 12 Traveling Part 13 Fork 13a Mast 14 Fork Driving Part 15 Detection Part 15a Position Sensor (Laser Navigator)
15b Obstacle sensor 15c Bumper sensor 15d Laser scanner 15e Light receiver 15f Facility 15g Reflector 16 Control unit 16a Operation unit 16b Travel data 16c Basic route 16d Marker 17 Safety unit 17a Emergency stop button 17b Voice/patrol lamp 17c Winker 20 Object detector 21 moving means 21a linear actuator 21b main body 21c movable part 22 sensor part 22a two-dimensional lidar 22b scan plane 22c two-dimensional point group 23 height sensor 24 shape recognition part 25 three-dimensional point group 26a, 26b... slice area 27a, 27b, 27c, 27d cluster 30 pallet 31 opening

Claims (5)

a vehicle body, a running section for running and steering the vehicle body, a fork attached to the front portion of the vehicle body and supported by a mast so as to be vertically movable , a fork driving section for vertically moving the fork, and a position. An automatically operating forklift equipped with a sensor and a control unit that controls the traveling unit and the fork driving unit based on traveling data while referring to the current position from the position sensor,
Furthermore, it is equipped with an object detection section consisting of a sensor section and a shape recognition section,
A two-dimensional lidar in which the sensor unit is attached to the vehicle body and arranged so that the direction of horizontal scanning is inclined downward toward the front, and a movement of moving the two-dimensional lidar in a direction different from the direction of horizontal scanning. consisting of means and
The moving means is fixed to the vehicle body separately from the movable fork, and the moving means detects a linear actuator for vertically translating the two-dimensional rider and the height position of the two-dimensional rider. and a height sensor,
The shape recognition unit acquires a three-dimensional point group by adding the height position detected by the height sensor to the two-dimensional point group acquired by horizontal scanning of the two-dimensional lidar, and acquires a three-dimensional point group belonging to a vertical plane. A self-driving forklift that selects the 3D point cloud to detect a vertical plane from a constant height slice area, and recognizes the shape and position of an object located in front. The self-operating forklift truck according to claim 1, wherein the linear actuator is fixed to a side surface of the mast outside the fork. The shape recognition unit recognizes the shape and position of the opening of the pallet by extracting the left and right column portions of the opening of the pallet to be conveyed by the fork from the three- dimensional point cloud belonging to the vertical plane. The automatic operation forklift according to claim 1 or 2 . Based on the relative positional relationship between the shapes and positions of the pallet and the opening recognized by the control unit and the current position detected by the position sensor, the fork reaches a loading position in which the fork directly faces the opening of the pallet. 4. The automatic operation forklift truck according to claim 3 , wherein the traveling data is created, and the traveling unit is controlled based on the traveling data. The automatic operation forklift truck according to any one of claims 1 to 4 , wherein the shape recognition unit recognizes the shape and position of a front floor surface and a marker installed on the floor surface from the three-dimensional point group.

Images