Autonomous control system and autonomous control method

Abstract

To provide an autonomous control system for enabling a control object to be efficiently and safely operated even when it is difficult to predict a behavior of a non-control object being a mobile body in the periphery of the control object.SOLUTION: An autonomous control system A100 includes: a state amount acquisition section A002 for acquiring information of a control object 001; a control object position specification section A003; a control object arrival area calculation section A007; an outer information acquisition section A001; an object identification section A004; a non-control object position specification section A005; a non-control object arrival area calculation section A006; a crossover area calculation section A008; a control parameter setting section A009; and a manipulated variable operation section A010 for calculating a control input so as to allow a value of an evaluation function J to be smaller than a previous value.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an autonomous control system and an autonomous control method relating to a safety support function in an autonomous or semi-autonomous vehicle. [Background technology]

[0002] In recent years, the development of autonomous driving technology has progressed in order to reduce traffic accidents and traffic congestion.

[0003] Autonomous driving technology is also attracting attention in the logistics industry, which is facing a serious labor shortage. Autonomous driving technology can be deployed in a wide range of applications, including not only trucks that collect and deliver goods on public roads, but also forklifts used to collect and store goods in factories and warehouses, transport robots known as AGVs (Automatic Guided Vehicles) and AMRs (Autonomous Mobile Robots), and transport vehicles between processes.

[0004] Except for fully automated large-scale logistics warehouses, autonomous driving technology is used in environments where people (mainly pedestrians) and non-automated vehicles (e.g., forklifts operated by workers) coexist. For this reason, autonomous vehicles must have safety features to avoid contact with pedestrians and non-automated vehicles.

[0005] Generally, to achieve such safety functions, the trajectory along which pedestrians or non-automated vehicles will move is predicted, and the automated vehicle is controlled to prevent the vehicle from coming into contact with the predicted trajectory.

[0006] The motion of non-automated vehicles is mainly governed by nonholonomic constraints, so they cannot move sideways or make sudden changes in direction, whereas pedestrians can move freely in various directions, making it difficult to calculate their predicted trajectory.

[0007] To address this issue, Patent Document 1 discloses that a driving assistance device starts decelerating a vehicle at a first deceleration at a first time point. If a crossing target still exists at a second time point immediately before a time point at which the vehicle cannot be stopped at a position immediately before the intersection area even if the vehicle starts decelerating at a second deceleration rate greater than the first deceleration, the driving assistance device starts decelerating the vehicle at the second deceleration at the second time point. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Patent Publication No. 2021-187207 Summary of the Invention [Problem to be solved by the invention]

[0009] In Patent Document 1, the collision condition is considered to be the intersection area between the area in which the vehicle could exist if it maintains its current speed (vehicle speed) and the area in which the obstacle could exist if it maintains its current target speed (candidate target speed) and heading.

[0010] For this reason, it may not be possible to provide ideal driving assistance when faced with obstacles that may behave differently from what is expected, such as a child who suddenly starts running or a drunk person who is unable to decide which direction to move.

[0011] The present invention has been devised to solve the above-mentioned problems, and an object of the present invention is to provide an autonomous control system and an autonomous control method that are capable of efficiently and safely driving a controlled object, such as an autonomous vehicle, even when it is difficult to predict the behavior of a non-controlled object, which is a moving body surrounding the controlled object. [Means for solving the problem]

[0012] In order to achieve the above object, the present invention is configured as follows.

[0013] The autonomous control system includes a state quantity acquisition unit that acquires a state quantity of a controlled object, a control object position identification unit that identifies a position of the controlled object based on the state quantity, a control object reachable area calculation unit that calculates a control object reachable area that the controlled object can reach within a predetermined time based on the position of the controlled object, an external information acquisition unit that acquires external information of the controlled object, a target identification unit that identifies an attribute of a non-controlled object based on the external information, a non-controlled object position identification unit that calculates the position of the non-controlled object based on the external information, a non-controlled object reachable area calculation unit that calculates a non-controlled object area that the non-controlled object can reach within a predetermined time using the attribute and the position, and an intersection area calculation unit that calculates an intersection area of ​​each set of positions using the set of positions of the object and the set of positions of the non-controlled objects calculated by the non-controlled object reachable area calculation unit; a control parameter setting unit that sets a control parameter of the controlled object such that the larger the intersection amount area of ​​the intersection amount area of ​​each set of positions, the greater the behavior of the controlled object in which the distance between the controlled object and the non-controlled object becomes; and a manipulated variable calculation unit that calculates a control input using the control parameters and an evaluation function whose value becomes smaller as the controlled object approaches a desired behavior, so that the value of the evaluation function becomes smaller than a previous value, within a range that satisfies a predetermined constraint condition.

[0014] Further, in an autonomous control method, a state quantity of a controlled object is acquired, a position of the controlled object is identified based on the state quantity, a controlled object reachable area that the controlled object can reach within a predetermined time based on the position of the controlled object is calculated, external information of the controlled object is acquired, attributes of non-controlled objects are identified based on the external information, a position of the non-controlled object is calculated based on the external information, a non-controlled object area that the non-controlled object can reach within a predetermined time using the attributes and the position is calculated, an intersection area of ​​each set of positions is calculated using the set of positions of the controlled object and the set of positions of the non-controlled objects, a control parameter of the controlled object is set such that the larger the intersection area of ​​the intersection area of ​​each set of positions, the greater the behavior of the controlled object in which the distance between the controlled object and the non-controlled object becomes, and a control input is calculated using the control parameters and an evaluation function whose value becomes smaller as the controlled object approaches a desired behavior, so that the value of the evaluation function becomes smaller than the previous value, within a range that satisfies a predetermined constraint condition. Effect of the Invention

[0015] It is possible to provide an autonomous control system and an autonomous control method that can drive a controlled object, such as an autonomous vehicle, efficiently and safely, even when it is difficult to predict the behavior of a non-controlled object, which is a moving body surrounding the controlled object. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 is a functional block diagram of an autonomous control system according to a first embodiment of the present invention. [Figure 2A] FIG. 1 is a diagram illustrating a first embodiment in which the present invention is applied to a vehicle. [Figure 2B] FIG. 1 is a diagram illustrating a first embodiment in which the present invention is applied to a vehicle. [Diagram 3] FIG. 13 is a functional block diagram of a non-control object reachable region prediction unit. [Figure 4A] FIG. 11 is a diagram illustrating an example of processing contents of a trajectory prediction unit. [Figure 4B] FIG. 11 is a diagram illustrating an example of processing contents of a trajectory prediction unit. [Figure 5A] FIG. 11 is a diagram illustrating an example of a process performed by a reachable region calculation unit; [Figure 5B] 13 is a diagram illustrating another example of the processing content of the reachable area calculation unit. FIG. [Figure 5C] 13 is a diagram illustrating yet another example of the processing content of the reachable area calculation unit. FIG. [Figure 6A] FIG. 11 is a diagram illustrating an example of a process performed by a control target reaching region calculation unit. [Figure 6B] FIG. 11 is a diagram illustrating an example of a process performed by a control target reaching region calculation unit. [Figure 7A] 11A and 11B are diagrams illustrating the processing content of an intersection area calculation unit. [Figure 7B] 11A and 11B are diagrams illustrating the processing content of an intersection area calculation unit. [Figure 7C] 11A and 11B are diagrams illustrating the processing content of an intersection area calculation unit. [Figure 7D] 11A and 11B are diagrams illustrating the processing content of an intersection area calculation unit. [Figure 7E] 11A and 11B are diagrams illustrating the processing content of an intersection area calculation unit. [Figure 8A] 5A and 5B are diagrams illustrating processing contents of a control parameter setting unit. [Figure 8B] 5A and 5B are diagrams illustrating processing contents of a control parameter setting unit. [Figure 8C] 5A and 5B are diagrams illustrating processing contents of a control parameter setting unit. [Figure 9] FIG. 2 is a diagram illustrating parameters of a vehicle to be controlled. [Figure 10] FIG. 2 is a diagram illustrating the positional relationship between a vehicle and a pedestrian. [Figure 11] This is an example of a penalty related to the position of vehicles and pedestrians. [Figure 12] 1A and 1B are diagrams illustrating a situation in which a vehicle takes unnecessary avoidance action. [Figure 13A] 1A and 1B are diagrams illustrating differences in pedestrian characteristics and a parameter setting method utilizing the differences. [Figure 13B] 1A and 1B are diagrams illustrating differences in pedestrian characteristics and a parameter setting method utilizing the differences. [Figure 14A] FIG. 13 is a diagram for explaining an innovation regarding a parameter setting method. [Figure 14B] FIG. 13 is a diagram for explaining an innovation regarding a parameter setting method. [Figure 15] 2 is an example of a flowchart of the autonomous control system of the present invention. [Figure 16A] FIG. 11 is a diagram for explaining a second embodiment in which the present invention is applied to a hydraulic excavator. [Figure 16B] FIG. 11 is a diagram for explaining a second embodiment in which the present invention is applied to a hydraulic excavator. [Figure 17] FIG. 11 is a diagram for explaining a second embodiment in which the present invention is applied to a hydraulic excavator. [Figure 18] FIG. 2 is a diagram illustrating components of a hydraulic excavator. [Figure 19] FIG. 2 is a diagram illustrating the positional relationship between a bucket of a hydraulic excavator and a worker. [Figure 20] FIG. 13 is a diagram for explaining a parameter setting method utilizing the proficiency of an operator. [Figure 21] FIG. 11 is a diagram for explaining a third embodiment in which the present invention is applied to a flying object. [Figure 22] 1 is a diagram illustrating the positional relationship between a flying object to be controlled and a flying object to be uncontrolled; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] An autonomous control system is a system that is mounted on a moving object to be controlled, collects information about the surroundings of the moving object, and controls the motion of the moving object to avoid contact with obstacles including moving objects other than the moving object to be controlled. The moving object to be controlled is a vehicle or a robot. The moving object other than the moving object to be controlled is a pedestrian, another vehicle, or another robot.

[0018] In addition, the moving object to be controlled is not limited to a fully autonomous vehicle, but may be a semi-autonomous vehicle that is normally driven by a driver and can intervene only in an emergency.

[0019] Furthermore, the moving object to be controlled may be a vehicle that runs on a general public road, or a vehicle (forklift) or robot that runs within a logistics warehouse.

[0020] Hereinafter, an embodiment of an autonomous control system of the present invention will be described with reference to the drawings. EXAMPLES

[0021] Example 1 FIG. 1 is a simplified functional block diagram of an example of components of an autonomous control system A100 according to a first embodiment of the present invention.

[0022] The autonomous control system A100 may be provided with controls other than the basic functions shown in FIG.

[0023] In the following, we consider an example in which the present invention is applied to a situation in which a vehicle 001, which is the controlled object traveling on a roadway 007, is the controlled object, and a pedestrian 002 walking on a sidewalk 009 is the non-controlled object, as shown in Figures 2A and 2B.

[0024] For simplicity of explanation, a situation with one controlled object and one uncontrolled object is shown, but the present invention can also be used in cases where there are multiple controlled objects and multiple uncontrolled objects.

[0025] The autonomous control system A100 receives as input the detection results of an external information acquirer A001 that acquires environmental information that is the state around the vehicle 001 to be controlled, and a state quantity acquirer A002 that acquires state quantities that are information that indicate the internal state of the vehicle 001 to be controlled. The state quantities of the vehicle 001 to be controlled include the position, direction, speed, and the like.

[0026] It should be noted that not all functions of the autonomous control system A100 need to be implemented in the vehicle 001. As shown in the example of a hydraulic excavator described later, if the area in which the controlled object moves is limited, it is also possible to provide the calculation function in a server that can communicate within that area.

[0027] The external information acquisition unit A001 is a collective term for sensors that acquire information about the outside of a moving object to be controlled. In other words, it refers to external recognition sensors such as LiDAR (Light Detection And Ranging), a stereo camera, and a millimeter wave radar mounted on the vehicle 001 to be controlled.

[0028] In FIG. 2A, the external information acquisition unit A001 is provided only at the front of the vehicle 001. However, a plurality of external information acquisition units A001 may be provided on the vehicle 001 so as to monitor the sides and rear of the vehicle 001.

[0029] In addition, the environment recognizer A001 does not need to be mounted on the vehicle itself as long as it can collect information about the surroundings of the vehicle 100 to be controlled. For example, as shown in Fig. 2B, if a recognition sensor (infrastructure sensor) 003 is installed at a location physically separated from the vehicle 001 to be controlled, and the sensor information of a pedestrian 002 acquired by the infrastructure sensor 003 can be provided to the vehicle 001 via a wireless system 004, the infrastructure sensor 003 also corresponds to the external information acquisition unit A001 of the present invention.

[0030] The state quantity acquisition unit A002 is a general term for sensors that acquire the state of the moving object itself to be controlled, and acquires state quantities that are internal information of the controlled object. The state quantity is a general term for information related to the movement of the controlled object, such as position, direction (orientation), speed, angular velocity, acceleration, and angular acceleration. In other words, the state quantity acquisition unit A002 corresponds to a Global Navigation Satellite System (GNSS) for acquiring the position information of the host vehicle, an Inertial Measurement Unit (IMU) for acquiring the acceleration and angular velocity of the host vehicle, a vehicle speed sensor, an encoder, and the like. Furthermore, it should be noted that the LiDAR mounted on the host vehicle can also be used for Simultaneous Localization and Mapping (SLAM), a technology for calculating the position and direction of the vehicle, and is therefore also a sensor that corresponds to the state quantity acquisition unit A002.

[0031] The control object position identification unit A003 integrates each sensor information acquired by the state quantity acquisition unit A002 to calculate the position information of the vehicle 001 to be controlled. That is, the control object position identification unit A003 identifies the position of the control object based on the internal information of the control object. For example, in the case of the vehicle 001 equipped with LiDAR, a function of estimating the position of the vehicle using SLAM corresponds to the control object position identification unit A003.

[0032] In addition, even if the vehicle 001 is equipped with a GNSS that can directly obtain the vehicle position, a sensor fusion function that uses an IMU to complement the update period of the position information provided by the GNSS is required, and this function corresponds to the control target position identification unit A003.

[0033] In addition, in the case of a vehicle 001 in which the sensor mounted on the control object only acquires position information, for example, a GNSS is mounted only, the GNSS is configured as both the state quantity acquisition unit A002 and the control object position identification unit A003.

[0034] In addition, if a system is configured in which a sensor with registered position information is present on the travel route and the vehicle 001 acquires its own position by passing over the sensor, this system may be treated as the controlled object position identifying unit A003. Such a system is used for controlling an autonomous mobile object in a limited area such as a logistics warehouse.

[0035] The object identification unit A004 uses techniques such as image recognition, Semantic SLAM, and Semantic Segmentation based on the sensor data (external information) acquired by the external information acquisition unit A001 to identify attributes of moving objects (non-controlled objects) around the controlled vehicle 001. This identification classifies the moving object based on characteristics related to its movement, such as whether it is a pedestrian, a bicycle, or a four-wheeled vehicle.

[0036] In addition, some stereo cameras and millimeter wave radars are equipped with a processing function for identifying moving objects. In this case, the external information acquisition unit A001 and the object identification unit A004 do not need to be separated.

[0037] Here, the characteristics related to movement refer to the equation of motion that governs the dynamic characteristics of the moving object and the maximum value of the moving speed. In the case of pedestrian 002, it can move freely on a two-dimensional plane. On the other hand, vehicles such as wheelchairs, bicycles, scooters, and cars include holonomic constraints, such as not being able to move straight sideways.

[0038] Furthermore, since the moving speed of pedestrians varies depending on whether they are adults, children, or the elderly, it is desirable to consider the characteristics of pedestrians classified in more detail.

[0039] The non-controlled object position identification unit A005, like the object identification unit A004, calculates the position of a moving object (non-controlled object) using Semantic SLAM technology or the like based on the sensor data (external information) acquired by the external information acquisition unit A001.

[0040] By correctly acquiring the orientation (direction) and speed of a moving object, it is expected that the accuracy of predicting the position of the moving object can be improved. For this reason, for example, it is desirable that the non-control target position identification unit A005 includes a process for estimating the orientation (direction), rotational speed (angular velocity) and moving speed of the moving object by acquiring position information (X, Y coordinates on a two-dimensional plane) at multiple times.

[0041] The object identification unit A004 and the non-controlled object position identification unit A005 can be executed simultaneously (in parallel processing) using the same technology.

[0042] 3, the non-control object reachable area calculation unit A006 is composed of a trajectory prediction unit A006a and a reachable area calculation unit A006b, and calculates a non-control object area that is reachable within a predetermined time. That is, the non-control object area that the non-control object can reach within a predetermined time is calculated using the attribute identified by the object identification unit A004 and the position of the non-control object calculated by the non-control object position identification unit A005.

[0043] The trajectory prediction unit A006a predicts the moving trajectory of the non-controlled object at an arbitrary time ahead based on the moving object characteristics of the non-controlled object identified by the object identification unit A004 and the position, direction (azimuth), and speed of the non-controlled object calculated by the non-controlled object position identification unit A005 before an arbitrary time. Note that the arbitrary time means the product of the sampling period Δta and the prediction step Na.

[0044] 4A and 4B show an example of a movement trajectory prediction by the trajectory prediction unit A006a.

[0045] FIG. 4A is an example of a predicted trajectory of the pedestrian 002 from time t0 to time t4 (t0+Na×Δt) four steps ahead (Na=4). The pedestrian 002 can not only move straight, but also move diagonally and horizontally freely, so the range of possible trajectories becomes wider as time progresses. The trajectory prediction unit A006a calculates possible trajectories based on the direction and speed of the non-controlled object at time t0, so it is characterized in that it maintains the moving direction, that is, continues to move to the left of the paper. Note that the walking speed and straightness of the pedestrian 002 vary, so the set of predicted trajectories has uncertainty, so it is desirable to calculate the elliptical predicted trajectory set shown in FIG. 4A.

[0046] 4B is an example of a similar predicted trajectory for bicycle 005. Because it is more difficult for bicycle 005 to move left and right than pedestrian 002, the candidate range does not expand in the vertical direction on the paper. However, because bicycle 005 moves faster than pedestrian 002, the candidate range expands in the horizontal direction on the paper.

[0047] The predicted trajectories shown in Figures 4A and 4B can be calculated by prediction based on a probability model. In other words, the average value and variance can be obtained. Therefore, it is desirable to give the spread of the candidate range by the variance as shown in Figures 4A and 4B.

[0048] For example, the X and Y coordinates of the non-controlled moving objects (pedestrian 002, bicycle 005) at time t0 are 0 , Y 0 , and the direction is θ 0 , angular velocity is ω 0 , the moving speed is v 0 In this case, the X and Y coordinates of the moving object at time k can be calculated sequentially by the following equation (1).

[0049]

number

[0050] Here, w X , w Y , w θ , w v , w ω is the process noise. In the case of pedestrian 002, who can move freely on the X-Y plane, the velocity v in the X direction is expressed as follows: X and the velocity in the Y direction v Y may be calculated individually.

[0051]

number

[0052] The reachable area calculation unit A006b calculates an area that the non-controlled object can reach at an arbitrary time ahead based on the moving object characteristics of the non-controlled object identified by the object identification unit A004 and the position, orientation, and speed of the non-controlled object calculated by the non-controlled object position identification unit A005 prior to an arbitrary time.

[0053] 5A, 5B, and 5C show examples of the calculation results of the reachable area.

[0054] 5A is an example of a reachable area of ​​the pedestrian 002 from time t0 to time t4 (t0+Na×Δt), which is four steps ahead (Na=4). Unlike the predicted trajectory calculated by the trajectory prediction unit A006a, the reachable area is calculated as an area in which the moving body can move, regardless of the direction of the moving body at time t0.

[0055] Therefore, the reachable area of ​​pedestrian 002 spreads out in a circular shape. This shape is obtained by arranging each value of the predicted trajectories t1 to t4 in FIG. 4A in all directions (on a circular arc). Note that the reachable area is the area that is reachable by each time, so the reachable area at time tk also includes the reachable area at the previous time tk-1. For example, the reachable area at t2 in FIG. 5A indicates an area that includes the circle of the reachable area at t1, and does not indicate the area of ​​the difference (ring) between the circle of the reachable area at t1 and the circle of the reachable area at t2.

[0056] It is desirable to use the larger value of the moving speed when calculating the reachable area between the moving speed at time k and the moving speed set according to the moving object characteristics. In other words, as shown in the following equation (3), the moving speed v obtained at time 0 is calculated as 0 v calculated by sequentially updating k and v, which is determined for each movement characteristic. pre The maximum value of v pred Calculate the reachable area using

[0057]

number

[0058] FIG. 5B is an example of a similar reachable area for bicycle 005. Since bicycle 005 can basically only move forward, the reachable area has a more limited shape than the reachable area for pedestrian 002 in FIG. 5A. Note that the reachable area may be extended in the backward direction of bicycle 005 as in FIG. 5C. It is also possible to assume movement in all directions as in FIG. 5A.

[0059] The control target reachable area calculation unit A007 calculates a control target reachable area where the moving object, which is the control target, can reach within a predetermined time based on the position information of the control target acquired by the control target position identification unit A003. Note that the arbitrary time is set in the same way as the non-control target reachable area calculation unit A006.

[0060] This calculation is basically the same as that of the reachable area calculation unit A006b of the non-control target reach area calculation unit A006. However, it should be noted that the control target reach area calculation unit A007 has little uncertainty because it is a function for calculating the movement area of ​​a moving object controlled by an autonomous control system. In other words, when dealing with a vehicle 001 traveling on a roadway 007 as shown in FIG. 6A, there is no need to consider the trajectory of the vehicle 001 going off the roadway 007, nor is there any need to consider the trajectory of the vehicle 001 going off into the opposite lane. Similarly, there is no need to evaluate the situation in which the vehicle 001, which is an automobile, moves backward.

[0061] As described above, it should be noted that the control target reachable area at time tk also includes the reachable area at the previous time tk-1.

[0062] It is desirable to calculate the farthest point pk (p1, p2, p3, ...) of the reachable area at each time tk as a point reached by uniform motion using the smaller of the maximum speed of the vehicle 001 or the maximum speed (such as the legal speed limit) set for the roadway 007 on which the vehicle is traveling.

[0063] As shown in FIG. 6B, in the case of a road 007 with two lanes on each side, the reachable area may be set taking into consideration actions including lane changing.

[0064] The intersection area calculation unit A008 uses the movement trajectory and reachable area of ​​the non-controlled object calculated by the non-controlled object reachable area calculation unit A006 and the reachable area of ​​the controlled object calculated by the control object reachable area calculation unit A007 to calculate an intersection area between the vehicle 001, which is a controlled object, and the pedestrian 002, which is a non-controlled object. In other words, the intersection area calculation unit A008 uses the set of positions of the controlled object calculated by the control object reachable area calculation unit A007 and the set of positions of the non-controlled object 002 calculated by the non-controlled object position identification unit A005 to calculate an intersection area of ​​each set of positions.

[0065] The specific processing contents of the intersection area calculation unit A008 will be described in detail with reference to Figures 7A to 7E. Note that Figures 7A to 7E are different from Figures 2A and 2B in that roadways and crosswalks are omitted in order to prioritize visibility of the figures.

[0066] Fig. 7A is a diagram showing, as an example, the relationship between the reachable areas of the controlled object and the non-controlled object up to four time ahead (t4). In Fig. 7A, the controlled object is a vehicle 001, and the non-controlled object is a pedestrian. Note that the present invention is not limited to performing evaluation up to four time ahead.

[0067] In Fig. 7A, the reachable areas of vehicle 001 and pedestrian 002 do not intersect at all. In other words, vehicle 001 will not come into contact with pedestrian 002 until time 4, no matter how pedestrian 002 moves.

[0068] Fig. 7B shows a situation where the position of pedestrian 002 at time t0 is different from that in Fig. 7A. In this situation, the reachable range of pedestrian 002 at time t4 intersects with the reachable range of vehicle 001 (A1 in the figure). That is, in the situation in Fig. 7B, vehicle 001 and pedestrian 002 may come into contact at time t4. In the present invention, the area A1 in the figure is called the first intersecting area.

[0069] Fig. 7C shows a case where the position of the pedestrian 002 at time t0 is the same as in Fig. 7B, but the orientation of the pedestrian 002 is different. Since the reachable area of ​​the non-controlled object calculated by the non-controlled object reachable area calculation unit A006 does not depend on the orientation of the non-controlled object (pedestrian 002), the reachable range of the pedestrian 002 at time t4 intersects with the reachable range of the vehicle 001 (first intersecting area A1 in the figure), similar to Fig. 7B.

[0070] In a situation where it is impossible to predict how the pedestrian 002 will move, the possibility of collision between the vehicle 001 and the pedestrian 002 cannot be denied in either the situation shown in Fig. 7B or 7C. However, generally speaking, it is difficult to imagine a situation in which the pedestrian 002 suddenly changes the direction of movement. To take such a situation into account, the intersection area calculation unit A008 also calculates a second intersection area A2, which is calculated using a different method from the first intersection area A1.

[0071] Fig. 7D shows an example of calculation of the second intersection area A2 in the same situation as Fig. 7B. The second intersection area A2 is an area where the movement trajectory of the non-controlled object calculated by the non-controlled object reachable area calculation unit A006 and the reachable area of ​​the controlled object calculated by the controlled object reachable area calculation unit A007 intersect.

[0072] Fig. 7E shows an example of calculation of the second intersection area A2 in the same situation as Fig. 7C. Since the movement trajectory of the non-control object calculated by the non-control object reachable area calculation unit A006 takes into account the orientation of the pedestrian 002, the reachable area of ​​the control object calculated by the control object reachable area calculation unit A007 does not intersect.

[0073] In this way, the possibility of a collision between the vehicle 001 and the pedestrian 002 is evaluated depending on the presence or absence of the second intersection area A2.

[0074] The control parameter setting unit A009 sets the control parameters of the controlled object to be used by the manipulated variable calculation unit A010, which will be described later, according to the calculation result of the intersection area calculation unit A008. That is, the control parameter setting unit A009 sets the control parameters to be larger so that the controlled object behaves in such a way that the greater the intersection amount area of ​​the intersection area calculated by the intersection area calculation unit A008, the greater the distance between the controlled object and non-controlled objects.

[0075] It is desirable to increase the values ​​of the control parameters W1 and W2 of the control object as the areas of the first intersection area A1 and the second intersection area A2 calculated by the intersection area calculation unit A008 are larger, as shown in Fig. 8A and Fig. 8B. As shown in Fig. 8A and Fig. 8B, different parameters W1 and W2 may be set for the first intersection area A1 and the second intersection area A2, or the same parameter W may be set.

[0076] 8C, the area of ​​the intersection region and the magnitude of the parameter W may be defined in a nonlinear relationship. The method of using the control parameters will be described later in detail.

[0077] The control parameters W1 and W2 may be defined as multiple parameters having different meanings depending on the control calculation executed by the manipulated variable calculation unit A010 described later. In such a case, it is preferable to provide multiple relational expressions or tables showing the parameters and the areas of the intersection regions shown in Figures 8A to 8C.

[0078] The operation amount calculation unit A010 calculates actuator command values ​​for controlling the behavior of the controlled object, using the position of the controlled object calculated by the controlled object position identification unit A003, the position of the non-controlled object calculated by the non-controlled object position identification unit A005, the predicted trajectory of the non-controlled object (a set of trajectories that the non-controlled object will pass through within a specified time) calculated by the non-controlled object arrival area calculation unit A006, and the control parameters set by the control parameter setting unit A009.

[0079] For example, when the controlled object is vehicle 001, calculations related to accelerator instructions and brake instructions for controlling the acceleration and deceleration of vehicle 001, and steering instructions for controlling the direction of vehicle 001 are executed by operation amount calculation unit A010.

[0080] Since the autonomous control system A100 of the present invention performs control based on the future behavior of controlled and uncontrolled objects, it is desirable to realize the manipulated variable calculation unit A010 using model predictive control (hereinafter, MPC). The manipulated variable calculation unit A010 uses the control parameters determined by the control parameter setting unit A009 and an evaluation function J whose value decreases as the controlled object approaches the desired behavior, and calculates a control input so that the value of the evaluation function J becomes smaller than the previous value within a range that satisfies a predetermined constraint condition.

[0081] Hereinafter, the specific calculation contents of the operation amount calculation unit A010 will be described using an example of a case where an autonomous vehicle is controlled using MPC.

[0082] If the vehicle 001 to be controlled is a four-wheeled vehicle as shown in FIG. 9, considering the vector x=[XY θ] that combines the coordinates (X, Y) and direction θ of the vehicle 001, simplified dynamics can be given by the following equations (4a) and (4b).

[0083]

number

[0084] In addition, the control input u in equations (4a) and (4b) is the vehicle speed v and the steering angle φ. Also, please note that the lowercase x represents a state quantity, and the uppercase X represents the X coordinate of vehicle 001.

[0085] The differential equation of equation (4a) can be discretized using the sampling period Δt as shown in the following equation (5).

[0086]

number

[0087] On the other hand, the simplified dynamics of the uncontrolled object can be given by equation (1) or equation (2).

[0088] The non-control object reach area calculation unit A006, the control object reach area calculation unit A007, and the operation amount calculation unit A010 do not need to use the same value for the sampling period Δt. For example, the sampling period of the control object reach area calculation unit A007 or the non-control object reach area calculation unit A006 may be Δta=100 ms, and the sampling period of the operation amount calculation unit A010 may be Δt=10 ms.

[0089] In order to automatically control a target vehicle using MPC, the desired behavior must be expressed as an evaluation function.

[0090] First, the desired behavior of the controlled vehicle 001 is that the position and orientation of the vehicle 001 follow the target trajectory r = [xr yr θr]T (T means transpose). This behavior can be expressed as an evaluation function as shown in the following equation (6).

[0091]

number

[0092] In equation (6), N is the prediction step, S and Q are weights, and the evaluation function J 1 is time k 0 It means a weighted sum of the target trajectory r and the vehicle position p from N steps ahead. Note that this prediction step N does not have to match the prediction steps of the controlled object reachable area calculation unit A007 and the non-controlled object reachable area calculation unit A006. However, it is desirable that the prediction time calculated by the manipulated variable calculation unit A010, that is, the product of the sampling period Δt and the prediction step N, be smaller than the product of the sampling period Δta and the prediction step Na of the controlled object reachable area calculation unit A007 and the non-controlled object reachable area calculation unit A006.

[0093] In other words, it is desirable that the manipulated variable calculation unit A010 does not calculate a trajectory that deviates from the reachable region in the calculation of the model predictive control.

[0094] The vehicle position at each time k can be predicted using equation (5). For example, in the case of a vehicle 001 traveling on a general public road, the target trajectory pr may be set at the center of a roadway 007 several meters away from the current vehicle position.

[0095] In addition, from the viewpoint of operating the vehicle 001 to be controlled, the smaller the control input (acceleration / deceleration and steering amount), the less energy consumption there will be, so it is desirable to add an evaluation function including the control input u as in the following equation (7).

[0096]

number

[0097] In equation (7), R is a weight.

[0098] Here, vector q is defined as including the position information of the non-controlled object (pedestrian 002) predicted using equation (1) or equation (2). Also, consider a circle with a radius rp surrounding vehicle 001, which is a controlled object, and a circle with a radius rq surrounding pedestrian 002, which is a non-controlled object, as shown in Fig. 10. In this situation, the distance rk between vehicle 001 and pedestrian 002 can be expressed by the following equation (8a).

[0099]

number

[0100] Here, pk is a vector that matches the position (Xk, Yk) of the vehicle 001 to be controlled at time k. With the above preparations, the condition for preventing contact between the vehicle 001 and the pedestrian 002 can be given by equation (8b).

[0101] When there are restrictions on the speed v and steering φ, which are the control inputs for the vehicle 001 to be controlled, the formula (8c) may be taken into consideration as a constraint on the control input. lb is the lower limit of the input, u ub means the upper limit of the input. If there are constraints on acceleration as well as speed, it is sufficient to consider the following equation (9), which adds speed to the state quantity in addition to equation (4), and simultaneously consider the constraint condition equation (8c) for the control input and the constraint condition equation (8d) for the state quantity.

[0102]

number

[0103] In formula (8d), x lb is the lower limit of the state quantity, x ub is the upper limit of the state quantity. Note that in equation (8d), it is not necessary to consider the lower and upper limits for all state quantities. For example, in the case of a moving object moving in free space, there is no need to impose constraints on the X and Y coordinates.

[0104] With the above preparations in mind, in model predictive control, a control input u is calculated to solve the optimization problem of the following equation (10). Since the optimization problem of equation (10) includes the constraint of equation (8b), if the behavior of pedestrian 002, which is a non-controlled object, complies with prediction model equation (1) or equation (2), it is possible to generate an action that prevents vehicle 001 from coming into contact with pedestrian 002.

[0105]

number

[0106] However, as described above, there is no guarantee that the pedestrian 002 will behave according to the prediction model. In such a case, even if control is performed taking into account the constraint condition of equation (8b), there is no guarantee that the vehicle 001 will be able to avoid contact with the pedestrian 002.

[0107] Therefore, in the present invention, the closer the predicted positions of the pedestrian 002 and the vehicle 001 are to each other, that is, the smaller the rk in the formula (8a) is, the lower the evaluation function J 3 Consider equation (11a). In equation (11a), W is a weighting parameter, and in equation (11b), ε is a small coefficient for preventing division by zero, and a is a coefficient for adjusting the shape of the function. The evaluation function used in the manipulated variable calculation unit A010 includes the product of a penalty function whose value increases as the distance between the controlled object and the uncontrolled object decreases, and the control parameters determined in the control parameter setting unit A009.

[0108]

number

[0109] l(p k , q k ) is the position p of vehicle 001 at time k k and pedestrian 002's position q k (Both at time k=k 0 is the measured position, and k=k 0 This is a function that gives a penalty to the relationship between the position and the predicted position.

[0110] Figure 11 shows the function l(p k , q k ) is an example. Note that the shape of the function changes depending on ε and a, so the shape shown in Figure 11 is only an example. As shown in Figure 11, the function l(p k , q k ) means that the distance between vehicle 001 and pedestrian 002 increases, that is, r k The larger r, the smaller the penalty. Conversely, the closer the distance between vehicle 001 and pedestrian 002, that is, r k The smaller the value, the larger the penalty.

[0111] Penalty J 3By using the evaluation function formula (12) with the addition of the above, in order to reduce the value of the evaluation function J, it becomes necessary to increase the distance between the vehicle and the pedestrian, making it possible to generate safer actions.

[0112]

number

[0113] On the other hand, Penalty J 3 The influence of other evaluation functions J 1 , J 2 , the vehicle 001 may take unnecessary avoidance action. For example, as shown in FIG. 12, if the penalty J 3 If the influence of is large, vehicle 001 will take unnecessary avoidance action. Such unnecessary avoidance action can be eliminated by adjusting the weight parameter W in equation (11a). In particular, if W = 0, J 3 =0, the optimization problems of equation (10) and equation (12) coincide, and vehicle 001 will not take avoidance action unless it meets constraint condition equation (8b). In order to realize such an action, it is desirable to make the weight parameter W variable according to the area of ​​the intersection region, as shown in Figures 8A to 8C.

[0114] When there is no first intersection area A1 (area is 0), there is no possibility of contact with the vehicle 001 no matter what action the pedestrian 002 takes, and therefore there is no need for the vehicle 001 to take evasive action, so the weighting parameter W1 is set to 0. For this reason, as shown in Figs. 8A to 8C, the relational expression between the area of ​​the intersection area and the weighting parameter W is characterized in that it passes through the origin. However, as described above, the relational expression between the area of ​​the intersection area and the weighting parameter W does not have to be a linear function. For example, it may be a nonlinear function, or a specific value may be set as the upper limit.

[0115] When the second intersection area A2 exists (the area is not 0), it is desirable to take a safer action since there is a high possibility that the vehicle 001 will come into contact with the pedestrian 002. In other words, it is desirable to further increase the weighting parameter W. For this reason, it is desirable to set the weighting parameter W2 to a large value according to the area of ​​the second intersection area A2, as shown in FIG. 8B.

[0116] The penalty J using the weight parameters W1 and W2 obtained by the above process is 3 The desired safe operation can be achieved by using equation (13) which utilizes

[0117]

number

[0118] In the situation shown in Fig. 7C, unless the pedestrian 002 changes his / her direction of travel, the vehicle 001 and the pedestrian 002 are unlikely to collide. An example of a situation in which the pedestrian 002 suddenly changes his / her direction of travel is when a child, who is the pedestrian 002, is called by a friend and runs off in the direction of the voice, as shown in Fig. 13A. On the other hand, elderly people generally do not have high agility, so the likelihood of such a situation occurring is low.

[0119] In this way, since the characteristics of pedestrians 002 vary greatly even among pedestrians 002, if the object identification unit A004 can classify pedestrians 002, it is desirable to set a different weighting parameter W for each classification (child, general, elderly) as shown in Fig. 13B. In other words, the information identified by the object identification unit A004 includes not only the movement mode of the non-controlled object but also behavior characteristics related to safe behavior that the non-controlled object can take, and the control parameter setting unit A009 sets a larger control parameter so that behavior that keeps a wider distance between the controlled object and the non-controlled object is generated as the non-controlled object has behavior characteristics that make it difficult to take safe behavior, even if the set value of the intersection area calculated by the intersection area calculation unit A008 is the same.

[0120] In the explanation so far, the area where the non-control target reach area and the control target reach area intersect is simply evaluated, and the weight parameter W is changed according to the area of ​​this intersection area. If such processing is performed, there remains a possibility that the vehicle 001 will take excessive avoidance action.

[0121] For example, consider a situation in which a vehicle 001 is traveling around a curve, as shown in Fig. 14A. To simplify the situation, the area that the vehicle 001 and the pedestrian 002 cannot reach is illustrated as three steps ahead (t3).

[0122] 14A, the area that pedestrian 002 can reach by time t3 and the area that vehicle 001 can reach by time t1 intersect at A1a, and similarly, the area that vehicle 001 can reach by time t2 intersects at A1b. However, the area that pedestrian 002 can reach at times t1 and t2 does not intersect with the area that vehicle 001 can reach at time t1.

[0123] Therefore, unless vehicle 001 is traveling at a low speed, it will not come into contact with pedestrian 002 in area A1a. Similarly, since the area reachable by pedestrian 002 at time t2 and the area reachable by vehicle 001 at time t2 do not intersect, vehicle 001 will not come into contact with pedestrian 002 in area A1b either, unless vehicle 001 is traveling at a low speed.

[0124] That is, in the situation of Fig. 14A, as long as the vehicle 001 is traveling at a normal speed, it is not necessary to take any action to avoid the pedestrian 002. In order to appropriately handle such a situation, it is preferable to use a weighted area obtained by adding values ​​obtained by multiplying the areas Aia, Aib, ... (i=1, 2, meaning the first intersection area A1 and the second intersection area A2) of the intersection areas as shown in Fig. 14B, rather than determining the weight parameter W simply according to the area of ​​the intersection area as shown in Figs. 8A to 8C.

[0125] It is desirable to set the weighting parameters ai, bi, ... to large values ​​when the arrival areas at the same time intersect. Furthermore, as in the situation shown in Fig. 14a, when the arrival area of ​​a controlled object at a certain time tk intersects with the arrival area of ​​a non-controlled object at a time tk+x (x≧1) that is earlier than the time tk, it is desirable to set the weighting parameters ai, bi, ... to small values. Furthermore, it is desirable to make the weighting parameters ai, bi, ... smaller as the above x increases.

[0126] If the above x exceeds a predetermined value, the weighting parameters ai, bi, . . . may be set to 0.

[0127] In other words, in a situation like that shown in FIG. 14A, unnecessary avoidance actions can be eliminated by setting the weighted area A1w to 0 by setting x=1 as the predetermined value, a1=0, a2=0, and thereby setting the control parameter W=0.

[0128] The processing contents of the autonomous control system A100 are shown in the flowchart of FIG.

[0129] In process FC01 in FIG. 15, the detection values ​​of various sensors (external information acquisition unit A001, state quantity acquisition unit A002) are updated.

[0130] Next, in process FC02, the current position of the control object is calculated. This process is executed by the control object position specifying unit A003.

[0131] Next, in process FC03, it is confirmed whether the external information acquisition unit A001 has detected a moving object (non-controlled object). If a moving object is detected (YES), the process proceeds to process FC04. On the other hand, if a moving object is not detected (NO) in process FC03, the process proceeds to process FC09. In this case, the controlled vehicle performs a control operation that takes into account only the host vehicle 001. In other words, model predictive control is used to minimize equation (13) without taking into account the constraint conditions regarding contact with the moving object. Note that W1 and W2 under this condition become 0.

[0132] In process FC04, non-control objects are identified from the data acquired by the external information acquisition unit A001. Then, according to the identification result, the moving speed limit of the non-control objects (v in equation (3)) is calculated. pre ) and the size of the uncontrolled object (radius r q This process corresponds to the object identification unit A004.

[0133] In process FC05, the current position of the non-controlled object is calculated. This process is executed by the non-controlled object position specifying unit A005.

[0134] In process FC06, the areas that each of the controlled object and non-controlled object can reach within a predetermined time from the current position are calculated. These processes are executed by the controlled object reachable area calculation unit A007 and the non-controlled object reachable area calculation unit A006, respectively. In addition, for the non-controlled object, a trajectory prediction up to a predetermined time ahead is also performed in parallel according to the characteristics of the object model (pedestrian (child, general, elderly), bicycle, motorcycle, etc.) identified in process FC003. This trajectory prediction process is executed by the trajectory prediction unit A006a of the non-controlled object reachable area calculation unit A006.

[0135] In process FC07, a first intersection area where the reach area of ​​the controlled object and the reach area of ​​the non-controlled object intersect, and a second intersection area A2 where the reach area of ​​the controlled object and the trajectory prediction range of the non-controlled object intersect are calculated.

[0136] In process FC08, the evaluation function J in formula (11b) is calculated using the area of ​​the first intersection area A1 and the area of ​​the second intersection area A2. 3 The parameter W to be used is determined.

[0137] In process FC09, a control input u of the controlled vehicle is calculated using the model predictive control shown in equation (13) so as to minimize the determined evaluation function J. This process is executed by the manipulated variable calculation unit A010.

[0138] In process FC10, each actuator is controlled so as to realize the control input u calculated in process FC09. This process is executed by actuator A011.

[0139] So far, we have explained a control method that utilizes the penalty function of formula (11b) and its weight parameter W in order to achieve safe driving. However, the essence of the present invention is to ensure safety by increasing the distance between the vehicle 001 (controlled object) and the pedestrian 002 (non-controlled object) when the reachable area of ​​the controlled object and the reachable area of ​​the non-controlled object intersect, that is, when there is a possibility of a collision.

[0140] Therefore, if the distance between the vehicle 001 and the pedestrian 002 can be increased, the penalty function of formula (11b) does not need to be used. For example, as shown in Figs. 8A to 8C, a model predictive control such as formula (14) can be used by using a parameter W that increases according to the area of ​​the intersection region. The constraint condition of formula (14) means that the distance between the vehicle 001 and the pedestrian 002 does not become less than W, that is, the distance between the vehicle 001 and the pedestrian 002 is W or more under any circumstances. By adopting such a method, the vehicle 001 can realize an operation to avoid the pedestrian 002. The constraint condition used by the operation amount calculation unit A010 includes a minimum distance at which the controlled object and the non-controlled object can approach each other, and the control parameter determined by the control parameter setting unit A009 is used for this minimum distance.

[0141]

number

[0142] According to the first embodiment, it is possible to provide an autonomous control system and an autonomous control method that can drive an autonomous vehicle efficiently and safely even in cases where it is difficult to predict the behavior of moving objects around the autonomous vehicle.

[0143] Example 2 Next, a second embodiment of the present invention will be described.

[0144] In the above-described first embodiment, an autonomous vehicle running on a general public road is the control target.

[0145] The present invention can be applied not only to moving bodies such as the vehicle 001, whose controlled object position changes greatly, but also to automatic operation of systems in which only part of the machine moves.

[0146] Second Embodiment As a second embodiment of the present invention, a case in which a hydraulic excavator performing construction work at a construction site is the control target will be described.

[0147] 16A and 16B show a hydraulic excavator 001A to which the second embodiment of the present invention is applied.

[0148] Fig. 16A is a diagram illustrating the initial posture of a hydraulic excavator 001A, which is a control object of this embodiment 2, from above, and Fig. 16B is a diagram illustrating the final posture of the hydraulic excavator 001A from the side. The hydraulic excavator 001A performs an operation of loading excavated earth and sand onto the bed of a dump truck 002A. The autonomous control system of the embodiment 2 of the present invention is used to control the loading operation without causing the toe position of the hydraulic excavator 001A to come into contact with the dump truck 002A.

[0149] FIG. 17 is a diagram showing an example of a construction site to which the second embodiment of the present invention is applied. As shown in FIG. 16A, FIG. 17 depicts the initial posture of the hydraulic excavator from above. In addition to the hydraulic excavator 001A and the dump truck 002A, there are a number of workers 002B in the construction site to perform detailed excavation work that is difficult for the hydraulic excavator 002A to handle and to monitor the progress of the construction site. The hydraulic excavator 001A is equipped with GNSS and IMU as the state quantity acquisition unit A002, and can acquire its own position. The hydraulic excavator 001A may be equipped with LiDAR as the external information acquisition unit A001, but the LiDAR mounted on the hydraulic excavator 001A can only acquire limited surrounding information, and blind spots are likely to occur. For this reason, it is desirable to install infrastructure sensors at the construction site and utilize the information acquired by the infrastructure sensors.

[0150] For the sake of simplicity, when the hydraulic excavator 001A loads earth and sand, the dump truck 002A is generally stopped, and therefore only the worker 002B is treated as a non-controlled moving object 002. In the first embodiment, a situation in which there is only one pedestrian 002 is assumed, but the present invention can also be applied to a situation in which there are multiple pedestrians 002 (workers 002B).

[0151] In order to control the position of the toe of the hydraulic excavator 001A, the autonomous control system A100 of this embodiment 2 controls each of the swing motor Ex01a for controlling the posture of the upper swing body Ex01 shown in Figure 18, the boom cylinder Ex02a for controlling the posture of the boom Ex02, the arm cylinder Ex03a for controlling the posture of the arm Ex03, and the bucket cylinder Ex04a for controlling the posture of the bucket Ex04.

[0152] The toe position of the hydraulic excavator 001A can be calculated by using the speed of each actuator Ex01a to Ex04a as the control input u and calculating the actual actuator displacement (swing angle, cylinder length). That is, for the control of the toe position of the hydraulic excavator 001A, the model of the control target can be given in the form of equation (5) as in the first embodiment. Furthermore, if a target toe position rk is given above the bed of the dump truck 002A, it can be defined as a problem of minimizing the evaluation function of equation (6) as in the first embodiment.

[0153] The vehicle 001 in the first embodiment can only move on the road (two-dimensional plane), so it is sufficient to consider only the X and Y coordinates. In the second embodiment, the tip of the hydraulic excavator 001A (bucket Ex04) operates in three-dimensional space, so it is necessary to consider the operation in X, Y, and Z coordinates. For this reason, it is also necessary to consider the conditions for preventing contact between the worker 002B and the tip (bucket Ex04) in three-dimensional space.

[0154] In other words, instead of considering the contact of circles on a two-dimensional plane as shown in Fig. 10, consider the contact of spheres in three-dimensional space as shown in Fig. 19. In three-dimensional space, the constraint condition for preventing the sphere of radius rp surrounding bucket EX04 and the sphere of radius rq surrounding worker 002B from contacting each other can be given by equation (14). By utilizing this constraint condition and using the model predictive control of equation (12), a safe loading operation can be achieved that prevents bucket Ex04 from contacting worker 002B.

[0155] When there are multiple workers 002B, model predictive control that considers the constraints of formulas (15a) to (15d) for each worker should be used. In other words, when there are M workers 002B, model predictive control that considers M constraints, as in formula (16), is used.

[0156]

number

[0157]

number

[0158] In this second embodiment, the loading operation has been described as an example, but it goes without saying that the scope of application of the present invention is not limited to the loading operation. In other words, the present invention can be applied to all the work required for construction, such as trench digging and leveling. Specifically, in trench digging, the position of the tip of the bucket of the hydraulic excavator is set to xk, and the control input u is calculated so as to minimize the evaluation function of equation (6) while successively changing the coordinates of the target position rk according to the shape of the trench.

[0159] Unlike public roads, a construction site is a managed area, so ordinary people (pedestrians 002, cyclists 005) cannot enter, and only construction workers are present. Under these conditions, it is possible to manage the safety awareness of worker 002B on-site.

[0160] For example, a skilled worker 002B who has attended a safety course or has years of service can be considered to have a higher level of safety awareness than a worker 002B who has less years of service. The highly safety-conscious worker 002B is familiar with the behavior of the hydraulic excavator 001A and can therefore proceed with the work while appropriately predicting the operation of the hydraulic excavator 001A.

[0161] In this way, if the worker 002B is working while paying attention to safety, it is desirable for the hydraulic excavator 001A to prioritize efficiency and operate along the shortest route. On the other hand, if the worker 002B is in an unusual situation where he or she must be careful about safety, it is desirable for the hydraulic excavator 001A to prioritize safety and operate the bucket Ex04 as far away from the worker 002B as possible.

[0162] These operations can be switched by changing the magnitude of the weight parameter W in equation (11a). Specifically, when the weight parameter W is small, the operation prioritizes efficiency, while when the weight parameter W is large, the operation prioritizes safety.

[0163] To realize such characteristics, it is desirable to change the weighting parameter W in the control parameter setting unit A009 according to the level of skill of the worker 002B at the work site, as shown in FIG. 20. When there are workers 002B with different levels of skill, it is possible to use characteristics according to the level of the worker 002B with the lowest level of skill, or characteristics according to the level of each worker 002B. When the latter method is selected, there will be M different weighting parameters W, and therefore the evaluation function J 3 Note that it is necessary to change to equations (17a) and (17b).

[0164]

number

[0165] In order for the autonomous control system A100 to acquire the proficiency level of each worker 002B, it is preferable that each worker 002B has an individual beacon and that each beacon is associated with the proficiency level of the worker 002B.

[0166] According to the second embodiment, even when it is difficult to predict the behavior of moving objects around the hydraulic excavator that is the object of control, it is possible to provide an autonomous control system and an autonomous control method that are capable of operating a hydraulic excavator efficiently and safely.

[0167] Example 3 Next, a third embodiment of the present invention will be described.

[0168] In the first and second embodiments, the uncontrolled objects are moving objects (pedestrian 002, worker 002B) that move on a two-dimensional plane, but in the third embodiment, the uncontrolled objects are moving objects that move in a three-dimensional space.

[0169] In the third embodiment, a situation is assumed in which a plurality of flying objects fly in a specific limited area, as shown in Fig. 21. The flying object 002C that is not to be controlled may be manually controlled by an operator, or may be automatically controlled by a control system different from the autonomous control system A100 of the present invention.

[0170] The flying object 001B to be controlled is equipped with GNSS and IMU as a state quantity acquisition unit A002, and can acquire its own position. However, since the flying object 001B has a weight limit, it is difficult to equip it with a LiDAR to check the surroundings, so it is desirable for the external information acquisition unit A001 to acquire information using an external sensor.

[0171] Since flying object 001B and flying object 002C can move freely in three-dimensional space, the dynamics for predicting the behavior of the uncontrolled flying object 002C is not equation (2) that was used to predict the behavior of pedestrian 002, but rather equation (18) that takes into account the height direction (Z).

[0172] Similarly, since both the flying object 001B, which is the controlled object, and the flying object 002C, which is the non-controlled object, move in a three-dimensional space, the controlled object reach area calculation unit A007, the non-controlled object reach area calculation unit A006, and the intersection area calculation unit A008 all need to be changed to perform calculations in a three-dimensional space. However, the basic concept is the same as in the first or second embodiment.

[0173] That is, as shown in FIG. 22, the reachable area is represented by a sphere, and the first intersection area A1 is treated as a volume V1 rather than an area. The subsequent processing follows the first or second embodiment, and therefore the description is omitted.

[0174]

number

[0175] According to the embodiments, it is possible to provide an autonomous control system and an autonomous control method that can operate a flying object efficiently and safely even when it is difficult to predict the behavior of moving objects around the flying object to be controlled.

[0176] Although the embodiments of the present invention have been described in detail above using as examples an autonomous vehicle 001 that runs on a general public road, a hydraulic excavator 001A that is a construction machine at a construction site, and a flying object 001B that flies in a limited area, it goes without saying that the application of the present invention is not limited to these cases. For example, the autonomous control system and autonomous control method of the present invention can also be used in transport vehicles at ports and robots that move within theme parks. [Explanation of symbols]

[0177] 001...vehicle, 001A...hydraulic excavator, 001B...flying object, 002...pedestrian, 002A...dump truck, 002B...worker, 002C...flying object, 003...infrastructure sensor, 004...wireless system, 005...bicycle, 007...roadway, 008...area managed by external information acquisition unit, 009...sidewalk, A1...first intersection area, A2...second intersection area, A001...external information acquisition unit , A002···State quantity acquisition unit, A003···Control object position identification unit, A004···Object identification unit, A005···Non-control object position identification unit, A006···Non-control object reachable area calculation unit, A006a···Trajectory prediction unit, A006b···Reachable area calculation unit, A007···Control object reachable area calculation unit, A008···Intersection area calculation unit, A009···Control parameter setting unit, A010···Operation amount calculation unit, A100···Autonomous control system

Claims

1. a state quantity acquisition unit that acquires a state quantity of a control object; a control object position specifying unit that specifies a position of the control object based on the state quantity; a control object reachable area calculation unit that calculates a control object reachable area that the control object can reach within a predetermined time based on the position of the control object; an external information acquisition unit that acquires external information of the control target; an object identification unit that identifies attributes of non-control objects based on the external information; a non-control object position specifying unit that calculates the position of the non-control object based on the external information; a non-control object reachable area calculation unit that calculates a non-control object area that the non-control object can reach within a predetermined time using the attribute and the location; an intersection area calculation unit that calculates an intersection area of ​​each set of positions using the set of positions of the control object and the set of positions of the non-control object calculated by the non-control object reach area calculation unit; a control parameter setting unit that sets a control parameter of the controlled object so that a behavior is generated in which the controlled object keeps a wider distance from the non-controlled object as the intersection amount area of ​​the intersection area of ​​each of the sets of positions increases; an operation amount calculation unit that calculates a control input using the control parameters and an evaluation function whose value decreases as the controlled object approaches a desired behavior, so that the value of the evaluation function becomes smaller than a previous value within a range that satisfies predetermined constraint conditions; An autonomous control system comprising:

2. 2. The autonomous control system according to claim 1, the non-control object reachable area calculation unit calculates a predicted trajectory, which is a set of trajectories that the non-control object will pass through within a predetermined time, and a non-control object area that the non-control object can reach within the predetermined time, using the attribute identified by the object identification unit and the position of the non-control object calculated by the non-control object position identification unit; the intersection area calculation unit calculates a first intersection area using the set of positions of the control objects calculated by the control object reach area calculation unit and the set of positions of the non-control objects calculated by the non-control object reach area calculation unit, and calculates a second intersection area using the set of positions of the control objects calculated by the control object reach area calculation unit and the set of predicted trajectories calculated by the non-control object reach area calculation unit; The control parameter setting unit sets the control parameters so that, even in a situation where the set values ​​of the first intersection areas are the same, the larger the set of the second intersection areas, the greater the behavior of keeping a distance between the controlled object and the non-controlled object. An autonomous control system characterized by:

3. 2. The autonomous control system according to claim 1, The information identified by the object identification unit includes, in addition to the movement method of the non-controlled object, behavior characteristics related to safe behavior that the non-controlled object can take, The control parameter setting unit sets the control parameters so that, even if the set value of the intersection area calculated by the intersection area calculation unit is the same, the more the uncontrolled object has behavior characteristics that make it difficult to take a safe action, the greater the distance between the controlled object and the uncontrolled object is generated. An autonomous control system characterized by:

4. 2. The autonomous control system of claim 1, the intersection area calculation unit calculates the intersection area of ​​the sets of positions at each time of the control object calculated by the control object reach area calculation unit and the sets of positions at each time of the non-control object calculated by the non-control object reach area calculation unit; The control parameter setting unit sets the control parameters so that, even if the set values ​​of the intersection areas calculated by the intersection area calculation unit are the same, the larger the set value of the intersection area between the set of positions of the controlled object at each time and the set of positions of the non-controlled object at each time at the same time, the wider the behavior of the controlled object and the non-controlled object. An autonomous control system characterized by:

5. 2. The autonomous control system of claim 1, The evaluation function used in the manipulated variable calculation unit includes a product of a penalty function whose value increases as the distance between the controlled object and the non-controlled object decreases, and the control parameter determined by the control parameter setting unit. An autonomous control system characterized by:

6. 2. The autonomous control system of claim 1, The constraint conditions include a minimum distance that the controlled object and the non-controlled object can approach each other, and the control parameters determined by the control parameter setting unit are used for the minimum distance. An autonomous control system characterized by:

7. 7. The autonomous control system according to claim 1, An autonomous control system characterized in that the controlled object is a vehicle.

8. 7. The autonomous control system according to claim 1, An autonomous control system characterized in that the controlled object is a hydraulic excavator.

9. 7. The autonomous control system according to claim 1, An autonomous control system characterized in that the controlled object is a flying object.

10. Acquire the state quantity of the controlled object, Identifying a position of the controlled object based on a state quantity; calculating a control object reachable region that the control object can reach within a predetermined time based on the position of the control object; Acquire external information of the controlled object; Identifying attributes of the uncontrolled object based on the external information; Calculating the position of the non-controlled object based on the external information; calculating a non-control object area that the non-control object can reach within a predetermined time using the attribute and the position; calculating an intersection area of ​​each set of positions using the set of positions of the controlled object and the set of positions of the non-controlled object; setting a control parameter of the controlled object so that a behavior of the controlled object is generated in which the controlled object keeps a wider distance from the non-controlled object as the intersection amount area of ​​the intersection area of ​​each of the sets of positions increases; using the control parameters and an evaluation function whose value decreases as the controlled object approaches a desired behavior, a control input is calculated so that the value of the evaluation function becomes smaller than the previous value within a range that satisfies a predetermined constraint condition; An autonomous control method characterized by:

11. The autonomous control method according to claim 10, Using the attributes and the position of the non-control object, a predicted trajectory is calculated, which is a set of trajectories that the non-control object will pass through within a predetermined time, and a non-control object region that the non-control object can reach within the predetermined time; calculating a first intersection region using the set of positions of the controlled object and the set of positions of the non-controlled object, and calculating a second intersection region using the set of positions of the controlled object and the set of predicted trajectories; The control parameters are set so that, even in a situation where the set values ​​of the first intersection areas are the same, the larger the set of the second intersection areas, the greater the behavior of keeping a greater distance between the controlled object and the non-controlled object. An autonomous control method characterized by:

Images