Autonomous Control System

Abstract

To efficiently and safely drive a vehicle even when it is hard to predict the behavior of surrounding moving bodies.SOLUTION: An autonomous control system calculates a position of a control object, identifies an attribute of a non-control object, calculates the position of the non-control object, evaluates a degree of deflection of a moving trajectory of the non-control object from a moving predictive trajectory of the non-control object associated with the attribute of the non-control object, determines a safety criterion concerning an action of the control object on the basis of the attribute of the non-control object and the degree of deflection from the moving predictive trajectory, and corrects the action of the control object on the basis of the position of the control object, the position of the non-control object, and the safety criterion, so as to prevent the control object from approaching the non-control object as the safety criterion increases.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an autonomous control system. [Background technology]

[0002] In recent years, the development of autonomous driving technology has progressed in order to reduce traffic accidents and traffic congestion. Hopes are high for autonomous driving technology in the logistics industry, which is facing a serious labor shortage. Autonomous driving technology can be applied to a wide range of applications, including trucks that collect and deliver goods on public roads, 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 inter-process transport vehicles.

[0003] With the exception of fully automated large-scale logistics warehouses, autonomous driving technology will be 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 coming into contact with pedestrians and non-automated vehicles.

[0004] Generally, to achieve such safety functions, the trajectory of pedestrians and non-automated vehicles is predicted, and the automated vehicle is controlled to prevent the vehicle from coming into contact with that predicted trajectory.

[0005] The motion of non-automated vehicles is primarily governed by nonholonomic constraints, meaning 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.

[0006] To address these issues, Patent Document 1 presents a driving assistance system that uses recorded past trajectory information of moving objects (including pedestrians) to calculate the probability (predictive information in the patent document) that a moving object is on the planned route of an autonomous vehicle, thereby supporting safe driving in environments that include pedestrians and are difficult to predict. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-53846 Summary of the Invention [Problem to be solved by the invention]

[0008] In Patent Document 1, in order to generate predictive information for a moving object, a large amount of information recording the past trajectories of the moving object is required. Therefore, if this record (database) is insufficient, it is not possible to provide appropriate predictive information. Furthermore, if appropriate predictive information cannot be obtained, it is expected that it will be difficult to provide a sufficient safe driving function.

[0009] The present invention was devised to solve the above-mentioned problems, and aims to provide an autonomous control system that can drive a vehicle efficiently and safely even when it is difficult to predict the behavior of surrounding moving objects. [Means for solving the problem]

[0010] An autonomous control system according to one aspect of the present invention is an autonomous control system that, in an area where control targets, which are moving objects whose behavior can be controlled, and non-control targets, which are moving objects whose behavior cannot be controlled, coexist, controls the behavior of the control targets so that the control targets do not come into contact with the non-control targets, and includes a control target position calculation unit that calculates the position of the control targets, and a target identification unit that identifies attributes of the non-control targets; Based on environmental information obtained from external recognition sensors a non-control object position calculation unit that calculates the position of the non-control object; using an average value and a variance of a trajectory calculated based on the position of the non-control object calculated a predetermined time ago by the non-control object position calculation unit and a time change in the position of the non-control object, A predicted movement trajectory of the non-control object corresponding to the attribute of the non-control object identified by the object identification unit Calculate The non-control target position calculated by the non-control target position calculation unit A movement trajectory of the non-controlled object is calculated based on a time change in the position, and the movement trajectory is calculated based on the time change in the position of the non-controlled object.the control object is configured to determine a safety standard for the behavior of the controlled object based on the attributes of the non-controlled object identified by the object identification unit and the degree of deviation from the predicted movement trajectory evaluated by the trajectory deviation evaluation unit; and the behavior modification unit is configured to modify the behavior of the controlled object based on the position of the controlled object calculated by the controlled object position calculation unit, the position of the non-controlled object calculated by the non-controlled object position calculation unit, and the safety standard determined by the safety standard determination unit so that the controlled object does not approach the non-controlled object as the safety standard becomes higher. [Effects of the Invention]

[0011] According to the present invention, it is possible to provide an autonomous control system that can drive a vehicle efficiently and safely even when it is difficult to predict the behavior of surrounding moving objects. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a functional block diagram of an autonomous control system according to a first embodiment of the present invention. [Figure 2] FIG. 2 is a schematic diagram showing a vehicle that is a controlled object and a pedestrian that is not a controlled object of the autonomous control system. [Figure 3] FIG. 3 is a functional block diagram of the trajectory deviation evaluation unit. [Figure 4] FIG. 4 is a schematic diagram showing an example of a predicted movement trajectory calculated by the predicted trajectory calculation unit. [Figure 5] FIG. 5 is a schematic diagram showing an example of a predicted trajectory and an actual trajectory. [Figure 6] FIG. 6 is a diagram illustrating a method for determining a safety standard by the safety standard determination unit according to the first embodiment. [Figure 7] FIG. 7 is a schematic diagram illustrating the motion control of a four-wheel vehicle. [Figure 8] FIG. 8 is a schematic diagram showing the positional relationship between controlled and non-controlled objects. [Figure 9]FIG. 9 is a flowchart showing an example of the flow of processing executed by the controller of the autonomous control system. [Figure 10] FIG. 10 is a schematic diagram showing a logistics warehouse to which the autonomous control system according to the second embodiment is applied. [Figure 11] FIG. 11 is a schematic diagram illustrating the motion control of the transport robot. [Figure 12] FIG. 12 is a diagram illustrating a method for determining a safety standard by the safety standard determination unit according to the second embodiment. [Figure 13] FIG. 13 is a diagram showing an example of the relationship between the function h(ε(k)) and the function ε(k). [Figure 14] FIG. 14 is a diagram showing the configuration of an autonomous control system according to Modification 1, and shows a vehicle as a moving body to be controlled and a pedestrian as a moving body not to be controlled. [Figure 15] FIG. 15 is a diagram illustrating a method for determining a safety standard and a priority of the host vehicle by a safety standard determination unit according to the second modification. DETAILED DESCRIPTION OF THE INVENTION

[0013] (First embodiment) FIG. 1 is a functional block diagram of an autonomous control system according to a first embodiment of the present invention. The autonomous control system A100 and motion control system B100 shown in FIG. 1 are mounted on a vehicle to be controlled. The autonomous control system A100 collects information about the surroundings of the vehicle to be controlled in an area where controlled and non-controlled vehicles coexist, and causes the motion control system B100 to control the motion (behavior) of the vehicle to be controlled so that the controlled vehicle does not come into contact with moving objects other than the controlled vehicle (e.g., pedestrians, other vehicles, etc.). In FIG. 1, parts that are not directly related to the function of the autonomous control system A100 according to this embodiment are not shown.

[0014] The vehicle to be controlled is not limited to a fully autonomous vehicle. For example, it may be a semi-autonomous vehicle that is normally driven by a driver and in which the autonomous control system A100 can intervene to perform control such as deceleration or stopping only in an emergency. The vehicle to be controlled may also be a single vehicle traveling on a public road, or a vehicle (robot) traveling within a logistics warehouse.

[0015] 2 is a schematic diagram showing a vehicle 001, which is a moving object to be controlled, and a pedestrian 002, which is a moving object that is not to be controlled. The following mainly describes an example in which the vehicle 001 traveling on a roadway 10 is the moving object to be controlled by the autonomous control system A100, and the pedestrian 002 walking on a sidewalk 20 is the moving object that is not to be controlled. In other words, the vehicle 001 is a moving object whose behavior can be controlled by the autonomous control system A100, and the pedestrian 002 is a moving object whose behavior cannot be controlled by the autonomous control system A100. Note that, for simplicity of explanation, a situation is shown in which there is one controlled object and one non-controlled object, but the present invention can also be used in cases in which there are multiple controlled objects and multiple non-controlled objects.

[0016] 1, the motion control system B100 includes an actuator controller B001 and an actuator B002. The actuator controller B001 controls the actuator B002 in accordance with instructions from the autonomous control system A100. The actuator B002 is connected to, for example, the steering, accelerator, brake, etc. of the vehicle.

[0017] 1, the autonomous control system A100 includes an environment recognizer A001, a state detector A002, a controlled object position calculation unit A003, an object identification unit A004, a non-controlled object position calculation unit A005, a trajectory deviation evaluation unit A006, a safety standard determination unit A007, and a vehicle motion calculation unit A008. The controlled object position calculation unit A003, the object identification unit A004, the non-controlled object position calculation unit A005, the trajectory deviation evaluation unit A006, the safety standard determination unit A007, and the vehicle motion calculation unit A008 are functions realized by the controller A101.

[0018] The controller A101 is configured, for example, by a computer equipped with a processing device such as a central processing unit (CPU), a read-only memory (ROM), non-volatile memory such as flash memory, a volatile memory called random access memory (RAM), an input / output interface, and other peripheral circuits. These pieces of hardware work together to run software and realize multiple functions. The controller A101 may be configured by one computer or multiple computers.

[0019] The nonvolatile memory stores programs capable of executing various calculations and data such as threshold values. In other words, the nonvolatile memory is a storage medium (storage device) from which the programs that realize the functions of this embodiment can be read. The volatile memory is a storage medium (storage device) that temporarily stores the results of calculations performed by the processing device and signals input from the input / output interface. The processing device is a device that loads the programs stored in the nonvolatile memory into the volatile memory and executes the calculations, and performs predetermined calculations on data taken from the input / output interface, the nonvolatile memory, and the volatile memory in accordance with the programs.

[0020] The autonomous control system A100 does not have to be implemented in the controlled object (i.e., vehicle 001). As in the second embodiment described below, if the area in which the controlled object moves is limited, it is also possible to provide the calculation function to a server that can communicate within that area.

[0021] The environment recognizer A001 acquires environmental information that represents the surrounding state of the controlled object. The environment recognizer A001 is, for example, an external recognition sensor such as a LiDAR (Light Detection and Ranging) sensor, a stereo camera, or a millimeter-wave radar mounted on the controlled object.

[0022] The state detector A002 acquires vehicle information (position, direction, speed, etc.) that indicates the state of the controlled object. The state detector A002 is, for example, a sensor such as a GNSS (Global Navigation Satellite System) receiver that acquires position information of the controlled object, or an IMU (Inertial Measurement Unit) that acquires the acceleration and angular velocity of the controlled object.

[0023] The environment recognizer A001 and the state detector A002 are not necessarily separate sensors. For example, a LiDAR sensor mounted on the controlled object functions as both the environment recognizer A001 and the state detector A002.

[0024] The control object position calculation unit A003 calculates the position of the control object by integrating the vehicle information acquired by the state detector A002. For example, if the state detector A002 is a LiDAR, the control object position calculation unit A003 estimates the position of the control object using a well-known SLAM (Simultaneous Localization and Mapping) technique. Also, if the state detector A002 is a GNSS receiver and an IMU, the control object position calculation unit A003 calculates the position of the control object by interpolating the update period of the position information output from the GNSS receiver with the IMU using a well-known sensor fusion technique.

[0025] The object identification unit A004 identifies attributes of non-controlled objects existing around the controlled object from the environmental information acquired by the environment recognizer A001 using well-known image recognition technology or Semantic SLAM technology.

[0026] The environment recognizer A001 and the object identification unit A004 may be integrated. For example, a stereo camera or a millimeter-wave radar may have a function for identifying moving objects. In this case, it is not necessary to explicitly separate the environment recognizer A001 and the object identification unit A004.

[0027] The object identification unit A004 identifies the attributes of the non-controllable object, such as characteristics related to the movement of a moving object, such as a pedestrian or a bicycle. Here, the characteristics related to movement refer to the equation of motion and maximum moving speed that govern the dynamic characteristics of the moving object. Pedestrians can move freely on a two-dimensional plane. On the other hand, vehicles such as wheelchairs, bicycles, scooters, and automobiles include holonomic constraints, such as not being able to move directly sideways. 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 more finely.

[0028] Similar to the object identification unit A004, the non-controlled object position calculation unit A005 calculates the positions of non-controlled objects present around the controlled object using well-known Semantic SLAM technology from the environmental information acquired by the environment recognizer A001. The object identification unit A004 and the non-controlled object position calculation unit A005 can execute processing simultaneously using the same technology. In other words, the identification process by the object identification unit A004 and the position calculation process by the non-controlled object position calculation unit A005 can be processed in parallel.

[0029] 3 is a functional block diagram of the trajectory deviation evaluation unit A006. The trajectory deviation evaluation unit A006 includes a predicted trajectory calculation unit A006a, an actual trajectory evaluation unit A006b, and a trajectory comparison unit A006c.

[0030] The trajectory deviation evaluation unit A006 evaluates the degree of deviation of the actual movement trajectory (hereinafter also referred to as actual trajectory) of the non-control object from the predicted movement trajectory of the non-control object corresponding to the attributes of the non-control object. The predicted trajectory calculation unit A006a calculates the predicted movement trajectory of the non-control object based on the attributes of the non-control object identified by the object identification unit A004 and the position of the non-control object calculated a predetermined time ago by the non-control object position calculation unit A005. For example, if the non-control object is a pedestrian, the position of the non-control object several seconds after its current position is calculated according to the equations of motion of Equations (3a) and (3b) described below. Also, if the non-control object is a vehicle, the position of the non-control object several seconds after its current position is calculated according to the equations of motion of Equations (2a) and (2b) described below. The actual trajectory evaluation unit A006b evaluates the change in the position of the non-control object calculated by the non-control object position calculation unit A005 as the actual movement trajectory of the non-control object. The trajectory comparison unit A006c compares the predicted movement trajectory of the non-controlled object calculated by the predicted trajectory calculation unit A006a with the movement trajectory (actual trajectory) of the non-controlled object evaluated by the actual trajectory evaluation unit A006b, and evaluates the degree of deviation of the movement trajectory (actual trajectory) from the predicted movement trajectory.

[0031] Fig. 4 is a schematic diagram showing an example of a predicted movement trajectory calculated by the predicted trajectory calculation unit A006a. Fig. 4 shows a schematic diagram of the predicted movement trajectories (hereinafter simply referred to as predicted trajectories) calculated by the predicted trajectory calculation unit A006a for a pedestrian 002 and a bicycle 005.

[0032] When pedestrian 002 is at position P0 at time t0 and facing leftward on the paper, the predicted trajectory calculation unit A006a predicts the position of pedestrian 002 as follows: position P1 at time t1, one step later, position P2 at time t2, two steps later, position P3 at time t3, three steps later, position P4 at time t4, four steps later, and position P5 at time t5, five steps later. The predicted trajectory calculation unit A006a calculates the predicted trajectory shown in FIG. 4 by prediction based on a probabilistic model. That is, the predicted trajectory calculation unit A006a calculates the average value and variance of the trajectory of the non-control object and sets them as the predicted trajectory. Positions P1 to P5 shown in FIG. 4 indicate the average value of the predicted trajectory, and the upper limit 40 and lower limit 30 shown are determined by the variance 50 of the predicted trajectory. As shown in FIG. 4, the predicted trajectory has a certain spread determined by the variance 50. In general, since pedestrians can move freely not only straight but also diagonally and sideways, the range of possible trajectories becomes wider as time progresses.

[0033] When bicycle 005 is facing leftward on the paper at position P10 at time t0, predicted trajectory calculation unit A006a predicts the position of bicycle 005 to be position P11 at time t1, one step later, position P12 at time t2, two steps later, position P13 at time t3, three steps later, position P14 at time t4, four steps later, and position P15 at time t5, five steps later. Because bicycle 005 has more difficulty moving left and right than pedestrian 002, the candidate range defined by upper limit 41 and lower limit 31 in the vertical direction on the paper does not expand as much as that of pedestrian 002. In other words, variance 51 of the predicted trajectory of bicycle 005 is smaller than variance 50 of the predicted trajectory of pedestrian 002. However, bicycle 005 is characterized by its faster movement speed than pedestrian 002, and therefore the candidate range expands left and right on the paper. That is, the intervals between the positions P10 to P15 of the bicycle 005 are wider than the intervals between the positions P0 to P5 of the pedestrian 002.

[0034] The actual trajectory evaluation unit A006b shown in FIG. 3 evaluates the actual movement trajectory (actual trajectory) of the non-control object by recording the positions of the non-control object calculated by the non-control object position calculation unit A005 for a predetermined period of time.

[0035] The trajectory comparison unit A006c compares the predicted trajectory calculated by the predicted trajectory calculation unit A006a with the actual trajectory evaluated by the actual trajectory evaluation unit A006b to evaluate the behavior index (behavior orientation) of the non-controlled object. The trajectory comparison unit A006c calculates an evaluation value using the following equation (1). In equation (1), k represents the time of the predicted trajectory, z(k) represents the coordinate of the predicted trajectory at time k, x(k) represents the coordinate of the actual trajectory at time k, and e(k) represents the trajectory deviation, which is the difference between the predicted trajectory and the actual trajectory at time k. The trajectory comparison unit A006c uses the trajectory deviation e calculated using equation (1) as an evaluation value that represents the degree of deviation of the actual trajectory from the predicted trajectory.

[0036]

number

[0037] 5A and 5B are schematic diagrams showing examples of predicted trajectories and actual trajectories. Fig. 5A shows a candidate range 60a of the predicted trajectory of a pedestrian 002 (i.e., a non-control target) calculated by the predicted trajectory calculation unit A006a at a certain time. In this example, a situation is assumed in which the pedestrian is moving upward on the paper (in the direction indicated by the arrow 61).

[0038] Figure 5(b) shows an example of an actual trajectory 62b of pedestrian 002 superimposed on a candidate range 60b of the predicted trajectory of pedestrian 002. The actual trajectory 62b of pedestrian 002 coincides with the center (average) of the candidate range 60b of the predicted trajectory. In other words, it can be determined that the behavior of this pedestrian 002 is easy to predict.

[0039] 5(c) shows an example of an actual trajectory 62c of pedestrian 002 superimposed on a candidate range 60c of the predicted trajectory of pedestrian 002. This actual trajectory 62c deviates from the candidate range 60c of the predicted trajectory. This type of behavior of pedestrian 002 may occur, for example, when pedestrian 002 is drunk, and therefore, it can be said that the behavior of pedestrian 002 is difficult to predict.

[0040] The safety standard determination unit A007 shown in FIG. 3 determines safety standards for the behavior of the controlled object based on the attributes of the non-controlled object identified by the object identification unit A004 and the degree of deviation from the predicted movement trajectory (trajectory deviation e) evaluated by the trajectory deviation evaluation unit A006. The safety standard determination unit A007 uses two types of safety standards: "standard," which is a standard safety standard, and "safest," which is the safety standard with the highest level of safety. These safety standards each affect the motion control of the controlled object by the vehicle motion calculation unit A008 (see FIG. 1). As will be described in detail later, when the safety standard is determined to be "safest," the vehicle motion calculation unit A008 performs motion control with a larger margin of distance between the non-controlled object and the controlled object compared to when the safety standard is determined to be "standard." In addition to these two types, the vehicle motion calculation unit A008 also continuously uses an intermediate safety standard that is safer than "standard" but lower than "safest."

[0041] FIG. 6 is a diagram illustrating how the safety standard determination unit A007 determines the safety standard. The safety standard determination unit A007 adjusts the control content to ensure safe motion control of the controlled object as the trajectory deviation e calculated by the trajectory comparison unit A006c increases. In FIG. 6, the safety standard is set to "standard" when the trajectory deviation e is equal to or less than the threshold th1, and set to "safest" when the trajectory deviation e is equal to or greater than the threshold th2. Note that the threshold th2 is greater than the threshold th1. These thresholds th1 and th2 are preferably designed according to the variance of the predicted trajectory calculated by the predicted trajectory calculation unit A006a. For example, using the standard deviation σ, which is the square root of the variance, makes it easy to determine the probability that the trajectory deviation e will occur relative to the prediction. In other words, if th1 is set to σ, control can be designed to respond to situations where the trajectory deviates from the trajectory that should be approximately 68% consistent. If th2 is set to 2σ, control can be designed to respond to situations where the trajectory deviates from the trajectory that should be approximately 95% consistent.

[0042] In Figure 6, two thresholds, th1 and th2, are provided and the safety standard is configured continuously, but the safety standard is not limited to this format. For example, it is also possible to set only one threshold and determine the safety standard to be "standard" when the trajectory deviation e is less than the threshold, and to determine the safety standard to be "safest" when the trajectory deviation e is equal to or greater than the threshold.

[0043] As described above, the controller A101 according to this embodiment shown in FIG. 1 detects a non-controlled object (moving object, pedestrian) a predetermined time (predetermined steps) beforehand, calculates a predicted trajectory, and then obtains an actual trajectory as the movement record for the predetermined time (predetermined steps), and compares the predicted trajectory with the actual trajectory. In other words, it is important to note that the timing at which a non-controlled object (moving object, pedestrian) is detected differs from the timing at which control becomes feasible. When safety is emphasized, it is desirable that the controller A101 sets the trajectory deviation e to an initial value equal to or greater than the threshold value th2 when a non-controlled object (moving object, pedestrian) is detected, and then changes the trajectory deviation e from the initial value when it becomes possible to actually calculate the trajectory deviation e.

[0044] The vehicle motion calculation unit A008 controls the motion of the controlled object based on the position of the controlled object calculated by the controlled object position calculation unit A003, the position of the non-controlled object calculated by the non-controlled object position calculation unit A005, and the safety standard determined by the safety standard determination unit A007.

[0045] If the controlled object is a semi-automated vehicle, the vehicle motion calculation unit A008 instructs the actuator controller B001 to interfere with vehicle speed control and slow down or stop the vehicle when the safety standard becomes higher than "standard." The actuator controller B001 follows this instruction and drives the actuator B002 connected to the brake. On the other hand, if the safety standard is lower than "standard," there is no need to interfere with the driver's operation.

[0046] If the vehicle to be controlled is an autonomous vehicle, the vehicle motion calculation unit A008 also uses control functions related to the overall vehicle motion.

[0047] FIG. 7 is a schematic diagram illustrating motion control of a four-wheeled vehicle. When the controlled object is four-wheeled vehicle 006 shown in FIG. 7, considering a vector p = [xy θ] that combines the coordinates (x, y) of the vehicle's position and the azimuth angle θ, simplified dynamics can be given by the following equations (2a) and (2b). The azimuth angle θ corresponds to the angle between the reference direction and the axis extending in the longitudinal direction of the vehicle. In equation (2a), L is the distance between the front and rear wheels of the vehicle, as shown in FIG. 7. Furthermore, the control input uc in equations (2a) and (2b) is the vehicle speed v and the steering angle φ.

[0048]

number

[0049] On the other hand, if the uncontrolled object is a pedestrian, and we consider the vector q = [xu yu] that summarizes the pedestrian's position coordinates (xu, yu), the simplified dynamics of the uncontrolled object can be given by the following equations (3a) and (3b). Equations (3a) and (3b) assume that the uncontrolled object is a pedestrian, and assume a model in which the position coordinates can move freely in either the x or y direction. The control input uu in equations (3a) and (3b) is the x-component vx and y-component vy of the velocity v of the uncontrolled object. If the uncontrolled object is a vehicle, the same equations of motion as equations (2a) and (2b) will be used.

[0050]

number

[0051] The differential equations (2a) and (2b) and (3a) and (3b) above can be discretized using the sampling period Δt as shown in the following equations (4a) and (4b). Here, p(k) is the state vector of the controlled object at time k (computation step k), and p(k+1) is the state vector of the controlled object at the next time k+1 (next computation step k+1). Similarly, q(k) is the state vector of the uncontrolled object (pedestrian) at time k (computation step k), and q(k+1) is the state vector of the uncontrolled object (pedestrian) at the next time k+1 (next computation step k+1).

[0052]

number

[0053] First, we will explain the control performed by the vehicle motion calculation unit A008 when the safety standard is "standard," that is, when the non-controlled object is within the range of the predicted trajectory. The movement of the controlled object can be predicted according to the above formula (4a). Furthermore, the movement of the non-controlled object can be predicted according to the above formula (4b). Assume that the position (xu, yu) and velocity (vx, vy) of the non-controlled object at the current time k0 are available from the environment recognizer A001 and the non-controlled object position calculation unit A005. In this case, the variance of the position of the non-controlled object changes according to the following formula (5). Here, P is the variance matrix of the position of the non-controlled object, and Qw is the variance of the process noise.

[0054]

number

[0055] Also, assume that the environment recognizer A001 has acquired the approximate size of the uncontrolled object and can surround the uncontrolled object with a circle of radius r. If the size (radius r) of the uncontrolled object cannot be acquired, the object identification unit A004 may set the radius r according to the attribute (pedestrian, bicycle, etc.) of the identified uncontrolled object. For example, the object identification unit A004 sets the radius r to 1.5 meters if the attribute of the uncontrolled object is pedestrian, and sets the radius r to 2 meters if the attribute of the uncontrolled object is bicycle.

[0056] On the other hand, if the total length of the controlled object is lc and the total width is wc, the controlled object can be surrounded by a circle with a radius rc as shown in the following equation (6).

[0057]

number

[0058] Figure 8 is a schematic diagram showing the positional relationship between the controlled object and the uncontrolled object. Based on the above assumptions, the uncontrolled object and the controlled object will not come into contact with each other as long as the positional relationship shown in Figure 8 is satisfied, that is, as long as the following equations (7a), (7b), and (7c) are satisfied. Note that dpq in equations (7a) and (7b) corresponds to the length of the line segment connecting the center of the controlled object and the center of the uncontrolled object, i.e., the distance between the controlled object and the uncontrolled object. Furthermore, ra in equation (7c) is the increase in radius that takes into account the uncertainty in the predicted trajectory of the uncontrolled object, and corresponds to the standard deviation calculated from the variance matrix P of the position of the uncontrolled object. α is an adjustment parameter (greater than 0). For example, if α = 2, the design will avoid contact with a 95% probability, assuming the predicted trajectory is correct.

[0059]

number

[0060] To automatically control the controlled object, the vehicle position can be controlled so that it follows the target trajectory pr = [xr yr θr]. For example, for a vehicle traveling on a public road, the target trajectory can be set at the center of the road a few meters ahead of the current vehicle position.

[0061] It is suitable to use the concept of model predictive control for such tracking control. In other words, it can be formulated as a control problem to find a control input uc so as to minimize the evaluation function J in the following equation (8) under the constraint that constraints (7a), (7b), and (7c) are satisfied. Note that Q and R are weighting matrices, and Np is the prediction step. Since the above is a general formulation of model predictive control, a detailed explanation will be omitted.

[0062]

number

[0063] Next, the control content by the vehicle motion calculation unit A008 when the safety standard is lower than the "standard", that is, when the non-controlled object deviates from the predicted trajectory but follows a safe trajectory, will be described.

[0064] In such a situation, safety can be maintained even if the vehicle to be controlled increases its speed. Therefore, the weight Q in the evaluation function (Equation (8)) used in model predictive control can be increased to improve the ability of the controlled vehicle to track the target trajectory. Alternatively, the weight R can be decreased to allow for the generation of a larger control input uc.

[0065] When the safety standard is lower than the "standard" level, the same control content as when the safety standard is the "standard" level may be applied.

[0066] Next, the control contents by the vehicle motion calculation unit A008 when the safety standard is higher than the "standard", that is, when the non-controlled object deviates from the predicted trajectory and is on a dangerous trajectory, will be described.

[0067] Although the behavior of the uncontrolled object cannot be predicted, it is assumed that its dynamics follow equation (4b). Furthermore, it is assumed that the object identification unit A004 is able to identify the attributes of the uncontrolled object (e.g., pedestrian, bicycle, etc.). The vehicle motion calculation unit A008 sets the maximum moving speed of the uncontrolled object according to this attribute. For example, if the uncontrolled object belongs to a pedestrian, the maximum moving speed is set to 5 kilometers per hour, and if it belongs to a bicycle, the maximum moving speed is set to 20 kilometers per hour. If the attributes of the uncontrolled object cannot be identified, the vehicle motion calculation unit A008 sets the maximum moving speed to an upper limit value according to the location where the uncontrolled object is located. For example, if the uncontrolled object is on a sidewalk, the maximum moving speed is set to 20 kilometers per hour, which is equivalent to a bicycle. Furthermore, if the uncontrolled object is on a roadway, the maximum moving speed is set to 60 kilometers per hour, which is equivalent to a general vehicle.

[0068] When safety standards are higher than the standard, model predictive control is considered, which calculates not only the control input u of the controlled object but also the control input uu of the uncontrolled object. However, since the uncontrolled object cannot be controlled, the control input uu of the uncontrolled object calculated by model predictive control is not actually used.

[0069] When the safety standard is higher than the standard, the following equation (9a) is given as an example of an evaluation function to be handled by model predictive control. The evaluation function J' in equation (9a) is formed by adding the evaluation function Jp expressed in equation (9b) to the evaluation function J expressed in equation (8). Here, Qu and Ru are weighting matrices.

[0070]

number

[0071] The effect of applying the above equation (9b) is explained below. First, the first term in equation (9b) is the product of the distance dpq between the controlled object and the uncontrolled object multiplied by the weighting matrix Qu. In other words, minimizing the evaluation function Jp means minimizing the distance dpq between the controlled object and the uncontrolled object. The second term is the product of the control input uu (velocity v of the uncontrolled object) of the uncontrolled object multiplied by the weighting matrix Ru, and this value is then multiplied by -1 to make it negative. In other words, minimizing the evaluation function Jp means maximizing the control input uu (velocity v of the uncontrolled object) of the uncontrolled object. Combining these effects, the newly added evaluation function Jp is a problem design that assumes that the uncontrolled object approaches the controlled object at the fastest speed. Under this problem design, by calculating the control input uc of the controlled object while taking into account the constraints of equations (7a), (7b), and (7c), it is possible to generate commands to execute safe movements that avoid the possibility of the uncontrolled object engaging in dangerous behavior.

[0072] As mentioned above, this control operation is also adopted when it is not possible to compare the predicted trajectory with the actual trajectory immediately after detecting a non-controlled object. However, if the detection range of the environment recognizer A001 is sufficiently wide, the controlled object and the non-controlled object are sufficiently far apart, so even if equation (9b) is taken into account, behavior that deviates from constraints (7a), (7b), and (7c) is unlikely to occur. In other words, when the safety standard is "standard," behavior similar to that generated when equation (8) is used is often generated.

[0073] When the safety standard determination unit A007 determines the safety standard discretely using only one threshold, the vehicle motion calculation unit A008 can switch between model predictive control that minimizes the evaluation function J expressed by equation (8) and model predictive control that minimizes the evaluation function J' expressed by equation (9a). When the safety standard determination unit A007 gradually increases the safety standard from a situation where the track deviation e exceeds the threshold th1, as shown in FIG. 6, the vehicle motion calculation unit A008 can use model predictive control that minimizes the evaluation function J' expressed by the following equation (10) using the adjustment variable β (0≦β≦1). That is, when the track deviation e is equal to or less than the threshold th1, if β=0, then equation (10) coincides with equation (8). On the other hand, when the track deviation e is equal to or greater than the threshold th2, if β=1, then equation (10) coincides with equation (9a). Furthermore, when the trajectory deviation e is greater than the threshold th1 and less than the threshold th2, the adjustment variable β is increased as the trajectory deviation e increases, thereby achieving behavior intermediate between equation (8) and equation (9a).

[0074]

number

[0075] An actuator controller B001 of the motion control system B100 controls an actuator B002 to change the accelerator opening and steering so as to realize the calculated control input uc, that is, the vehicle speed v and the steering angle φ.

[0076] FIG. 9 is a flowchart showing an example of the flow of processing executed by the controller A101 of the autonomous control system. The processing shown in FIG. 9 is repeatedly executed at a predetermined control period. As shown in FIG. 9, in step S01, the controller A101 acquires detection values ​​from the environment recognizer A001 and the state detector A002 and updates the detection values ​​stored in memory. In step S02, the controller A101 determines whether the environment recognizer A001 has detected a moving object that is an uncontrolled object. If the environment recognizer A001 has detected an uncontrolled object, the processing proceeds to step S03. On the other hand, if the environment recognizer A001 has not detected an uncontrolled object, the processing proceeds to step S10. In this case, the vehicle motion calculation unit A008 performs a control operation that takes into account only the controlled object. In other words, model predictive control is performed to minimize equation (8) without considering constraints (7a), (7b), and (7c).

[0077] In step S03, the object identification unit A004 identifies non-control objects based on the environmental information acquired from the environment recognizer A001. The object identification unit A004 determines the movement speed limit (maximum value of control input uu) and size (radius r) of the non-control objects according to the identification results. In step S04, the predicted trajectory calculation unit A006a calculates the predicted trajectory of the non-control objects according to the attributes of the non-control objects identified in step S03.

[0078] In step S05, the actual trajectory evaluation unit A006b calculates the movement trajectory (actual trajectory) actually taken by the non-controlled object. In step S06, the trajectory comparison unit A006c determines whether a predetermined time has passed since the time the non-controlled object was detected in step S02. If the actual trajectory is too short, it is not possible to compare the predicted trajectory with the actual trajectory, and therefore such a determination is made in step S05. If the predetermined time has not passed, the process proceeds to step S07. On the other hand, if the predetermined time has passed, the process proceeds to step S08. If the process proceeds to step S07, the actual orbit is too short to compare the predicted orbit with the actual orbit, so the orbit comparison unit A006c sets the orbit deviation e to a predetermined initial value (a value equal to or greater than threshold th2).If the process proceeds to step S08, the orbit comparison unit A006c compares the predicted orbit with the actual orbit to calculate the orbit deviation e.

[0079] In step S09, the safety standard determination unit A007 determines a safety standard and a model predictive control method to be used in accordance with the trajectory deviation e set in step S07 or calculated in step S08. In step S10, the vehicle motion calculation unit A008 calculates a control input uc for the vehicle to be controlled in accordance with the model predictive control method determined in step S09. In step S11, the vehicle motion calculation unit A008 issues a control instruction to the actuator controller B001 to realize the control input uc calculated in step S10. The actuator controller B001 controls each actuator B002 in accordance with this control instruction.

[0080] According to the above-described embodiment, the following advantageous effects are achieved.

[0081] (1) The trajectory deviation evaluation unit A006 evaluates the degree of deviation of the movement trajectory of the non-controlled object from the predicted movement trajectory of the non-controlled object corresponding to the attributes of the non-controlled object. The safety standard determination unit A007 determines the safety standard for the behavior of the controlled object based on the attributes of the non-controlled object and the degree of deviation from the predicted movement trajectory. The vehicle motion calculation unit A008 (behavior modification unit) modifies the behavior of the controlled object based on the position of the controlled object, the position of the non-controlled object, and the safety standard so that the controlled object does not approach the non-controlled object as the safety standard becomes higher. This makes it possible to drive an autonomous or semi-autonomous vehicle efficiently and safely even when it is difficult to predict the behavior of surrounding moving objects.

[0082] (2) The predicted trajectory calculation unit A006a calculates the average value and variance of the trajectory based on the attributes of the non-controlled object and the position of the non-controlled object calculated a predetermined time ago, and sets this as the predicted movement trajectory of the non-controlled object. The actual trajectory evaluation unit A006b evaluates the change in the position of the non-controlled object as the movement trajectory of the non-controlled object. The trajectory comparison unit A006c evaluates the degree of deviation of the movement trajectory of the non-controlled object from the predicted movement trajectory of the non-controlled object. In this way, the range of the predicted movement trajectory can be accurately determined according to the attributes of the non-controlled object.

[0083] (3) The predicted trajectory calculation unit A006a calculates the predicted trajectory of the non-controlled object based on the attributes of the non-controlled object, using the equation of motion that the non-controlled object follows, the maximum value of the movement speed of the non-controlled object, and the size of the non-controlled object. This makes it possible to accurately calculate the predicted trajectory according to the attributes of the non-controlled object.

[0084] (Second embodiment) FIG. 10 is a schematic diagram showing a logistics warehouse to which an autonomous control system according to the second embodiment is applied. A total of three transport robots 101a, 101b, and 101c are arranged in a warehouse 100. The transport robots 101a, 101b, and 101c perform transport work together with workers. In this embodiment, the transport robots 101a, 101b, and 101c are controlled objects, and pedestrians 102a and 102b, such as workers, are not controlled objects. In the following description, the transport robots 101a, 101b, and 101c are collectively referred to as transport robots 101. Similarly, the pedestrians 102a and 102b are collectively referred to as pedestrians 102.

[0085] A plurality of infrastructure sensors 103 are provided within the warehouse 100. The plurality of infrastructure sensors 103 monitor moving objects present within the warehouse 100. Monitoring information from the plurality of infrastructure sensors 103 is transmitted to a server 105 via a wireless communication access point 104. The server 105 has a function of calculating a movement control plan for the transport robot 101 within the warehouse 100 and is responsible for the calculation function of the autonomous control system according to the second embodiment. The server 105 has the same hardware (processing device, volatile memory, non-volatile memory, etc.) as the controller A101 described in the first embodiment. The server 105 and the transport robot 101 communicate bidirectionally. That is, the server 105 receives sensor information acquired by the transport robot 101, and the transport robot 101 receives a behavior plan from the server 105. The transport robot 101 is equipped with sensors equivalent to the environment recognizer A001 and the state detector A002, such as a LiDAR sensor, an IMU, and an encoder.

[0086] 11 is a schematic diagram illustrating the motion control of the transport robot 101. The transport robot 101 is driven by a differential two-wheel system. When the coordinates (xi, yi) of the position of the transport robot 101 and the azimuth angle θi are given as a vector pi = [xi yi θi] using the subscript i corresponding to the ID assigned to each transport robot 101, the transport robot 101 follows the equation of motion shown in the following equation (11). In equation (11), vi is the speed of the transport robot 101, ωi is the angular velocity of the transport robot 101, and uc,i is the control input of the transport robot 101.

[0087]

number

[0088] An ID is also assigned to the pedestrians 102 in the order in which they are recognized by the infrastructure sensor 103. When the coordinates ((xu,j), (yu,j)) of the position of the pedestrian 102 are given as a vector qj=[xu,j yu,j] using the subscript j corresponding to each ID, the pedestrian 102 follows the equation of motion of the following equation (12). In equation (12), vx,i is the x-direction component of the velocity v of the pedestrian 102, vy,j is the y-direction component of the velocity v of the pedestrian 102, and uu,j is the control input for the pedestrian 102.

[0089]

number

[0090] As in the first embodiment described above, when the infrastructure sensor 103 corresponding to the environment recognizer A001 or the LiDAR sensor mounted on the transport robot 101 detects an uncontrolled object (pedestrian), the object identification unit A004 identifies the attributes (movement characteristics) of the uncontrolled object.

[0091] Unlike general public roads, in limited areas such as logistics warehouses, the behavior of robots and vehicles (including forklifts operated by humans) moving within the warehouse is managed by the WMS (Warehouse Management System) that operates the warehouse. For this reason, the attributes (movement characteristics) of uncontrolled objects can be identified by linking the positions of uncontrolled objects detected by controlled objects with the operational status of vehicles managed by the WMS.

[0092] Similarly, in a limited area such as a logistics warehouse, the majority of uncontrolled objects that cannot be matched by the attribute identification using the WMS described above will be classified as pedestrians. However, even with the same pedestrian classification, in a limited area it is possible to perform more detailed classification.

[0093] For example, employees with longer tenure are generally more familiar with warehouse operations and less likely to engage in risky behavior. Similarly, part-time and temporary workers are less familiar with warehouse operations and are more likely to engage in risky behavior. Furthermore, members of the public who visit the warehouse have never seen the Transport Robot 101, so they are likely to engage in even more risky behavior.

[0094] It is difficult to infer the above classification from images or point cloud information acquired by the infrastructure sensor 103 or LiDAR sensor. However, since entry into a restricted area is not possible without an employee ID or guest card, the server 105 can acquire the attributes of the card holder by communicating with this card. The object identification unit A004 identifies the attributes of the pedestrian 102, which is a non-controlled object, based on the attribute information received from the card held by the pedestrian 102.

[0095] Fig. 12 is a diagram illustrating a method for determining safety standards by the safety standard determination unit A007 according to the second embodiment. According to the above-described classification, the safety standard determination unit A007 can be subdivided as shown in Fig. 12. In other words, even if the values ​​of the trajectory deviation e between the predicted trajectory and the actual trajectory calculated by the trajectory deviation evaluation unit A006 are the same, different safety standards are determined.

[0096] For example, for a pedestrian 102 whose attribute is identified by the object identification unit A004 as either a new employee, a part-time worker, or a casual worker, the safety standard determination unit A007 adopts the standard safety standard characteristics 70. The standard safety standard characteristics 70 are similar to the characteristics shown in FIG. 6, and therefore a description thereof will be omitted.

[0097] The safety standard determination unit A007 adopts the veteran safety standard characteristic 71 for a pedestrian 102 whose attribute is identified by the object identification unit A004 as a long-serving employee. Long-serving employees are more safety-conscious than new employees, part-timers, and casual workers, and therefore are expected to avoid a collision with the transport robot 101 even if the safety standard is not raised to "highest safety." Therefore, the veteran safety standard characteristic 71 is a characteristic that sets the safety standard to "high safety," which is between "highest safety" and "standard," when the trajectory deviation e is equal to or greater than the threshold th2. The veteran safety standard characteristic 71 is a characteristic that increases the safety standard as the trajectory deviation e increases within the range where the trajectory deviation e is greater than the threshold th1 and less than the threshold th2. The veteran safety standard characteristic 71 is a characteristic that sets the safety standard to "standard" when the trajectory deviation e is equal to or less than the threshold th1, similar to the standard safety standard characteristic 70.

[0098] The safety standard determination unit A007 adopts a guest safety standard characteristic 72 for a pedestrian 102 whose attribute is identified as guest by the object identification unit A004. For a guest card holder who is likely not safety-conscious, it is preferable to set the safety standard higher than "standard" even if the predicted trajectory and the actual trajectory match. This makes it easier to avoid contact between the pedestrian (guest) 102 and the transport robot 101 even if the pedestrian (guest) 102 suddenly approaches the transport robot 101. The guest safety standard characteristic 72 is a characteristic that sets the safety standard to "high safety," which is between "standard" and "safest," when the trajectory deviation e is greater than or equal to 0 and less than a threshold value th1. Furthermore, the guest safety standard characteristic 72 is a characteristic that sets the safety standard to "high safety," which is between "standard" and "safest," when the trajectory deviation e is greater than or equal to 0 and less than a threshold value th2, as the trajectory deviation e increases, and sets the safety standard to "safest" when the trajectory deviation e is greater than or equal to a value intermediate between the threshold values ​​th1 and th2.

[0099] The following describes the control calculations performed by the vehicle motion calculation unit A008 when there are multiple controlled objects and multiple non-controlled objects.

[0100] First, the control content by the vehicle motion calculation unit A008 when the safety standard determination unit A007 determines the safety standard to be "standard" will be described. The conditions for preventing contact between the transfer robot 101 and the pedestrian 102 are expressed by the following equations (13a), (13b), (13c), and (13d). The radius rc,i surrounding the transfer robot 101(i) with ID i can be given by equation (13a). Furthermore, the radius rj surrounding the pedestrian 102(j) with ID j is calculated according to the detection result of the environment recognizer A001. Furthermore, the increase in radius ra,j taking into account the uncertainty in the predicted trajectory of the pedestrian 102(j) can be given by equation (13b). The distance dij between the transfer robot 101(i) and the pedestrian 102(j) can be given by equation (13c). With the above preparations, the condition for preventing contact between the transfer robot 101(i) and the walker 102(j) can be given by equation (13d).

[0101]

number

[0102] In addition, when there are multiple transport robots 101 to be controlled, conditions are also required to prevent the transport robots 101 from coming into contact with each other, but the method for calculating these conditions is similar to equations (13a), (13b), (13c), and (13d), so the explanation will be omitted.

[0103] When a target path ri that the transport robot 101(i) should follow is given, the evaluation function Ji to be considered for the transport robot 101(i) can be given by the following equation (14), following equation (8) when there is one controlled object.

[0104]

number

[0105] To optimize the behavior of all transport robots 101, the control input uc,i that minimizes the following equation (15) considering the sum J of the evaluation functions Ji of each transport robot 101 can be found by taking into consideration equations (13a), (13b), (13c), and (13d), which are constraints for preventing contact between the transport robots 101 and the pedestrian 102. In equation (15), N is the number of transport robots 101, and N=3 in the example shown in FIG.

[0106]

number

[0107] The server 105 distributes the calculated control input uc,i to each transfer robot 101. The transfer robot 101 drives the actuator B002 using the motion control system B100 provided therein.

[0108] Next, the control content by the vehicle motion calculation unit A008 when the safety standard determination unit A007 determines the safety standard to be "maximum safety" will be described. When there are multiple uncontrolled objects, the evaluation function J' to be considered can be given by the following equations (16a) and (16b), similar to equations (9a) and (9b) when there is one uncontrolled object. Here, M is the number of uncontrolled objects, and in the example shown in FIG. 10, M=2. In equation (16b), vj is the speed of the pedestrian 102 with ID j. Furthermore, the function mj(ε(k)) included in equation (16b) can be given by equations (16c), (16d), (16e), and (16f). As with the problem setting described above, the evaluation function J' in equation (16a) is formed by adding an evaluation function Jp, which aims to make the uncontrolled object approach the controlled object as quickly as possible, to the evaluation function J in equation (15). However, when there are multiple controlled and non-controlled objects, the effect of improving safety cannot be fully achieved unless consideration is given to which controlled object the vehicle is approaching. To take this issue into account, the function h(ε(k)) of equation (16f) is included in equation (16c). In equation (16c), W is a predetermined coefficient.

[0109]

number

[0110] FIG. 13 is a diagram showing an example of the relationship between the function h(ε(k)) and the function ε(k). The function ε(k) calculated by equations (16d) and (16e) corresponds to the square of the distance between the transfer robot 101(i) with ID i and the pedestrian 102(j) with ID j. When the function ε(k) is large, that is, when the transfer robot 101(i) and the pedestrian 102(j) are sufficiently far apart, the formula (16f) becomes 0. In this case, the formula (16c) also becomes 0. On the other hand, when the function ε(k) is small, that is, when the transfer robot 101(i) and the pedestrian 102(j) are close, the function h(ε(k)) shown in the formula (16f) becomes 1. Only under such conditions does the evaluation function Jp be added to the evaluation function J, as in the formulas (9a) and (9b). In other words, the problem is to calculate a control input uc that prevents the transport robot 101(i) from coming into contact with the pedestrian 102(j) when the pedestrian 102(j) approaches the nearby transport robot 101(i).

[0111] By using the control input uc calculated according to model predictive control that takes into account the evaluation functions of equations (16a) and (16b) and the constraints of equations (13a), (13b), (13c), and (13d), it is possible to make the transport robot 101 take even safer actions than when the safety standard is "standard."

[0112] According to the above-described second embodiment, the following advantageous effects are achieved.

[0113] (1) The object identification unit A004 identifies the attributes of the non-controllable object using a card (device) that is attached to the non-controllable object and can identify the attributes of the non-controllable object. This makes it possible to determine the attributes of the non-controllable object in advance and use them in the motion control of the control object.

[0114] (2) The safety standard determination unit A007 determines different safety standards if the evaluation values ​​related to the safety behavior of the non-controllable object set on the card (device) capable of identifying the attributes of the non-controllable object are different, even if the attributes of the non-controllable object identified by the object identification unit A004 and the degree of deviation of the non-controllable object from the predicted movement trajectory evaluated by the trajectory deviation evaluation unit A006 are the same. This enables more flexible motion control using not only the attributes and behavior of the non-controllable object but also the evaluation values ​​of the non-controllable object.

[0115] Although the embodiments of the present invention have been described in detail above using examples of vehicles that travel on public roads and transport robots in logistics warehouses, it goes without saying that the application of the present invention is not limited to these. For example, the present invention can also be used in transport vehicles at ports and robots that move within theme parks.

[0116] The following modified examples are also within the scope of the present invention, and it is possible to combine the configuration shown in the modified example with the configuration described in the above embodiment, to combine the configurations described in the different embodiments above, or to combine the configurations described in the different modified examples below.

[0117] <Variation 1> FIG. 14 is a diagram showing the configuration of an autonomous control system according to Modification 1, illustrating a vehicle 001, which is a moving object to be controlled, and a pedestrian 002, which is a moving object not to be controlled. In the first embodiment, an example has been described in which an environment recognizer A001 is mounted on the moving object to be controlled. However, as long as peripheral information about the moving object can be collected, the environment recognizer A001 does not need to be mounted on the moving object to be controlled. In the example shown in FIG. 14, an infrastructure sensor 003 serving as the environment recognizer A001 is installed in a location physically separated from the vehicle 001, which is the moving object to be controlled. The infrastructure sensor 003 provides the acquired sensor information of the pedestrian 002 to the vehicle 001 via a wireless system 004. In this way, for example, when multiple vehicles are used in a limited space, the cost, weight, fuel consumption, and the like per vehicle can be reduced.

[0118] <Variation 2> In the example described in Fig. 6, the safety standard is determined based only on the trajectory deviation e, but the safety standard determination unit A007 is not limited to this configuration. For example, even if the behavior of pedestrian 002 as shown in Fig. 5(c) where the trajectory deviation e is large is confirmed, if pedestrian 002 notices a vehicle 90 approaching from a distance and takes evasive action as shown in Fig. 5(d), this pedestrian 002 is determined to be highly safety-conscious. On the other hand, even in a situation similar to Fig. 5(d), if an actual trajectory is confirmed that shows the pedestrian approaching a distant vehicle 90 as shown in Fig. 5(e), this pedestrian is determined to be engaging in risky behavior.

[0119] Therefore, the safety standard may be determined using a corrected trajectory deviation e' (see FIG. 15), which is negative when the non-controlled object moves away from the controlled object as shown in FIG. 5(d) and positive when the non-controlled object moves closer to the controlled object as shown in FIG. 5(e). The trajectory deviation evaluation unit A006 calculates and stores a predicted value of the distance between the controlled object and the non-controlled object after a predetermined time has elapsed (hereinafter also referred to as the relative distance), and compares it with the actual measured value of the relative distance when the predetermined time has actually elapsed. If the actual measured value of the relative distance is greater than the predicted value, the trajectory deviation evaluation unit A006 determines that the non-controlled object is moving away from the controlled object. If the actual measured value of the relative distance is equal to or less than the predicted value, the trajectory deviation evaluation unit A006 determines that the non-controlled object is approaching the controlled object.

[0120] FIG. 15 is a diagram illustrating a method for determining the safety standard and the priority of the host vehicle by the safety standard determination unit A007 according to Modification 2. When the corrected trajectory deviation e' is negative, that is, when the non-controlled object takes an action to avoid the controlled object, safety is not compromised even if the controlled object behaves selfishly. The vertical axis of FIG. 15 schematically shows that the safety standard becomes higher as it moves above the horizontal axis, and the priority of the host vehicle (controlled object) becomes higher as it moves below the horizontal axis. FIG. 15 shows an example in which, when the corrected trajectory deviation e' is negative, the safety standard is maintained (β=0 is maintained), and the priority of the host vehicle is increased as the absolute value of the corrected trajectory deviation e' becomes larger.

[0121] In this modification, the evaluation function J' is obtained by multiplying the first term on the right-hand side of equation (10) by a coefficient γ (J' = γJ + βJp). When the corrected trajectory deviation e' is negative, the safety standard determination unit A007 increases the coefficient γ as the absolute value of the corrected trajectory deviation e' increases, that is, as the priority of the host vehicle increases.

[0122] In the example shown in FIG. 15 , when the corrected trajectory deviation e' is negative, if the corrected trajectory deviation e' is equal to or less than 0 and equal to or greater than threshold th3, the priority is set to standard. Note that the priority is also set to standard when the corrected trajectory deviation e' is equal to or greater than 0. In this case, the coefficient γ is set to 1. Note that threshold th3 is a value less than 0, for example, a value obtained by multiplying threshold th1 by −1. If the corrected trajectory deviation e' is equal to or less than threshold th4, the priority is set to the highest priority. In this case, the coefficient γ is set to a predetermined value γ1 greater than 1. Note that threshold th4 is a value less than threshold th3, for example, a value obtained by multiplying threshold th2 by −1. If the corrected trajectory deviation e' is less than threshold th3 and greater than threshold th4, the priority increases as the absolute value of the corrected trajectory deviation e' increases. In other words, the coefficient γ increases as the absolute value of the corrected trajectory deviation e' increases.

[0123] When the coefficient γ is large, the speed of the controlled object can be increased, thereby improving the efficiency of the work performed by the controlled object. For example, applying this control method to an inter-process transport vehicle operating in a factory during busy periods can contribute to improving the productivity of the factory.

[0124] As described above, when the degree of deviation from the predicted trajectory evaluated by the trajectory deviation evaluation unit A006 is greater than a predetermined threshold (e>th1) and behavior of the non-controlled object approaching the controlled object is confirmed, the safety standard determination unit A007 according to this modification changes the safety standard for the behavior of the controlled object to a safer side (higher side) than the current standard. On the other hand, when the degree of deviation from the predicted trajectory evaluated by the trajectory deviation evaluation unit A006 is greater than a predetermined threshold (e>th1) and behavior of the non-controlled object moving away from the controlled object is confirmed, the safety standard determination unit A007 does not change the safety standard to a higher side. In other words, according to this modification, it is possible to determine accurate safety standards that take into account not only the attributes of the non-controlled object but also the behavior of the non-controlled object.

[0125] The characteristics used to determine the safety standard and the priority of the host vehicle are not limited to the example shown in Fig. 15. For example, the characteristics used to determine the priority may be a characteristic that sets the priority to the highest priority when the corrected trajectory deviation e' is equal to or smaller than the threshold value th3.

[0126] Although the embodiments of the present invention have been described above, the above embodiments merely illustrate some of the application examples of the present invention, and it is not intended that the technical scope of the present invention be limited to the specific configurations of the above embodiments.

[0127] Note that the components of the autonomous control system A100 described above, as well as their functions and execution processes, may be partially or entirely implemented in hardware (for example, by designing the logic that executes each function as an integrated circuit). The control lines and information lines shown in the figure are those considered necessary for explanation, and do not necessarily represent all control lines and information lines required for the product. In reality, it may be assumed that almost all components are interconnected. [Explanation of symbols]

[0128] A001...environment recognizer, A002...state detector, A003...control object position calculation unit, A004...object identification unit, A005...non-control object position calculation unit, A006...trajectory deviation evaluation unit, A006a...predicted trajectory calculation unit, A006b...actual trajectory evaluation unit, A006c...trajectory comparison unit, A007...safety standard determination unit, A008...vehicle motion calculation unit, A100...autonomous control system, A101...controller, B001...actuator controller, B002...actuator, B100...motion control system

Claims

1. An autonomous control system that controls the behavior of a control target, which is a moving object whose behavior can be controlled, and a non-control target, which is a moving object whose behavior cannot be controlled, in an area where the control target and the non-control target do not come into contact with each other, a control object position calculation unit that calculates a position of the control object; an object identification unit that identifies attributes of the non-control object; a non-control object position calculation unit that calculates the position of the non-control object based on environmental information acquired from an external environment recognition sensor; a trajectory deviation evaluation unit that calculates a predicted movement trajectory of the non-control object corresponding to the attribute of the non-control object identified by the object identification unit, using an average value and a variance of the trajectory calculated based on the position of the non-control object calculated a predetermined time ago by the non-control object position calculation unit and a change in the position of the non-control object over time, calculates a movement trajectory of the non-control object based on the change in the position of the non-control object calculated by the non-control object position calculation unit over time, and evaluates the degree of deviation of the movement trajectory from the predicted movement trajectory; a safety standard determination unit that determines a safety standard for the behavior of the controlled object based on the attributes of the non-controlled object identified by the object identification unit and the degree of deviation from the predicted movement trajectory evaluated by the trajectory deviation evaluation unit; and a behavior modification unit that modifies the behavior of the controlled object based on the position of the controlled object calculated by the controlled object position calculation unit, the position of the non-controlled object calculated by the non-controlled object position calculation unit, and the safety standard determined by the safety standard determination unit, so that the controlled object does not approach the non-controlled object as the safety standard becomes higher; An autonomous control system equipped with:

2. 2. The autonomous control system according to claim 1, The trajectory deviation evaluation unit calculates the predicted movement trajectory of the non-controllable object based on the attributes of the non-controllable object identified by the object identification unit, using an equation of motion followed by the non-controllable object, a maximum movement speed of the non-controllable object, and a size of the non-controllable object.

3. 2. The autonomous control system according to claim 1, The safety standard determination unit changes the safety standard to a safer side than the current standard when the degree of deviation from the predicted movement trajectory evaluated by the trajectory deviation evaluation unit is greater than a predetermined threshold, but does not change the safety standard when behavior of the non-controlled object moving away from the controlled object is confirmed.

4. 2. The autonomous control system according to claim 1, The object identification unit is an autonomous control system that identifies the attributes of the non-controlled object using a device that is provided in the non-controlled object and is capable of identifying the attributes of the non-controlled object.

5. 5. The autonomous control system according to claim 4, The safety standard determination unit determines different safety standards if the evaluation value regarding the safety behavior of the non-controlled object set in a device that can identify the attributes of the non-controlled object is different, even if the attributes of the non-controlled object identified by the object identification unit and the degree of deviation of the non-controlled object from the predicted movement trajectory evaluated by the trajectory deviation evaluation unit are the same.

Images