Obstacle Avoidance Device

The obstacle avoidance device for autonomous vehicles reduces processing load and collision risk by deriving an outer edge from obstacle points, representing it as a polygon or convex hull, and ensuring the vehicle avoids obstacles within its travelable width, addressing the inefficiencies of handling large point groups.

JP7593272B2Active Publication Date: 2024-12-03TOYOTA INDUSTRIES CORP
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
JP2021147008
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-09-09
Publication Date
2024-12-03
Estimated Expiration
2041-09-09

AI Technical Summary

Technical Problem

Autonomous vehicles face increased processing load as the number of obstacle points observed by sensors increases, leading to inefficiencies in obstacle avoidance due to the need to consider and process a large number of points, which can result in redundant calculations and potential collisions.

Method used

An obstacle avoidance device that derives an outer edge from a group of obstacle points, ignoring points inside the edge and representing it as a polygon or convex hull, reducing the load by focusing on the edge to avoid obstacles, and generating additional points to ensure the distance between edges is within the vehicle's travelable width.

Benefits of technology

Reduces processing load and prevents redundant routes by focusing on the outer edge of obstacles, ensuring the vehicle avoids obstacles efficiently while minimizing the risk of collision with unseen areas.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007593272000001
    Figure 0007593272000001
  • Figure 0007593272000002
    Figure 0007593272000002
  • Figure 0007593272000003
    Figure 0007593272000003
Patent Text Reader

Abstract

To reduce a load applied to an obstacle avoiding device when an autonomous traveling body avoids an obstacle.SOLUTION: A control unit 32 executes obstacle point group derivation processing of driving an obstacle point group PCob composed of a plurality of obstacle points Pob indicating the positions of obstacles Ob1, Ob2 relative to an autonomous traveling body 10 based on a result of observation performed by a sensor 31. Subsequently, the control unit 32 executes processing of deriving, from among the plurality of obstacle points Pob derived in the obstacle point group derivation processing, the obstacle points Pob located on outer edges E1, E2 including the obstacle point group PCob as outer edge points Pe1, Pe2. The intervals between the adjacent outer edge points Pe1, Pe2 on the outer edges E1, E2 are less than a possible travel width W1 of the autonomous traveling body 10.SELECTED DRAWING: Figure 7
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an obstacle avoidance device. [Background technology]

[0002] Patent Document 1 discloses an autonomous vehicle equipped with a sensor capable of observing obstacles in the vicinity. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2014-089740 A Summary of the Invention [Problem to be solved by the invention]

[0004] In some cases, an obstacle avoidance device is installed in an autonomous vehicle to avoid obstacles observed by a sensor. The obstacle avoidance device causes the autonomous vehicle to avoid obstacles by taking into account an obstacle point cloud that indicates the positions of the obstacles observed by the sensor. In this case, as the number of points constituting the obstacle point cloud increases, the load on the obstacle avoidance device increases when processing the obstacle point cloud. [Means for solving the problem]

[0005] An obstacle avoidance device that solves the above problem is an obstacle avoidance device for causing an autonomous vehicle to avoid an obstacle observed by a sensor equipped on the autonomous vehicle, and includes an obstacle point cloud derivation unit that derives an obstacle point cloud composed of a plurality of obstacle points indicating the position of the obstacle relative to the autonomous vehicle based on the observation results of the sensor, an outer edge derivation unit that derives, among the plurality of obstacle points, obstacle points located on an outer edge that contains the derived obstacle point cloud as outer edge points, and an avoidance unit that causes the autonomous vehicle to avoid the obstacle indicated by the outer edge points, wherein the distance on the outer edge between adjacent outer edge points is less than the drivable width of the autonomous vehicle.

[0006] According to this, the outer edge that includes the obstacle point cloud includes all of the obstacles indicated by the obstacle point cloud. The outer edge indicates the boundary of the area in which the obstacles exist. The obstacle avoidance device causes the autonomous vehicle to avoid the area indicated by the outer edge points arranged on such an outer edge. This allows the autonomous vehicle to avoid the obstacle based on the outer edge points. At this time, for example, the obstacle point cloud that exists inside the outer edge is not taken into consideration when avoiding the obstacle. This reduces information on the obstacle points arranged inside the outer edge when the autonomous vehicle avoids the obstacle. Therefore, it is possible to reduce the load on the obstacle avoidance device when the autonomous vehicle avoids the obstacle.

[0007] The obstacle avoidance device may also be provided with an outer edge point generation unit that generates outer edge points on the outer edge between adjacent outer edge points so that the distance between adjacent outer edge points on the outer edge is less than the drivable width when the distance between adjacent outer edge points on the outer edge is greater than or equal to the drivable width of the autonomous vehicle.

[0008] According to this, even if the distance on the outer edge between the outer edge points derived from the obstacle point cloud is equal to or greater than the travelable width, the distance becomes less than the travelable width by generating the outer edge points on the outer edge. Therefore, it is possible to prevent the occurrence of the inconvenience that the autonomous vehicle avoids an obstacle by heading toward an area where the distance between the outer edge points is equal to or greater than the travelable width, despite the presence of the obstacle.

[0009] In the above obstacle avoidance device, the outer edge may be a polygon having vertices that are at least some of the outer edge points derived by the outer edge derivation section. According to this, the outer edge is a straight line connecting adjacent outer edge points. Therefore, an outer edge located between two adjacent outer edge points can be expressed as a line segment of the adjacent outer edge points. Therefore, the degree of freedom of the outer edge is smaller than when the outer edge is expressed as a curve connecting adjacent outer edge points. Therefore, the outer edge generating unit can easily determine the position on the outer edge where the outer edge point is to be generated. Therefore, the calculation load on the obstacle avoidance device when the outer edge generating unit generates the outer edge points can be reduced.

[0010] In the obstacle avoidance device described above, the outer edge may be a convex hull having the outer edge points derived by the outer edge derivation section as vertices. This makes it possible to prevent the obstacles represented by the outer edge points from becoming excessively large compared to the obstacles represented by the obstacle point cloud, and therefore to prevent the route taken by the autonomous vehicle to avoid the obstacle from becoming redundant.

[0011] In the above obstacle avoidance device, the obstacle point cloud derivation unit may include an observation point cloud acquisition unit that acquires an observation point cloud indicating the position of the obstacle within an observation area of ​​the sensor based on observation results of the sensor, and an auxiliary point cloud generation unit that generates an auxiliary point cloud within a masked area of ​​the observation area that is blocked by the obstacle based on the observation point cloud acquired by the observation point cloud acquisition unit, and the obstacle point cloud may include the observation point cloud and the auxiliary point cloud.

[0012] According to this, obstacles observed by the sensor are represented as an observation point cloud, and obstacles that may exist in a blocked area that the sensor could not observe due to the obstacle are represented as an auxiliary point cloud. The obstacle avoidance device derives and generates outer edge points based on an obstacle point cloud including the observation point cloud and the auxiliary point cloud, and causes the autonomous vehicle to avoid the obstacles represented by the outer edge points. Therefore, the autonomous vehicle can be caused to avoid the obstacles while taking into consideration the possibility that an obstacle may exist in an area that the sensor could not observe due to the obstacle.

[0013] In the above obstacle avoidance device, the obstacle point cloud may include a virtual observation point cloud indicating the position of the obstacle that was observed by the sensor in the past and that has become outside the observation area of ​​the sensor due to movement of the autonomous vehicle, and the obstacle point cloud derivation unit may include: a movement amount acquisition unit that acquires the amount of movement of the autonomous vehicle from the acquisition of the past observation point cloud to the acquisition of the observation point cloud by the observation point cloud acquisition unit, where the observation point cloud acquired earlier than the observation point cloud acquired by the observation point cloud acquisition unit is defined as a past observation point cloud; and a virtual observation point cloud derivation unit that derives the virtual observation point cloud by moving the past observation point cloud by the movement amount.

[0014] According to this, the virtual observation point cloud is obtained by moving the previously observed observation point cloud by the movement amount of the autonomous vehicle. The virtual observation point cloud indicates where an obstacle indicated by the previously observed observation point cloud is located at the time when the observation point cloud is acquired by the observation point cloud acquisition unit. Therefore, it is possible to cause the autonomous vehicle to avoid the obstacle while taking into consideration the possibility that the previously observed obstacle is located outside the observation area.

[0015] In the above-mentioned obstacle avoidance device, the obstacle point cloud derivation unit may include a virtual auxiliary point cloud derivation unit that derives a virtual auxiliary point cloud located outside an observation area of ​​the sensor by moving the auxiliary point cloud generated previously by the movement amount, and the obstacle point cloud may include the virtual auxiliary point cloud.

[0016] When an obstacle blocks the observation of the sensor, it becomes difficult for the autonomous vehicle to observe the area behind the obstacle with the sensor. Therefore, according to the present configuration, an auxiliary point cloud is generated in such a blocked area. When the previously generated auxiliary point cloud moves outside the observation area of ​​the sensor due to the movement of the autonomous vehicle, the auxiliary point cloud is moved by the movement amount of the autonomous vehicle to derive a virtual auxiliary point cloud. The virtual auxiliary point cloud derived in this manner indicates an area that was in the observation area of ​​the sensor but could not be observed by the sensor due to the obstacle. Since the obstacle point cloud includes the virtual auxiliary point cloud, the autonomous vehicle can avoid the area represented by the virtual auxiliary point cloud when avoiding an obstacle. The autonomous vehicle can avoid the obstacle while avoiding the area that could not be observed by the sensor due to the obstacle. Therefore, the possibility that the autonomous vehicle will come into contact with an object in an area blocked by the obstacle can be suppressed. Effect of the Invention

[0017] According to the present invention, the load on the obstacle avoidance device can be reduced. [Brief description of the drawings]

[0018] [Figure 1] FIG. [Diagram 2] FIG. 2 is a block diagram of an autonomous vehicle. [Diagram 3] 13 is a flowchart showing an obstacle avoidance process. [Figure 4] 13 is a flowchart showing an obstacle point cloud derivation process. [Diagram 5] 13 is a flowchart showing outer edge processing. [Figure 6] FIG. 13 is a diagram showing the transition of a point group during outer edge processing. [Figure 7] FIG. 13 is a diagram showing the transition of a point group during outer edge processing. [Figure 8] FIG. 13 is a diagram showing the transition of a point group during outer edge processing. [Figure 9] FIG. 13 is a diagram showing the transition of a point group during outer edge processing. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] <Configuration> Hereinafter, an embodiment of an autonomous vehicle to which an obstacle avoidance device is applied will be described. As shown in Fig. 1, the autonomous vehicle 10 includes a vehicle 20, a sensor 31 mounted on the vehicle 20, and a control device 32 mounted on the vehicle 20. The vehicle 20 includes a vehicle body 21 and a plurality of wheels 22. The autonomous vehicle 10 is a transport cart that transports loads. The autonomous vehicle 10 of this embodiment is configured to be movable in all directions and to be able to turn on the spot around the vehicle center position P.

[0020] The wheels 22 in this embodiment are omnidirectional wheels. An omnidirectional wheel is a wheel that rotates together with the axle and also allows movement in the axial direction of the axle. Four wheels 22 are provided. By controlling the number of rotations and the direction of rotation of the wheels 22, the vehicle 20 can move in all directions while maintaining the orientation of the body 21, move while changing the orientation of the body 21, and change the orientation of the body 21 without moving. Note that the above-mentioned "omnidirectional" refers to the direction of movement on the road surface or floor surface on which the vehicle 20 runs. Hereinafter, for convenience of explanation, the surface on which the vehicle 20 runs is referred to as a horizontal plane.

[0021] 2, the autonomous vehicle 10 includes a drive mechanism 41 that drives the wheels 22. The drive mechanism 41 includes a motor 42 for rotating the wheels 22, and a motor driver 43 that drives the motor 42. Although not shown in the figure, the number of motors 42 and motor drivers 43 provided is the same as the number of wheels 22. The motor driver 43 controls the rotation speed of the motor 42 in response to a command from the control device 32. The control device 32 can control the traveling direction of the vehicle 20 by controlling the rotation speed of the motor 42 via the motor driver 43.

[0022] Next, the sensor 31 and the control device 32 will be described in detail. 1, the sensor 31 is configured to be able to observe an obstacle Ob1 within an observation area A1. The sensor 31 in this embodiment is an RGB camera including an imaging element such as a CMOS (Complementary Metal Oxide Transistor).

[0023] The observation area A1 is an area extending from the sensor 31 in the observation direction D1. The observation direction D1 is a direction parallel to the horizontal plane. As the observation area A1 moves from the sensor 31 toward the observation direction D1, it expands in a direction parallel to the horizontal plane and perpendicular to the observation direction D1. The observation direction D1 indicates the center of the observation area A1 based on the sensor 31. The observation area A1 is included in a range of a predetermined observation angle from the observation direction D1 in the horizontal plane. The observation angle is, for example, in a range of ±60 degrees with respect to the observation direction D1. The observation angle in this embodiment is the angle of view of the sensor 31. The sensor 31 is mounted at a sensor position Ps that is a position away from the running object center position P. Hereinafter, for convenience of explanation, a virtual boundary line that extends from the sensor position Ps and whose angle with the observation direction D1 is the above-mentioned observation angle is referred to as the observation boundary line B1, B2. In this embodiment, the direction from the running object center position P toward the sensor position Ps in the horizontal plane coincides with the observation direction D1.

[0024] As shown in FIG. 6, when obstacles Ob1 and Ob2 exist within an observation area A1, the observation area A1 is made up of a blocked area A11 and an observable area A12. The shielded area A11 is an area of ​​the observation area A1 where the observation by the sensor 31 is blocked by the obstacles Ob1 and Ob2. In other words, the shielded area A11 is a blind spot of the obstacles Ob1 and Ob2 as seen from the sensor 31, such as the depth of the obstacles Ob1 and Ob2 based on the sensor 31. For example, when the obstacles Ob1 and Ob2 exist in the observation area A1, the sensor 31 observes the parts of the obstacles Ob1 and Ob2 facing the sensor 31. However, the sensor 31 may not be able to observe the areas of the obstacles Ob1 and Ob2 that are blocked by the parts observed by the sensor 31. The areas blocked by the parts of the obstacles Ob1 and Ob2 correspond to the shielded area A11. In this case, the sensor 31 cannot observe the obstacles that may be located in the shielded area A11. Note that the obstacles that may be located within the blocked area A11 are, for example, parts of the obstacles Ob1 and Ob2, and obstacles different from the obstacles Ob1 and Ob2.

[0025] The observable area A12 is an area of ​​the observation area A1 other than the blocked area A11. The observable area A12 is an area in which the field of view of the sensor 31 is not blocked by the obstacles Ob1 and Ob2. The observable area A12 is an area in which the sensor 31 can directly observe the obstacles Ob1 and Ob2.

[0026] As shown in FIG. 2, the control device 32 includes a CPU 33, a storage unit 34 including a RAM and a ROM, and an auxiliary storage device 35. The storage unit 34 stores various programs for controlling the vehicle 20. The control device 32 may include dedicated hardware for executing at least some of the various processes, such as an application specific integrated circuit (ASIC). The control device 32 may be configured as a circuit including one or more processors that operate according to a computer program, one or more dedicated hardware circuits such as an ASIC, or a combination thereof. The processor includes a CPU and memories such as a RAM and a ROM. The storage unit 34 stores program code or instructions configured to cause the CPU to execute the processes. The storage unit 34, i.e., the computer readable medium, includes anything that can be accessed by a general-purpose or dedicated computer. The auxiliary storage device 35 is a rewritable non-volatile memory. Examples of the auxiliary storage device 35 include an EEPROM (Electrically Erasable Programmable Read-Only Memory) and a flash memory. The auxiliary storage device 35 stores information indicating an environmental map. The environmental map is information regarding the physical structure of the surrounding environment of the autonomous vehicle 10, such as the shape and size of the environment in which the autonomous vehicle 10 is used.

[0027] As shown in Fig. 1, when the vehicle 20 is driven, the control device 32 generates a driving route R. The driving route R is a route from the current position, which is the starting point of the vehicle 20, to the final destination. When the vehicle 20 is moving straight, the observation direction D1 is parallel to the driving route R. In Fig. 1, the driving route R and the observation direction D1 overlap. The driving route R can be generated by, for example, converting a drivable route into a grid map and generating a route, or by a potential method using a potential field.

[0028] The control device 32 sets a position on the travel route R a predetermined distance away from the vehicle 20 as a target point Pg, and controls the drive mechanism 41 to travel toward the target point Pg. The control device 32 drives the vehicle 20 so that the observation direction D1 of the sensor 31 faces the traveling direction of the vehicle 20. For ease of explanation, the traveling of the vehicle 20 such that the observation direction D1 coincides with the traveling direction of the vehicle 20 will be referred to as normal traveling below. The traveling direction of the vehicle 20 in normal traveling coincides with the direction in which the travel route R extends.

[0029] The control device 32 performs obstacle avoidance processing. The control device 32 performs obstacle avoidance processing to cause the autonomous vehicle 10 to avoid the obstacles Ob1 and Ob2 observed by the sensor 31. At this time, the control device 32 maintains the distance between the vehicle 20 and the obstacles Ob1 and Ob2 at a predetermined value or more so that the vehicle 20 and the obstacles Ob1 and Ob2 do not come into contact with each other. In this embodiment, the distance between the first obstacle Ob1 and the second obstacle Ob2 is equal to or greater than the travelable width W1 of the autonomous vehicle 10. The travelable width W1 of the autonomous vehicle 10 is the width in which the autonomous vehicle 10 can travel. The travelable width W1 is longer than the maximum dimension of the autonomous vehicle 10 in the horizontal direction by a predetermined margin width. The margin width can be set arbitrarily by the user based on the allowable distance between the vehicle 20 and the obstacles Ob1 and Ob2.

[0030] An example of the obstacle avoidance processing will be described below with reference to Figures 3 and 4. The obstacle avoidance processing is performed while the vehicle 20 is traveling. As shown in Fig. 3, the control device 32 performs an obstacle point cloud derivation process in step S1. As a result, as shown in Fig. 6, the control device 32 derives an obstacle point cloud PCob based on the observation results of the sensor 31. The obstacle point cloud derivation process is repeatedly performed at a predetermined update period. As a result, the obstacle point cloud PCob is updated at the update period. The control device 32 that executes the process of step S1 corresponds to an obstacle point cloud derivation unit of this embodiment.

[0031] The obstacle point cloud PCob indicates the positions of the obstacles Ob1 and Ob2 relative to the autonomous mobile unit 10. For example, the obstacle point cloud PCob is the relative coordinates of the obstacles Ob1 and Ob2 relative to the autonomous mobile unit 10 in a horizontal plane. The obstacle point cloud PCob is composed of one or more obstacle points Pob. The obstacle point Pob is a point indicating the position of the obstacles Ob1 and Ob2. In this embodiment, the obstacle point cloud PCob is a point cloud expressed in the mobile unit coordinates. The mobile unit coordinates are expressed in a two-dimensional coordinate system with the mobile unit center position P as the origin. In the mobile unit coordinates, the direction parallel to the observation direction D1 is the Y direction, and the direction perpendicular to the observation direction D1 is the X direction. The XY plane is parallel to the horizontal plane. The obstacle point cloud PCob in this embodiment includes an observation point cloud PC1, an auxiliary point cloud PC2, a virtual observation point cloud PCi1, and a virtual auxiliary point cloud PCi2. Moreover, the obstacle points Pob in this embodiment include an observation point P1, an auxiliary point P2, a virtual observation point Pi1, and a virtual auxiliary point Pi2.

[0032] The observation point group PC1 indicates the positions of obstacles Ob1, Ob2 within the observation area A1 of the sensor 31. The observation point group PC1 is made up of one or more observation points P1. The auxiliary point group PC2 indicates the position of an obstacle that may be located in the blocked area A11. An obstacle that may be located in the blocked area A11 is, for example, a part of the obstacles Ob1 and Ob2, or an obstacle different from the obstacles Ob1 and Ob2. The auxiliary point group PC2 is composed of one or more auxiliary points P2.

[0033] The virtual observation point group PCi1 indicates the positions of obstacles Ob1, Ob2 previously observed by the sensor 31. The virtual observation point group PCi1 is made up of one or more virtual observation points Pi1.

[0034] The virtual auxiliary point group PCi2 indicates the positions of obstacles Ob1, Ob2 that may be located within the blocked area A11 derived based on the past observation results of the sensor 31. The virtual auxiliary point group PCi2 is made up of one or more virtual auxiliary points Pi2.

[0035] Here, the obstacle point cloud derivation process in step S1 will be described in detail. 4, in step S11, the control device 32 acquires an observation point group PC1 and an observation time t1 based on the observation results of the sensor 31. The control device 32 that executes the process of step S11 corresponds to the observation point group acquisition unit of this embodiment.

[0036] Observation time t1 is the time when the observation result of sensor 31 is obtained. Observation time t1 can also be said to be the time when observation point group PC1 is acquired by control device 32. Control device 32 stores observation point group PC1 and observation time t1 acquired in step S11 in auxiliary storage device 35. Note that acquisition and storage of observation time t1 are optional.

[0037] An example of a method for deriving the observation point group PC1 expressed in vehicle coordinates will be described. The control device 32 acquires image data captured by the sensor 31. The image data may include an object located within the observation area A1. Each pixel of the acquired image data is associated with coordinates on the image data and depth information. The image coordinates are, for example, two-dimensional coordinates with an arbitrary point on the image as the origin, and indicate the direction of the object represented in each pixel relative to the sensor 31. The depth information indicates the distance between the object represented in each pixel and the sensor 31. The depth information is obtained, for example, by performing depth estimation on the image data. The specific aspect of the depth estimation is arbitrary, and examples include SfM (Structure from Motion) and machine learning using a convolution neural network (CNN). The depth information may be obtained separately from a distance sensor or the like.

[0038] The control device 32 derives sensor coordinates corresponding to each pixel from the acquired image coordinates and depth information. The sensor coordinates are, for example, a three-dimensional coordinate system with the sensor position Ps as the origin. In the sensor coordinates, the direction parallel to the observation direction D1 is the Y direction, the direction perpendicular to the observation direction D1 is the X direction, and the direction perpendicular to both the X direction and the Y direction is the Z direction. Here, the X direction and the Y direction are each parallel to the horizontal plane. The sensor coordinates may be expressed as polar coordinates with the sensor position Ps as the origin. In this way, the control device 32 derives a point cloud on the sensor coordinates of the target indicated by each pixel.

[0039] The control device 32 converts the point cloud into the vehicle coordinates based on the X and Y coordinates of the point cloud on the derived sensor coordinates. The control device 32 derives the point cloud obtained by the conversion as the observation point cloud PC1. The conversion deletes the Z coordinate included in the point cloud on the sensor coordinates. In other words, the conversion is a projection from the three-dimensional sensor coordinates to the two-dimensional vehicle coordinates. The control device 32 may convert only the Z coordinates of the derived point cloud on the sensor coordinates that are within a predetermined allowable range into the vehicle coordinates. The allowable range may be appropriately set based on, for example, the height of the step on which the autonomous vehicle 10 can travel or the vehicle height of the autonomous vehicle 10. This prevents, for example, the floor on which the autonomous vehicle 10 travels or the step that the autonomous vehicle 10 can overcome from being derived as the observation point cloud PC1.

[0040] For convenience of explanation, each point group described below is represented in a vehicle coordinate system unless otherwise specified. After step S11 is executed, the process proceeds to step S12, where the control device 32 generates an auxiliary point group PC2 in the masked region A11. The masked region A11 includes the boundary between the masked region A11 and the observable region A12. In this embodiment, at least a part of the auxiliary point group PC2 is generated on the boundary between the masked region A11 and the observable region A12. In addition, at least a part of the auxiliary point group PC2 is disposed on the search region A2.

[0041] As shown in FIG. 1, the search area A2 includes at least a portion of the autonomous vehicle 10. The search area A2 includes the vehicle center position P and the sensor position Ps. In this embodiment, the search area A2 includes the entire autonomous vehicle 10. The search area A2 is rectangular. The size of the search area A2 may be set in advance by a user, or may be set appropriately based on the traveling state of the autonomous vehicle 10, for example, the braking distance of the autonomous vehicle 10 corresponding to the traveling speed.

[0042] The auxiliary point group PC2 is composed of a plurality of auxiliary points P2. The distance between the auxiliary points P2 is less than the travelable width W1 of the autonomous vehicle 10. The travelable width W1 of the autonomous vehicle 10 is the width in which the autonomous vehicle 10 can travel. The travelable width W1 is longer than the maximum dimension in the horizontal direction of the autonomous vehicle 10 by a predetermined margin width. The margin width can be arbitrarily set by the user based on the allowable distance between the vehicle 20 and the obstacle Ob1. Note that, since the observation point group PC1 is used to derive the auxiliary point group PC2, the auxiliary point group PC2 is also a point group indicating the positions of the obstacles Ob1 and Ob2 in the shielded area A11 at the observation time t1. The control device 32 that executes the process of step S12 corresponds to the auxiliary point group generating unit of this embodiment.

[0043] The method of generating the auxiliary point cloud PC2 in the shielded region A11 is arbitrary. For example, the control device 32 may generate the auxiliary point P2 at a position farther away from the observation point P1 with respect to the sensor position Ps on a straight line connecting the sensor position Ps and the observation point P1. The control device 32 may generate the auxiliary point cloud PC2 by deriving the shielded region A11 based on the observation point cloud PC1 and arranging the auxiliary point P2 at an arbitrary position within the derived shielded region A11. In particular, in this embodiment, when generating the auxiliary point cloud PC2, the control device 32 selects the observation point P1 having the largest angle in the direction from the sensor position Ps toward the observation point P1 with respect to the observation direction D1 as a reference. The control device 32 generates the auxiliary point P2 at a position farther away from the observation point P1 with respect to the sensor position Ps on a straight line connecting the sensor position Ps and the selected observation point P1. This allows the control device 32 to generate the auxiliary point cloud PC2 on the boundary between the shielded region A11 and the observable region A12. The distance between the auxiliary points P2 generated on the boundary between the blocked area A11 and the observable area A12 is an arbitrary value less than the travelable width of the autonomous mobile body 10.

[0044] Next, proceeding to step S13, the control device 32 acquires the past observation point cloud, the past auxiliary point cloud, and the past observation time t2. The past observation point cloud is the observation point cloud PC1 acquired earlier than the observation point cloud PC1 acquired in step S11. The past auxiliary point cloud is the auxiliary point cloud PC2 generated earlier than the auxiliary point cloud PC2 generated in step S12. The past observation time t2 is the time when the observation point cloud PC1 was acquired in the past step S11. In this embodiment, the past observation time t2 is also the time when the auxiliary point cloud PC2 was generated in the past step S12.

[0045] In this embodiment, the observation point group PC1, auxiliary point group PC2, and observation time t1 stored in the auxiliary memory device 35 in the obstacle point group derivation process performed in the previous update period are acquired as the past observation point group, past auxiliary point group, and past observation time t2, respectively.

[0046] The control device 32 acquires the amount of movement δ of the autonomous mobile unit 10 from the acquisition of the past observation point group in step S13 to the acquisition of the observation point group PC1 in step S11. The amount of movement δ includes the traveling direction and traveled distance of the autonomous mobile unit 10. In this embodiment, the amount of movement δ is the change in the mobile unit center position P in the absolute coordinate system from the past observation time t2 to the observation time t1. The absolute coordinate system is a coordinate system in which the ground surface on which the wheels 22 are in contact appears stationary.

[0047] The method of acquiring the movement amount δ is arbitrary, but for example, the control device 32 uses odometry to estimate the travel distance and travel direction of the autonomous mobile body 10 from the rotation amount of the wheels 22 and the rotation amount of the motor 42. The control device 32 acquires the movement amount δ from the estimated travel distance and travel direction of the autonomous mobile body 10. Note that the control device 32 may derive the movement amount δ using Simultaneous Localization and Mapping (SLAM). The control device 32 that executes the process of step S14 corresponds to the movement amount acquisition unit of this embodiment.

[0048] Next, the process proceeds to step S15, where the control device 32 derives a virtual observation point group PCi1 and a virtual auxiliary point group PCi2 by moving the past observation point group and the past auxiliary point group by a movement amount δ. In particular, the control device 32 moves past observation point groups whose distances from other past observation point groups are equal to or greater than the omission distance and less than the driving width W1. The omission distance indicates the resolution of the obstacles Ob1, Ob2 recognized by the control device 32. The omission distance can be set arbitrarily based on, for example, the observation accuracy of the obstacles Ob1, Ob2. The control device 32 that executes the process of step S15 corresponds to the virtual observation point group derivation unit and the virtual auxiliary point group derivation unit of this embodiment.

[0049] In this embodiment, control device 32 selects the past observation point group from the previous update cycle, i.e., the past observation point group acquired in step S13, whose distance from the other past observation point groups is equal to or greater than the omittable distance and less than driving width W1. Control device 32 translates the selected past observation point group by the movement amount δ acquired in step S14, to derive virtual observation point group PCi1.

[0050] As described above, the moving object center position P moves by the movement amount δ from the past observation time t2 to the observation time t1. Therefore, the obstacles Ob1 and Ob2 shown in the past observation point group obtained at the past observation time t2 move in the direction in which the moving object center position P moved, i.e., in the opposite direction to the traveling direction of the autonomous moving object 10, by the same distance as the movement distance of the moving object center position P.

[0051] Therefore, the virtual observation point group PCi1 is located at a position moved by the movement amount δ from the past observation point group PC1. In detail, the virtual observation point group PCi1 is located at a position moved from the past observation point group PC1 in the opposite direction to the movement direction of the moving object center position P, the same distance as the movement distance of the moving object center position P. Similarly, the virtual auxiliary point group PCi2 is located at a position moved by the movement amount δ from the past auxiliary point group PC2. Therefore, the point group obtained by translating the past observation point group by the movement amount δ becomes the virtual observation point group PCi1 representing the positions of obstacles Ob1, Ob2 at observation time t1.

[0052] Here, due to the movement of the autonomous mobile unit 10, obstacles Ob1 and Ob2 that were located within the observation area A1 at the past observation time t2 may be located outside the observation area A1 at the observation time t1. In this case, the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 are located outside the observation area A1. Therefore, the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 indicate the positions of obstacles located outside the observation area A1.

[0053] Next, the process proceeds to step S16, where the control device 32 removes the virtual observation point group PCi1 in the observation area A1 and the virtual auxiliary point group PCi2 in the observation area A1. In other words, the control device 32 leaves the virtual observation point group PCi1 outside the observation area A1 and the virtual auxiliary point group PCi2 outside the observation area A1 as the obstacle point group PCob. Whether the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 are within the observation area A1 can be determined by whether the angles formed by the directions in the horizontal plane of the virtual observation point Pi1 and the virtual auxiliary point Pi2 with respect to the sensor position Ps and the observation direction D1 are within the range of the observation angle from the observation direction D1. If the angles are within the range of the observation angle, the control device 32 determines that the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 are within the observation area A1. On the other hand, if the angles are outside the range of the observation angle, the control device 32 determines that the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 are outside the observation area A1. The virtual observation point group PCi1 and the virtual auxiliary point group PCi2 that remain without being removed in step S16 are located outside the observation area A1. Therefore, the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 indicate the positions of the obstacles Ob1, Ob2 that are now located outside the observation area A1 of the sensor 31 due to the movement of the autonomous mobile body 10. The control device 32 that executes the processing of step S16 corresponds to the auxiliary point group deletion unit of this embodiment.

[0054] Next, the process proceeds to step S17, where the control device 32 removes the auxiliary point group PC2 and the virtual auxiliary point group PCi2 from within the observable area A12. Note that in this embodiment, the virtual auxiliary point group PCi2 from within the observable area A12 has been removed in step S16.

[0055] Next, proceeding to step S18, the control device 32 removes the set of observation points PC1, the set of virtual observation points PCi1, the set of auxiliary points PC2, and the set of virtual auxiliary points PCi2 that are located outside the search area A2. 4, the process proceeds to step S19, where the control device 32 derives, as an obstacle point group PCob, those remaining without being removed in the processes from step S16 to step S18, among the observation point group PC1, auxiliary point group PC2, virtual observation point group PCi1, and virtual auxiliary point group PCi2 acquired or derived in the processes from step S11 to step S15. In other words, the control device 32 derives, as an obstacle point group PCob, those located within the search area A2, among the observation point group PC1 in the observation area A1, the auxiliary point group PC2 in the observation area A1, the virtual observation point group PCi1 outside the observation area A1, and the virtual auxiliary point group PCi2 outside the observation area A1.

[0056] The control device 32 updates the obstacle point cloud PCob of the previous update cycle by storing the observation point cloud PC1 and virtual observation point cloud PCi1 derived as the obstacle point cloud PCob as past observation point clouds, the auxiliary point cloud PC2 and virtual auxiliary point cloud PCi2 as past auxiliary point clouds, and the observation time t1 as past observation time t2 in the auxiliary storage device 35. After completing the process of step S19, the control device 32 terminates the obstacle point cloud derivation process.

[0057] After the obstacle point cloud derivation process in step S1 is completed, the process proceeds to step S2, where the control device 32 executes outer edge processing to derive outer edge points Pe1, Pe2 constituting outer edges E1, E2 from the obstacle point cloud PCob.

[0058] Here, the details of the outer edge processing performed in step S2 will be described. In the following description, it is assumed that the obstacle point cloud PCob has been derived at the stage when the processing in step S1 is completed, as shown in FIG.

[0059] 5, first, in step S21, the control device 32 performs clustering on the obstacle point cloud PCob derived in step S1. As a result, the control device 32 classifies which obstacle position each obstacle point Pob indicates.

[0060] The control device 32 performs clustering of each obstacle point Pob based on the similarity between the obstacle points Pob. The similarity is, for example, the distance between the obstacle points Pob. The distance can be expressed in any form, such as Manhattan distance, Euclidean distance, or cosine distance, as long as it can be compared with the travelable width W1. If the distance between the two obstacle points Pob is less than the travelable width W1, the control device 32 classifies the two obstacle points Pob into the same obstacle point group. On the other hand, if the distance between the two obstacle points Pob is equal to or greater than the travelable width W1, the control device 32 classifies the two obstacle points Pob into different obstacle point groups. Any specific algorithm can be used for the clustering, and examples of the algorithm include the group average method, Ward's method, shortest distance method, and longest distance method.

[0061] As shown in Fig. 6 and Fig. 7, in this embodiment, as a result of the clustering, the obstacle point cloud PCob is classified into two obstacle point clouds PCob1 and PCob2. The first obstacle point cloud PCob1 indicates the position of the first obstacle Ob1. The second obstacle point cloud PCob2 indicates the position of the second obstacle Ob2. Note that the clustering is not limited to two obstacles and can be applied to any number of obstacles.

[0062] The control device 32 may store which of the observation point group PC1, the virtual observation point group PCi1, the auxiliary point group PC2, and the virtual auxiliary point group PCi2 each of the outer edge points Pe1 and Pe2 derived in step S21 originates from. In this case, in the next obstacle point cloud derivation process, the control device 32 treats the outer edge points Pe1 and Pe2 in the same way as the point group from which they originate.

[0063] As shown in Fig. 5, the process proceeds to step S22, where the outer edge points Pe1 and Pe2 constituting the outer edges E1 and E2 are derived from the obstacle points Pob constituting each of the obstacle point groups PCob1 and PCob2. As a result, the outer edges E1 and E2 and the outer edge points Pe1 and Pe2 as shown in Fig. 7 are obtained. The control device 32 that executes the process of step S22 corresponds to the outer edge derivation unit of this embodiment. The outer edge points Pe1 and Pe2 of this embodiment are parts of the obstacle point groups PCob1 and PCob2, respectively.

[0064] The first outer edge E1 defines the outer edge of the area that includes the first obstacle point group PCob1. The outer edge of a certain area includes a point group means that all of the points that constitute the point group are located in the area or on the outer edge of the area. For example, when the first outer edge E1 includes the first obstacle point group PCob1, it includes a case where at least a part of the first obstacle point Pob1 is located on the first outer edge E1. The first outer edge E1 is a polygon with the first outer edge point Pe1 as a vertex. In this embodiment, the first outer edge E1 is a convex hull of the first obstacle point group PCob1. Note that a convex hull is the smallest convex polygon that includes all given points. A convex polygon is a polygon in which all interior angles are 180 degrees or less.

[0065] The second outer edge E2 defines the outer edge of the area that includes the second obstacle point cloud PCob2. When the second outer edge E2 includes the second obstacle point cloud PCob2, it includes a case where at least a part of the second obstacle point Pob2 is located on the second outer edge E2. The second outer edge E2 is a polygon having the second outer edge point Pe2 as a vertex. The second outer edge E2 is a polygon having the second outer edge point Pe2 as a vertex. In this embodiment, the second outer edge E2 is a convex hull of the second obstacle point cloud PCob2.

[0066] Here, an example of a method for deriving the first outer edge point Pe1, in other words, a method for constructing the first outer edge E1 that is a convex hull, will be described. First, the control device 32 selects the first obstacle point Pob1 that has the smallest Y coordinate from among the first obstacle points Pob1 that constitute the first obstacle point group PCob1. Next, the control device 32 selects the first obstacle point Pob1 that has the smallest deflection angle with respect to the obstacle point Pob1. The control device 32 may sequentially select the first obstacle point Pob1 that has the smallest deflection angle with respect to the positive direction of the X-axis with the selected first obstacle point Pob1 as the origin. The obstacle point Pob1 selected in this manner constitutes the first outer edge point Pe1. The control device 32 stores the first outer edge E1 as a point sequence of the first outer edge points Pe1 arranged in the order of selection. This allows the control device 32 to grasp the first outer edge points Pe1 that are adjacent to each other on the first outer edge E1 based on the order of the point sequence. The control device 32 can obtain the second outer edge E2 and the second outer edge point Pe2 from the second obstacle point cloud PCob2 by applying the above method to the second obstacle point cloud PCob2.

[0067] In addition, the specific method of deriving the outer edge points Pe1 and Pe2, in other words, the method of constructing the convex hull, is not limited to the above and can be any method, such as the sequential construction method, the divide and conquer method, the Quickhull method, the Graham scan, the gift wrapping method, etc.

[0068] In this embodiment, a part of the outer edge points Pe1 and Pe2 is configured by an auxiliary point P2 generated on the boundary between the masked area A11 and the observable area A12. As a result, the boundary between the masked area A11 and the observable area A12 is included in the outer edges E1 and E2. Therefore, the outer edges E1 and E2 can be used to more accurately represent an area where an obstacle may exist.

[0069] 5, the process proceeds to step S23, where the control device 32 removes the obstacle point cloud PCob included inside the outer edges E1, E2. The inside of the outer edges E1, E2 is the area surrounded by the outer edges E1, E2, and does not include the area that coincides with the outer edges E1, E2. In other words, the inside of the outer edges E1, E2 is an open set surrounded by the outer edges E1, E2.

[0070] As shown in Fig. 8, by executing the process of step S23, obstacle points Pob other than the outer edge points Pe1 and Pe2 are removed. In other words, only the outer edge points Pe1 and Pe2 on the outer edges E1 and E2 remain. The first obstacle point cloud PCob1 is composed of the first outer edge points Pe1, and the second obstacle point cloud PCob2 is composed of the second outer edge points Pe2.

[0071] As shown in FIG. 5, the process proceeds to step S24, where the control device 32 determines whether or not there is an area on the first outer edge E1 where the distance between adjacent first outer edge points Pe1 is equal to or greater than the travelable width W1. In this embodiment, the determination is made based on whether or not the ratio of the distance between the first outer edge points Pe1 to the travelable width W1 is equal to or greater than 1. Note that the determination method is not limited to this and may be any method, and may be performed using the difference between the travelable width W1 and the first outer edge point Pe1. The control device 32 also makes a similar determination for the second outer edge point Pe2.

[0072] If the determination result in step S24 is negative, that is, if the intervals between all of the first outer edge points Pe1 and the intervals between all of the second outer edge points Pe2 are less than the travelable width W1, outer edge processing ends.

[0073] On the other hand, if the judgment result in step S24 is positive, that is, if there is an area on the outer edges E1, E2 where the distance between the first outer edge points Pe1 or the distance between the second outer edge points Pe2 is greater than or equal to the drivable width W1, the control device 32 proceeds to step S25.

[0074] In step S25, the control device 32 generates the outer edge points Pe1 and Pe2 in the area on the outer edges E1 and E2 where the interval between the outer edge points Pe1 and Pe2 on the outer edges E1 and E2 is equal to or greater than the travelable width W1. As a result, as shown in FIG. 9, the interval between the first outer edge points Pe1 and the interval between the second outer edge points Pe2 on the area are less than the travelable width W1. Hereinafter, for convenience of explanation, the outer edge points Pe1 and Pe2 generated in step S25 may be referred to as additional outer edge points Pe1n and Pe2n. Note that the outer edge points Pe1 and Pe2 generated in step S25 may be different from those derived in step S22. In other words, the outer edge points Pe1 and Pe2 generated in step S25 may not be included in the obstacle points Pob1 and Pob2 derived in step S1. The control device 32 that executes the process of step S25 corresponds to the outer edge point generating unit of this embodiment.

[0075] Here, an example of a method for generating the additional outer perimeter points Pe1n and Pe2n in step S25 will be described. For convenience of explanation, only the first additional outer perimeter point Pe1n will be described below, but the method shown below can also be applied to the second additional outer perimeter point Pe2n in the same manner.

[0076] First, the control device 32 derives the ratio of the interval between the first outer edge points Pe1 to the travelable width W1. The ratio derived in the judgment of step S24 may be used as the ratio. Since the judgment result in step S24 is positive, the ratio takes a value of 1 or more. Next, the value after the decimal point of the ratio is rounded down. As a result, the ratio takes a value of a natural number of 1 or more. The ratio expressed as a natural number indicates the number of first outer edge points Pe1 that need to be newly generated on the area of ​​the outer edges E1 and E2. Hereinafter, for convenience of explanation, the ratio expressed as a natural number is referred to as the generation number N. The control device 32 generates the first additional outer edge points Pe1n at equal intervals between the first outer edge points Pe1 located at both ends of the area. The position of the first additional outer edge points Pe1n is arbitrary as long as the interval between the adjacent first outer edge points Pe1 and first additional outer edge points Pe1n in the area is less than the travelable width W1.

[0077] After executing the process of step S25, the control device 32 ends the outer edge process and proceeds to step S3. As shown in FIG. 3, in step S3, the control device 32 determines whether or not obstacles Ob1, Ob2 are present within the search area A2 based on the outer edge points Pe1, Pe2 derived in step S2. The outer edge points Pe1, Pe2 derived in step S2 include the outer edge points Pe1, Pe2 derived in step S22 and the outer edge points Pe1, Pe2 generated in step S25, i.e., the additional outer edge points Pe1n, Pe2n. Any method of determination may be used, but for example, the control device 32 determines whether or not the outer edge points Pe1, Pe2 are included within the search area A2. If the determination result in step S3 is negative, that is, if it is determined that obstacles Ob1, Ob2 are not present within the search area A2, the process proceeds to step S4, where the control device 32 causes the autonomous vehicle 10 to continue normal traveling and ends the obstacle avoidance process. On the other hand, if the determination result in step S3 is positive, that is, if it is determined that obstacles Ob1, Ob2 are present within the search area A2, the process proceeds to step S5.

[0078] In step S5, the control device 32 causes the autonomous vehicle 10 to perform avoidance travel, thereby causing the autonomous vehicle 10 to avoid the obstacles Ob1 and Ob2 indicated by the obstacle point cloud PCob. Avoidance travel is travel in a direction in which the obstacles Ob1 and Ob2 determined to exist in step S3 can be avoided. As an example of avoidance travel, the control device 32 determines the travel direction based on the positions of the outer edges E1 and E2. The control device 32 searches whether there is an area in the search area A2 where the distance between the outer edge points Pe1 and Pe2 is equal to or greater than the travelable width W1. The control device 32 determines, as the travel direction of the autonomous vehicle 10, the direction in which the angle between the direction from the vehicle center position P toward the area and the direction of the travel route R is the smallest, among the areas equal to or greater than the travelable width W1. The control device 32 then causes the autonomous vehicle 10 to travel in the travel direction. In other words, the control device 32 causes the autonomous vehicle 10 to avoid the obstacles Ob1, Ob2 by making the autonomous vehicle 10 detour around the area where the distance between the outer edge points Pe1, Pe2 is less than the travelable width W1. The control device 32 that executes the process of step S5 corresponds to the avoidance unit of this embodiment.

[0079] After a predetermined update period has elapsed since the process of step S5 is executed, the control device 32 returns to step S1 and performs the obstacle point cloud derivation process again. As a result, the control device 32 derives the obstacle point cloud PCob every time an update period elapses, and derives the outer edge points Pe1 and Pe2 based on the derived obstacle point cloud PCob.

[0080] If it is determined that the obstacles Ob1, Ob2 are not present within the search area A2 as a result of the avoidance driving, the determination result in step S3 is negative, and the control device 32 causes the autonomous mobile body 10 to perform normal driving.

[0081] <effect> The operation of this embodiment will now be described. By executing the process of step S22, the distance between the outer edge points Pe1, Pe2 on the outer edges E1, E2 may become equal to or greater than the travelable width W1. Therefore, there is a risk that the obstacles Ob1, Ob2 are present in an area where the distance between the outer edge points Pe1, Pe2 is equal to or greater than the travelable width W1. If the determination of step S3 is made based on such outer edge points Pe1, Pe2, there is a risk that the control device 32 causes the autonomous mobile body 10 to travel toward the obstacles Ob1, Ob2.

[0082] Therefore, by performing the process of step S24, the distance between the first outer edge points Pe1 on the first outer edge E1 and the distance between the second outer edge points Pe2 on the second outer edge E2 are less than the travelable width W1 in the entire area of ​​the outer edges E1, E2. Therefore, the area in which the obstacles Ob1, Ob2 may exist is surrounded by the outer edges E1, E2 composed of the outer edge points Pe1, Pe2 that are spaced apart by less than the travelable width W1.

[0083] <Effects> The effects of this embodiment will be described below. (1) The control device 32 executes the process of step S1 to derive an obstacle point cloud PCob composed of multiple obstacle points Pob indicating the positions of obstacles Ob1, Ob2 relative to the autonomous vehicle 10 based on the observation results of the sensor 31. Next, the control device 32 executes the process of step S22 to derive, as outer edge points Pe1, Pe2, obstacle points Pob located on outer edges E1, E2 that contain the obstacle point cloud PCob derived in step S1, from the multiple obstacle points Pob derived in step S1. The distance between adjacent outer edge points Pe1, Pe2 on the outer edges E1, E2 is less than the travelable width W1 of the autonomous vehicle 10.

[0084] According to this, the outer edges E1, E2 that encompass the obstacle point groups PCob1, PCob2 encompass the entire obstacles Ob1, Ob2 indicated by the obstacle point groups PCob. The outer edges E1, E2 indicate the boundaries of the areas in which the obstacles Ob1, Ob2 exist. The control device 32 causes the autonomous vehicle 10 to avoid the areas indicated by the outer edge points Pe1, Pe2 that are located on such outer edges E1, E2. This allows the autonomous vehicle 10 to avoid the obstacles Ob1, Ob2 based on the outer edge points Pe1, Pe2.

[0085] At this time, for example, the obstacle point cloud PCob present inside the outer edges E1, E2 is not taken into consideration when the autonomous mobile unit 10 avoids the obstacles Ob1, Ob2. This reduces the information on the obstacle points Pob located inside the outer edges E1, E2 when the autonomous mobile unit 10 avoids the obstacles Ob1, Ob2. This reduces the load on the control device 32 when the autonomous mobile unit 10 avoids the obstacles Ob1, Ob2.

[0086] (2) If the judgment result of step S24 is positive, i.e., if the distance on the outer edges E1, E2 between adjacent outer edges E1, E2 is less than the drivable width W1 of the autonomous vehicle 10, the control device 32 executes the processing of step S25 to generate new additional outer edge points Pe1n, Pe2n on the outer edges E1, E2 between the adjacent outer edges E1, E2, thereby making the distance on the outer edges E1, E2 between adjacent outer edges E1, E2 less than the drivable width W1.

[0087] According to this, even if the distance on the outer edges E1, E2 between the outer edge points Pe1, Pe2 derived from the obstacle point groups PCob1, PCob2 is equal to or greater than the travelable width W1, the outer edge points Pe1, Pe2 are generated on the outer edges E1, E2, so that the distance becomes less than the travelable width W1. Therefore, it is possible to prevent the autonomous mobile body 10 from avoiding an area where the distance between the outer edge points Pe1, Pe2 is equal to or greater than the travelable width W1, despite the presence of the obstacles Ob1, Ob2.

[0088] (3) The outer edges E1 and E2 are polygons having the outer edge points Pe1 and Pe2 as vertices derived by executing the process of step S22. According to this, the outer edges E1, E2 are straight lines connecting the adjacent outer edge points Pe1, Pe2. Therefore, the outer edges E1, E2 located between the two adjacent outer edge points Pe1, Pe2 can be expressed as a line segment of the adjacent outer edge points Pe1, Pe2. Therefore, the degree of freedom of the outer edges E1, E2 is smaller than when the outer edges E1, E2 are expressed as a curve connecting the adjacent outer edge points Pe1, Pe2. Therefore, the control device 32 can easily determine the positions on the outer edges E1, E2 to generate new outer edge points Pe1, Pe2, i.e., additional outer edge points Pe1n, Pe2n. Therefore, the calculation load on the control device 32 when generating the additional outer edge points Pe1n, Pe2n can be reduced.

[0089] (4) The outer edges E1 and E2 are convex hulls having the outer edge points Pe1 and Pe2 derived in step S22 as vertices. This makes it possible to prevent the obstacles Ob1 and Ob2 from becoming excessively large when represented by the outer edge points Pe1 and Pe2, compared to when the obstacles Ob1 and Ob2 are represented by the obstacle point cloud PCob. Therefore, it is possible to prevent the route when the autonomous mobile unit 10 avoids the obstacles Ob1 and Ob2 from becoming redundant.

[0090] (5) The control device 32 executes the process of step S11 to acquire an observation point cloud PC1 indicating the positions of obstacles Ob1, Ob2 in the observation area A1 of the sensor 31 based on the observation results of the sensor 31. Next, the control device 32 executes the process of step S12 to generate an auxiliary point cloud PC2 in the obstructed area A11 blocked by the obstacles Ob1, Ob2 within the observation area A1, based on the observation point cloud PC1 acquired in step S11. The obstacle point cloud PCob includes the observation point cloud PC1 and the auxiliary point cloud PC2.

[0091] According to this, obstacles observed by the sensor 31 are represented as an observation point group PC1, and obstacles that may exist in the shielded area A11 are represented as an auxiliary point group PC2. The control device 32 derives and generates outer edge points Pe1, Pe2 based on the obstacle point group PCob including the observation point group PC1 and the auxiliary point group PC2, and causes the autonomous mobile device 10 to avoid the obstacles Ob1, Ob2 represented by the outer edge points Pe1, Pe2. Therefore, it is possible to cause the autonomous mobile device 10 to avoid the obstacles Ob1, Ob2, taking into consideration the possibility that there may be obstacles in an area that the sensor 31 could not observe due to the obstacles Ob1, Ob2.

[0092] (6) If the observation point group PC1 acquired earlier than the observation point group PC1 acquired in step S11 is taken as the past observation point group, the control device 32 executes the process of step S14 to acquire the movement amount δ of the autonomous mobile device 10 from the past observation time t2 when the past observation point group was acquired to the observation time t1 when the observation point group PC1 is acquired in step S11. Next, the control device 32 executes the process of step S15 to move the past observation point group by the movement amount δ, thereby deriving the virtual observation point group PCi1. Then, the obstacle point group PCob includes the virtual observation point group PCi1.

[0093] According to this, the virtual observation point cloud PCi1 is obtained by moving the previously observed observation point cloud PC1 by the movement amount δ of the autonomous mobile unit 10. The virtual observation point cloud PCi1 indicates where the obstacles Ob1, Ob2 indicated by the previously observed observation point cloud PC1 are located at the time when the observation point cloud PC1 is acquired in step S11. Therefore, it is possible to cause the autonomous mobile unit 10 to avoid the obstacles Ob1, Ob2 while taking into consideration the possibility that the previously observed obstacles Ob1, Ob2 are located outside the observation area A1.

[0094] (7) In the obstacle point cloud derivation process, the control device 32 executes the process of step S15 of deriving a virtual auxiliary point cloud PCi2 located outside the observation area A1 by moving the past auxiliary point cloud, which is the auxiliary point cloud PC2 generated in the past, by the movement amount δ based on the observation point cloud PC1 acquired in step S11. Then, the obstacle point cloud PCob includes the virtual auxiliary point cloud PCi2.

[0095] When the observation of the sensor 31 is blocked by the obstacles Ob1 and Ob2, it becomes difficult for the sensor 31 to observe the area behind the obstacles Ob1 and Ob2 with respect to the autonomous vehicle 10. Therefore, according to the present configuration, an auxiliary point cloud PC2 is generated in such a blocked area A11. When the previously generated auxiliary point cloud PC2 moves outside the observation area A1 of the sensor 31 due to the movement of the autonomous vehicle 10, the auxiliary point cloud PC2 is moved by the movement amount δ of the autonomous vehicle 10, and a virtual auxiliary point cloud PCi2 is derived. The virtual auxiliary point cloud PCi2 derived in this manner indicates an area that was in the observation area A1 of the sensor 31 but could not be observed by the sensor 31 due to the obstacles Ob1 and Ob2. Since the obstacle point cloud PCob includes the virtual auxiliary point cloud PCi2, the autonomous vehicle 10 can avoid the area represented by the virtual auxiliary point cloud PCi2 when avoiding an obstacle. It is possible to avoid the obstacles Ob1 and Ob2 while avoiding the area that the sensor 31 could not observe due to the obstacles Ob1 and Ob2. Therefore, it is possible to reduce the possibility that the autonomous mobile body 10 will come into contact with an object in the area blocked by the obstacles Ob1 and Ob2.

[0096] <Modification> The embodiment can be modified as follows: The embodiment and the following modified examples can be combined with each other to the extent that there is no technical contradiction.

[0097] The outer edges E1 and E2 change as the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 outside the search area A2 are deleted in step S18. In this case, there is a risk that the area where the obstacles Ob1 and Ob2 may exist may be located outside the outer edges E1 and E2. Therefore, when at least a part of the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 derived in step S15 is located in the search area A2, the control device 32 may represent the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 located outside the search area A2 as points on the search area A2. For example, the control device 32 may derive the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 on the search area A2 by complementing the movement paths of the virtual observation point group PCi1 and the virtual auxiliary point group PCi2 based on the past observation point group, the past auxiliary point group, and the movement amount δ.

[0098] This reduces the change in the outer edges E1, E2 before and after the process of step S18, allowing the outer edges E1, E2 to more accurately encompass the area in which the obstacles Ob1, Ob2 may exist, thereby further improving the obstacle avoidance performance of the autonomous mobile unit 10.

[0099] The search area A2 can be set arbitrarily, and the shape of the search area A2 can be any closed area, such as a polygon such as a rectangle or a trapezoid, a circle, a semicircle, or the like. The control device 32 may cause the autonomous vehicle 10 to perform avoidance driving even when an obstacle point cloud PCob exists outside the search area A2. Accordingly, the control device 32 may omit the process of step S18 for removing each point cloud PC1, PCi1, PC2, and PCi2 outside the search area A2. In this case, the search area A2 does not need to be set.

[0100] When deriving the virtual observation point cloud PCi1 in step S15, the control device 32 may use all of the past observation point cloud, not just a part of it. Also, the control device 32 may use any interval between the virtual observation points Pi1 in the virtual observation point cloud PCi1 derived in step S15.

[0101] The control device 32 may omit the process of step S15 of deleting the auxiliary point group PC2 and the virtual auxiliary point group PCi2 in the observable area A12. The control device 32 may omit the process of step S13 for generating the auxiliary point group PC2 within the masked area A11.

[0102] The control device 32 may derive the outer edge points Pe1 and Pe2 in any manner as long as the distance between the outer edge points Pe1 and Pe2 on the outer edge E1 and E2 is less than the travelable width W1 by executing the process of step S22. For example, the control device 32 may omit the process of step S25. For example, if the distance between the outer edge points Pe1 and Pe2 on the outer edge E1 and E2 is equal to or greater than the travelable width W1 as a result of the process of step S22, the control device 32 may change the outer edge E1 and E2 so that the obstacle points Pob1 and Pob2 within a predetermined distance from the outer edge E1 and E2 become the outer edge points Pe1 and Pe2. The control device 32 may repeat the above process until the distance between the outer edge points Pe1 and Pe2 on the outer edge E1 and E2 becomes less than the travelable width W1.

[0103] The obstacle point cloud PCob may have any specific form and may not include the auxiliary point cloud PC2, the virtual observation point cloud PCi1, or the virtual auxiliary point cloud PCi2. It is preferable that the obstacle point cloud PCob includes the observation point cloud PC1, but is not limited to this.

[0104] The shape of the outer edges E1, E2 does not have to be a polygon such as a convex hull, and any shape including at least one of straight lines and curved lines can be adopted. For example, the outer edges E1, E2 may be a circle that contains the obstacle point groups PCob1, PCob2, specifically, the minimum containing circle of the obstacle point groups PCob1, PCob2.

[0105] The specific form of the sensor 31 is arbitrary as long as it can measure the relative coordinates of the obstacle Ob1 with respect to the autonomous vehicle 10. For example, a plurality of ultrasonic sensors may be used as the sensor 31. The ultrasonic sensors can measure distance by emitting ultrasonic waves. The ultrasonic sensors are mounted on the vehicle body 21 at intervals. The control device 32 can obtain the distance and angle of the obstacle Ob1, i.e., the relative coordinates of the obstacle Ob1, based on the intensity of the reflected waves observed by each ultrasonic sensor. As another example, the sensor 31 may be one capable of observing the distance and angle to an object, such as a LIDAR (Laser Imaging Detection and Ranging), a laser range finder, a ToF (Time of Flight) camera, or the like.

[0106] The method of acquiring the observation point cloud PC1 in step S11 is arbitrary. For example, the control device 32 may acquire the observation point cloud by converting the point cloud obtained by the sensor 31 using a LIDAR into an appropriate coordinate system.

[0107] The vehicle center position P and the sensor position Ps may coincide with each other. The autonomous vehicle 10 may be implemented in any desired manner. For example, any wheels may be used as the wheels 22, not limited to omnidirectional wheels. The number of wheels 22 is not limited to four, and may be any number. [Explanation of symbols]

[0108] 10...autonomous vehicle, 31...sensor, 32...control device, A1...observation area, A11...obstructed area, E1, E2...outer edge, Ob1, Ob2...obstacle, PC1...observation point cloud, PC2...auxiliary point cloud, PCi1...virtual observation point cloud, PCi2...virtual auxiliary point cloud, PCob, PCob1, PCob2...obstacle point cloud, Pe1, Pe2...outer edge points, Pe1n, Pe2n...additional outer edge points, W1...traveling width, δ...movement amount.

Claims

1. An obstacle avoidance device for causing an autonomous vehicle to avoid an obstacle observed by a sensor provided in the autonomous vehicle, an obstacle point cloud derivation unit that derives an obstacle point cloud configured from a plurality of obstacle points that indicate positions of the obstacles relative to the autonomous vehicle based on the observation results of the sensor; an outer edge derivation unit that derives, as outer edge points, obstacle points that are located on the outer edge of a region that is surrounded by an outer edge that passes through at least some of the obstacle points that constitute the derived obstacle point cloud, and that is defined such that all of the obstacle points that constitute the obstacle point cloud are located in the region or on the outer edge; an avoidance unit that causes the autonomous vehicle to avoid the obstacle indicated by the outer edge point; An obstacle avoidance device, wherein the distance on the outer edge between adjacent outer edge points is less than the travelling width of the autonomous vehicle.

2. 2. The obstacle avoidance device of claim 1, further comprising an outer edge point generation unit that generates outer edge points on the outer edge between adjacent outer edge points so that the distance between adjacent outer edge points on the outer edge is less than the drivable width when the distance between adjacent outer edge points on the outer edge is greater than or equal to the drivable width of the autonomous vehicle.

3. 3. The obstacle avoidance device according to claim 1, wherein the outer edge is a polygon having vertices at least some of the outer edge points derived by the outer edge derivation unit.

4. The obstacle avoidance device according to claim 3 , wherein the outer edge is a convex hull having the outer edge points derived by the outer edge derivation unit as vertices.

5. The obstacle point cloud derivation unit is an observation point cloud acquisition unit that acquires an observation point cloud indicating the position of the obstacle within an observation area of ​​the sensor based on the observation result of the sensor; an auxiliary point cloud generation unit that generates an auxiliary point cloud within a blocked area that is blocked by the obstacle, of the observation area, based on the observation point cloud acquired by the observation point cloud acquisition unit; 5. The obstacle avoidance device according to claim 1, wherein the obstacle point cloud includes the observation point cloud and the auxiliary point cloud.

6. the obstacle point cloud includes a virtual observation point cloud indicating the positions of the obstacles that were observed by the sensor in the past and that have come to be located outside the observation area of ​​the sensor due to the movement of the autonomous vehicle; The obstacle point cloud derivation unit is a movement amount acquisition unit that acquires a movement amount of the autonomous vehicle from the acquisition of the past observation point group to the acquisition of the observation point group by the observation point group acquisition unit, where the observation point group acquired earlier than the observation point group acquired by the observation point group acquisition unit is a past observation point group; 6. The obstacle avoidance device according to claim 5, further comprising: a virtual observation point cloud derivation unit that derives the virtual observation point cloud by moving the past observation point cloud by the movement amount.

7. The obstacle point cloud derivation unit is a virtual auxiliary point cloud derivation unit that derives a virtual auxiliary point cloud located outside an observation area of ​​the sensor by moving the previously generated auxiliary point cloud by the movement amount, The obstacle avoidance apparatus of claim 6 , wherein the obstacle point cloud includes the virtual auxiliary point cloud.

Citation Information

Patent Citations

  • Movable region extraction device and movable region extraction method

    JP2011253377A

  • Robot system and map update method

    JP2014089740A

  • Mounting determination method, mounting method, mounting determination device, and robot

    JP2015123534A

  • Display processing device

    JP2019087980A

  • Autonomous travel body

    JP2020140477A