JP6299957B2 - MOBILE BODY CONTROL DEVICE, MOBILE BODY CONTROL PROGRAM, AND MOBILE BODY CONTROL METHOD - Google Patents

Description

  The present invention relates to a moving body control device, a moving body control program, and a moving body control method, and more particularly to, for example, a moving body control device, a moving body control program, and a moving body control method that control movement of a moving body that travels autonomously.

  An example of background art of the present invention is disclosed in Patent Document 1. The robot of Patent Document 1 stores a pedestrian model indicating the trajectory of a pedestrian that passes by itself, detects the position of the pedestrian and the pedestrian, and walks passing by himself based on the detection result and the pedestrian model. The person's trajectory is predicted, and self-avoidance behavior is determined according to the prediction result. Specifically, it is determined whether or not the robot passes the pedestrian after a predetermined time. If the robot passes, it is determined whether or not the pedestrian falls within a determination circle centered on the position of the robot. Then, when the pedestrian enters the determination circle, the robot is moved so as to keep a certain distance from the pedestrian.

JP2012-200081 [B25J 13/08, G05D 1/02]

  In this background art, when a pedestrian enters the determination circle, the robot is moved so as to maintain a certain distance from the pedestrian. However, since the way of passing with the robot differs depending on the pedestrian, it is not appropriate movement control to move the robot uniformly so as to keep a certain distance. For example, when a certain distance is not sufficient for a pedestrian, there is a risk of giving a dangerous impression to the pedestrian. In addition, when a certain distance is sufficient for the pedestrian, the robot is excessively turned, and the movement time is unnecessarily long. That is, the movement efficiency is poor.

  Therefore, a main object of the present invention is to provide a novel mobile control device, mobile control program, and mobile control method.

  Another object of the present invention is to provide a moving body control device, a moving body control program, and a moving body control method capable of performing appropriate movement control.

A first aspect of the invention is a moving body control device that controls movement of a moving body that travels autonomously, and includes a storage unit, a position prediction unit, a target position calculation unit, and a control unit. Storage means, a set of feature vectors, feature vectors generated for each relative distance between the mobile body and the unspecified number of passing object for each of the avoidance behavior of the unspecified number of passing objects pass each other with a mobile And for each relative distance, a database of a plurality of feature vectors of each clustered class and an average vector of the plurality of feature vectors is stored in advance. The position prediction means selects a database corresponding to the relative distance between the moving object and the current passing object from the storage means, and selects a class including a feature vector that approximates the current feature vector of the current passing object from the database. Then, using the average vector of the class, the relative position after the predetermined time of the current passing object with respect to the moving body at the predicted position after the predetermined time is predicted. The target position calculation means calculates the target position of the moving body based on the relative position predicted by the position prediction means and the destination of the moving body. Then, the control means moves the moving body to the target position calculated by the target position calculation means .

According to the first invention, the relative position of the current passing target with respect to the moving object is predicted from the feature vector regarding the avoiding behavior of the passing object stored in advance, and the predicted relative position and the destination of the moving object are determined. Based on this, since the target position to which the mobile body should move next is calculated, the mobile body can be moved so that the passing target performs a normal avoidance action. Therefore, the moving object does not get too close to the passing object or avoid the passing object too much. That is, appropriate movement control can be performed. Therefore, a dangerous impression is not given to a pedestrian as a passing object, and the moving body can be moved efficiently.
According to the first invention, the prediction is performed using the average vector corresponding to the current avoidance action of the passing object in consideration of the relative distance between the moving object and the passing object as well as the type of avoidance action. Can do. Therefore, it is possible to control the movement of the moving body so that the passing object is caused to perform a normal avoidance action more accurately.

The second invention, the feature vector for each of the avoidance behavior of the unspecified number of passing objects moving body and passing each other to autonomous, generated for each relative distance between the mobile body and the unspecified number of passing object Clustered into a set of feature vectors , and for each relative distance, a storage unit that stores in advance a database of a plurality of feature vectors of each clustered class and an average vector of the plurality of feature vectors is provided to control movement of the moving object A moving body control program executed by a computer of a moving body control device, wherein a database corresponding to a relative distance between a moving body and a current passing object is selected from a storage unit in a processor of the computer, A class containing a feature vector that approximates the current feature vector Therefore, using the average vector of the class, the position prediction step for predicting the relative position of the current passing target with respect to the moving body of the predicted position after the predetermined time after the predetermined time, and the relative position and movement predicted at the position prediction step body target position calculating step of calculating a target position of the moving object based on the destination, and to perform the control step of moving the moving body to the calculated target position at the target position calculation step, is a mobile control program .

A third invention is the feature vector for each of the avoidance behavior of the unspecified number of passing objects moving body and passing each other to autonomous, generated for each relative distance between the mobile body and the unspecified number of passing object Clustered into a set of feature vectors , and for each relative distance, a storage unit that stores in advance a database of a plurality of feature vectors of each clustered class and an average vector of the plurality of feature vectors is provided to control movement of the moving object In the moving body control method of the moving body control device, the computer of the moving body control device selects (a) a database corresponding to the relative distance between the moving body and the current passing object from the storage means, and from the database, Find the class that contains the feature vector that approximates the current feature vector of the current passing target, and calculate the average vector of that class. (B) predicting the relative position of the current passing target with respect to the mobile object at the predicted position after a predetermined time using the threshold, and (b) the relative position predicted at step (a) and the destination of the mobile object And (c) a step of moving the moving body to the target position calculated in step (b) .

Also in the second and third inventions, appropriate movement control can be performed as in the first invention.

  According to the present invention, the relative position of the current moving object after the movement of the passing object is predicted from all the feature vectors regarding the avoiding action of the passing object stored in advance, and the predicted relative position and the destination of the moving object are determined. Based on this, since the target position to which the mobile body should move next is calculated, the mobile body can be moved so that the passing target performs a normal avoidance action. Therefore, the moving object does not get too close to the passing object or avoid the passing object too much. That is, appropriate movement control can be performed. Therefore, a dangerous impression is not given to a pedestrian as a passing object, and the moving body can be moved efficiently.

  The above object, other objects, features and advantages of the present invention will become more apparent from the following detailed description of embodiments with reference to the drawings.

FIG. 1 is an illustrative view showing an example of a mobile control system. FIG. 2 is a block diagram showing an example of the electrical configuration of the moving body. FIG. 3 is an illustrative view showing one example of an external configuration of the moving body shown in FIG. 2. FIG. 4 is an illustrative view for explaining a detection range of the distance sensor shown in FIGS. 2 and 3. FIG. 5 is a block diagram showing an example of the electrical configuration of the measuring apparatus shown in FIG. FIG. 6 is an illustrative view for explaining parameters for calculating the position and orientation of the moving body shown in FIGS. FIG. 7 is an illustrative view for explaining a feature amount when a passing object passes by a moving object. FIG. 8 is a diagram schematically showing a feature vector to be passed at a certain relative distance. FIG. 9 is an illustrative view showing a feature vector of each class and an average vector for each class when all feature vectors at a certain relative distance are clustered. FIG. 10 is an illustrative view for explaining a method of movement control of a moving body when passing a passing object. FIG. 11 is an illustrative view for explaining a method of calculating a target position of a moving body. FIG. 12 is another illustrative view for explaining a method of calculating the target position of the moving body. FIG. 13 is an illustrative view showing one example of a memory map of the RAM shown in FIG. FIG. 14 is an illustrative view showing specific contents of the average vector data shown in FIG. FIG. 15 is a flowchart showing a part of the movement control process of the CPU shown in FIG. 16 is another part of the movement control process of the CPU shown in FIG. 2, and is a flowchart subsequent to FIG. FIG. 17 is a flowchart showing post-movement position calculation processing of the CPU shown in FIG. FIG. 18 is an illustrative view showing an example in which a distance sensor is arranged in an environment with poor visibility. FIG. 19 is an illustrative view for describing a method of controlling movement of a moving object in the environment as shown in FIG.

  FIG. 1 shows a mobile body control system 1 according to this embodiment, which includes a mobile body 10 and a measurement device 100, and the mobile body 10 and the measurement device 100 are communicably connected via a network 2. However, what actually communicates is a computer 12 (see FIG. 2) included in the moving body 10 and a computer 102 (see FIG. 5) included in the measurement apparatus 100.

  FIG. 2 is a block diagram showing an electrical configuration of the moving body 10 shown in FIG. The moving body 10 includes a computer 12 that also functions as a movement control device and a route calculation device. The computer 12 is a computer such as a general-purpose personal computer or workstation, and includes components such as a CPU 12a, a RAM 12b and an HDD 12c, and a communication device (not shown). The computer 102 to be described later is also provided with such components.

  Further, an input / output interface (hereinafter simply referred to as “interface”) 14, an operation lever 16, and distance sensors 18 a and 18 b are connected to the computer 12. A motor 22a is connected to the interface 14 via a motor driver 20a, and a motor 22b is connected via a motor driver 20b. Furthermore, encoders 24 a and 24 b are connected to the interface 14.

  The rotating shaft of the motor 22a and the rotating shaft of the left rear wheel 34a (see FIG. 3) are connected using a gear, and the rotating shaft of the motor 22b and the rotating shaft of the right rear wheel 34b (see FIG. 3) are the gears. Are connected using

  Here, with reference to FIG. 3 (A), the moving body 10 of this embodiment is a moving body such as an electric wheelchair, and the left and right front wheels (casters) 32a and 32b and the left and right rear wheels are provided below the moving body. 34a and 34b are provided. Further, the above-described operation lever 16 is provided on the upper part of the left frame of the moving body 10. Furthermore, the distance sensor 18a described above is attached to the left side of the moving body 10 and on the left side of the step 36a where the user places his left foot. However, the distance sensor 18a is a frame on the right side of the moving body 10 and may be attached to the right side of the step 36b where the user puts the right foot. Further, as shown in FIG. 3B, which is a schematic view of the moving body 10 as viewed from above, the distance sensor 18 b described above is provided behind the moving body 10.

  In this embodiment, the distance sensors 18 a and 18 b are attached to the front and rear of the moving body 10, but may be attached to the left and right of the moving body 10. The reason why the two distance sensors 18a and 18b are provided on the front and rear or the left and right of the moving body 10 is to detect (measure) the distance around the moving body 10 over the entire circumference (360 °). However, this is merely an example, and there may be one distance sensor or three or more distance sensors.

  In this embodiment, the distance sensors 18a and 18b are not only fixed objects (static obstacles) such as walls and pillars in a certain environment, but are also moved like humans (pedestrians). Measure (detect) the distance to the target object (dynamic obstacle).

  Returning to FIG. 3A, a box 40 is provided below the seat of the moving body 10 and between the left rear wheel 34 a and the right rear wheel 34 b. In this, the above-described computer 12, interface 14, motor drivers 20a and 20b, motors 22a and 22b, and encoders 24a and 24b are provided. However, the computer 12 and the interface 14 may be provided outside the box 40.

  The computer 12 (CPU 12a) governs overall control of the moving body 10 of this embodiment. In this embodiment, the CPU 12a calculates the moving path of the moving body 10. The CPU 12a controls driving of the motors 22a and 22b automatically or in response to an operation input from the operation lever 16. That is, the CPU 12a controls the movement of the moving body 10. Further, the CPU 12a acquires and stores distance data and rotation speed data detected by the distance sensors 18a and 18b and the encoders 24a and 24b. Then, the CPU 12a measures the distance to the obstacle and predicts (determines) the position of the moving body 10 based on the detected data.

  The operation lever 16 is operated by a user on the moving body 10, and a signal (operation input) corresponding to the operation is given to the computer 12. For example, operations such as forward, backward, stop, left turn, right turn, and turn can be performed.

  However, an operation input can be given to the computer 12 by remote control using a remote controller (not shown). Further, in this embodiment, the operation lever 16 may not be provided in order to make the moving body 10 autonomously travel.

  The distance sensors 18a and 18b are, for example, general-purpose laser range finders (LRF), and measure the distance to the object from the time it takes to irradiate the laser and reflect it back to the object. For example, in the LRF, a laser beam emitted from a transmitter (not shown) is reflected by a rotating mirror (not shown), and scanned in a fan shape at a certain angle (for example, 0.25 degrees). In this embodiment, the period for scanning the entire detection range (see FIG. 4) is 25 msec.

  FIG. 4 is an illustrative view for explaining detection ranges (measurement ranges) of the distance sensors 18a and 18b. The measurement ranges of the distance sensors 18a and 18b of this embodiment are shown in a fan shape with a radius R (R≈10 to 15 m). The angle of the fan is 135 ° on each of the left side and the right side with respect to the front direction. That is, the distance can be detected for a range of 270 ° centered on the front.

  Returning to FIG. 2, the motor drivers 20 a and 20 b drive the motors 22 a and 22 b based on an instruction from the computer 12. The encoder 24 a detects the rotation speed (rps) of the motor 22 a and inputs data about the rotation speed (rotation speed data) to the computer 12 via the interface 14. Similarly, the encoder 24b detects the rotational speed of the motor 22b, and inputs rotational speed data regarding the rotational speed to the computer 12 via the interface 14.

  The computer 12 acquires distance data detected by the distance sensors 18a and 18b and rotation speed data detected by the encoders 24a and 24b, and stores them in the RAM 12b (or the HDD 12c). However, the distance data detected by the distance sensor 18a and the distance data detected by the distance sensor 18b are distinguished, and the distance data detected at the same time point and the rotation speed data are associated with each other.

  As shown in FIG. 5, the measuring apparatus 100 includes a computer 102, and a plurality of distance sensors 104 are connected to the computer 102. Although not shown, the plurality of distance sensors 104 are installed in an environment in which the mobile body 10 travels autonomously. The distance sensor 104 may be the same as the distance sensors 18a and 18b described above. Therefore, each distance sensor 104 is disposed so as to partially overlap the detection range of the other distance sensor 104, and the distance can be measured over the entire range of the environment in which the moving body 10 autonomously travels.

  However, as the distance sensors 18a, 18b, and 104, an ultrasonic distance sensor or a millimeter wave radar may be used instead of the LRF.

  In this embodiment, the measuring apparatus 100 tracks the position of the moving body 10 and also includes one or more passing target candidates (hereinafter referred to as “passing candidates”) that exist in an environment in which the moving body 10 travels autonomously. And each passing candidate is also tracked. The measuring apparatus 100 determines one passing candidate predicted to pass the moving body 10 earliest in time among passing candidates moving so as to face the moving body 10 as a passing target. When the passing object is determined, the measuring apparatus 100 transmits the passing object information (hereinafter, referred to as “passing object information”) to the moving body 10.

  Here, the passing object information is the distance between the moving object 10 and the passing object (the relative distance D. See FIG. 7), the current position of the passing object, the moving speed at the current time of the passing object, and the current time of the passing object. And the feature vector calculated for the passing object. The feature vector and its calculation method will be described later in detail.

  However, since the passing position (passing candidate) is tracked by the measuring apparatus 100, the current position of the passing target and the moving direction of the passing target up to the current time can be easily known. The passing object information is transmitted to the moving body 10 every time the relative distance D changes by a predetermined distance (for example, 1 m) until the passing object passes the moving object 10. In addition, each time, a feature vector for a passing object is calculated (recalculated).

  Here, the passing object means a pedestrian who passes the moving body 10 and includes not only a pedestrian who does not use an auxiliary device such as a walker but also a pedestrian, a cart or a shopping cart which uses a walker or a cane. Pedestrians to be pushed (may include carts and shopping carts) and other wheelchairs different from the moving body 10 are applicable.

  As described above, since the passing objects include pedestrians and wheelchairs, the distance sensors 104 described above are installed at a height of about 70 cm to 100 cm from the floor in order to detect them.

  In addition, the measuring apparatus 100 stores data on a two-dimensional map (world coordinate system map) in which the environment in which the moving body 10 autonomously travels is viewed from above.

  Therefore, when the measuring apparatus 100 detects an object that does not appear on the map based on the distance data detected by the distance sensor 104, the measuring apparatus 100 first determines whether it is a passing candidate. However, since the start position of the moving body 10 is input to the measuring apparatus 100, the measuring apparatus 100 recognizes that it is the moving body 10 and tracks it. Therefore, it is not determined whether or not the moving body 10 is a passing candidate.

  In this embodiment, the measuring apparatus 100 detects the size of the detected object and the presence or absence of movement, the size of the object is within a predetermined size range (first condition), and When the object is moving (second condition), the object is determined as a passing candidate.

  For example, the size of the object approximates the shape of the object specified based on the distance data output from the distance sensor 104 with a circle, and if the approximate circle has a diameter of about 70 cm to 150 cm, Judge that it is a wheelchair. That is, it is determined that the first condition as a passing candidate is satisfied. However, when the size of the object is out of the above range, the object is determined as some obstacle that is not a passing object.

  Further, by detecting the position of the object that satisfies the first condition every predetermined time (for example, several hundred milliseconds), the presence or absence of the movement of the object is detected. If the position of the object does not change even after a certain period of time (for example, several seconds), the object does not move, so the second condition is not satisfied, and the object is regarded as an obstacle. To be judged. On the other hand, when the position of the object changes without exceeding a certain period, the object has moved, the second condition is satisfied, and the object is determined as a passing candidate. Identification information is attached to an object determined as a passing candidate, and its position is tracked. That is, the current position of the passing candidate is detected every predetermined time (for example, several hundred milliseconds), and the locus is acquired.

  In the measuring apparatus 100, the computer 102 estimates changes in the current position of the moving object 10 and each passing candidate using a particle filter based on the output (distance data) from the distance sensor 104. Here, the particle filter is a technique for estimating the current state from the prediction at the previous time and the current observation information using a large number of particles. The particle filter tracks the moving object 10 passing candidate by repeating prediction, observation, and resampling.

  For example, when scanned by the distance sensor 104, a visible area where the moving object 10 passing candidate does not exist, a shadow area where the moving object 10 passing candidate exists, and an edge of the moving object 10 passing candidate are detected. Further, particles are evenly distributed in a virtual space corresponding to the real space (environment in which the moving body 10 autonomously travels), and the likelihood is obtained for each distance sensor 104. Furthermore, each particle is updated by integrating the likelihood for each distance sensor 104. Then, a change in the current position of the moving object 10 passing candidate is estimated by each updated particle. The likelihood is a constant value in the visible area, and is the sum of the constant value and the likelihood of the edge in the shadow area. Based on the change in the current position estimated in this way, the position of the moving object 10 passing candidate is obtained, and numerical values (position data) indicating the plane coordinates of the position are stored in time series, thereby moving the moving object. Trajectory data is generated for 10 passing candidates.

  However, the particle filter for tracking the moving object 10 and the particle filter for tracking the passing candidate are independent, and the particle filter for tracking each passing candidate is also independent.

  In this embodiment, the moving body 10 inputs (sets) a start point and a goal point (destination) to calculate a movement route (performs a movement route plan) and performs automatic traveling (autonomous movement). )can do.

  The normal travel route plan is a method for calculating the shortest distance and the shortest time route. In this embodiment, the travel route that is the shortest distance is calculated. When the movement route is calculated, the moving body 10 is moved according to the calculated movement route.

  The mobile body 10 autonomously travels according to the calculated travel route. At this time, the motor drivers 20a and 20b drive the motors 22a and 22b based on an instruction from the computer 12. The encoder 24 a detects the rotation speed (rps) of the motor 22 a and inputs data about the rotation speed (rotation speed data) to the computer 12 via the interface 14. Similarly, the encoder 248 b detects the rotational speed of the motor 22 b and inputs rotational speed data regarding the rotational speed to the computer 12 via the interface 14.

  The computer 12 acquires distance data detected by the distance sensors 18a and 18b and rotation speed data detected by the encoders 24a and 24b, and stores them in the RAM 12b (or the HDD 12c). However, the distance data detected by the distance sensor 18a and the distance data detected by the distance sensor 18b are distinguished, and the distance data detected at the same time point and the rotation speed data are associated with each other.

  The distance data measured by each of the distance sensors 18a and 18b mounted on the moving body 10 is a two-dimensional coordinate system (moving body coordinates based on the position and orientation of the moving body 10 when the distance data is detected. System) position (point) data. At this time, the mounting positions of the two distance sensors 18a and 18b are considered. That is, the displacement between the position of the moving body 10 and the attachment positions of the distance sensors 18a and 18b is adjusted. Of course, the angle in the measurement range of the distance sensors 18a and 18b when the distance data is detected is also taken into consideration. Furthermore, in this embodiment, position data in the moving body coordinate system is converted into position data in a two-dimensional world coordinate system. In other words, in the moving body coordinate system, since the position and orientation of the moving body 10 when the detection of the distance data (movement of the moving body 10) measured by the distance sensors 18a and 18b is started is used as a reference. It converts to the world coordinate system about the real space.

Here, in this embodiment, the position and orientation of the moving body 10 are obtained based on the measured values by odometry. As shown in FIG. 6, the position and orientation of the moving body 10 are determined based on the movement [vm of the moving body 10 calculated using the moving speeds V _L and V _R of the left and right wheels (34a, 34b) and the wheel distance DT. , ω] ^T is obtained by tracking ^T.

Specifically, the operation [vm, ω] ^T of the moving body 10 is calculated according to Equation 1. However, vm is the speed of the moving body 10, and ω is the center of the axle of the wheels 34a and 34b (see FIG. 6), and is the angular speed around the axis perpendicular to the axle.

[Equation 1]

However, the moving speeds V _L and V _R of the left and right wheels (34a, 34b) can be calculated from the rotational speeds of the left and right wheels (34a, 34b), and the wheel interval DT can be known in advance. However, the moving velocity V _L corresponding to the rotational speed, is prepared a table of V _R, without computing, it may acquire the moving speed V _L, V _R by referring to the table.

Then, by integrating the calculated operation [vm, ω] ^T , the previously calculated movement distance dm and rotation angle θ from the position and orientation of the moving body 10 can be obtained. Accordingly, the map is obtained by cumulatively adding the movement distance dm and the rotation angle θ calculated according to the time series to the initial position (start point) and the initial direction (angle θ = 0) of the moving body 10. It is possible to know the position and orientation of the moving body 10 over time. That is, it is possible to know whether or not the moving body 10 is moving according to the calculated movement route. When the moving object 10 is deviated from the movement route, the movement is controlled so as to move on the movement route.

  As described above, since the measuring device 100 tracks the moving body 10, the coordinate data of the current position of the moving body 10 may be acquired from the measuring device 100.

  In this way, the moving body 10 moves toward the destination according to the calculated movement route. However, when passing a passing object such as a pedestrian or the like, it is necessary to move the moving body 10 while avoiding the passing object so as not to collide with the passing object, regardless of the previously calculated movement route.

  For example, as a method of moving the moving body 10 while avoiding the passing object, the position of the passing object after a predetermined time is set such that the moving object 10 maintains a certain distance from the position of the moving body 10 after the predetermined time. It is conceivable to control the movement.

  However, there are various actions that a passing object such as a pedestrian avoids the moving body 10, and a certain distance may be appropriate for a pedestrian or the like, but is appropriate for another pedestrian or the like. Is not limited. Therefore, a pedestrian who feels that a certain distance is too close may give fear. Moreover, the pedestrian who feels that a certain distance is too far turns the moving body 10 more than necessary, and the moving body 10 cannot be moved efficiently.

  Therefore, in this embodiment, the movement of the moving body 10 is controlled so that the normal avoidance action is gradually performed on the passing target from the initial stage of passing, thereby realizing safe passing and performing appropriate movement control. It is like that.

  Briefly, for each of a plurality of (unspecified many) passing objects in which the moving body 10 has passed, a feature amount for an action (avoidance action) that avoids the moving body 10 is acquired, and based on the acquired feature amount Create a database in advance. Then, using the database, the position of the current passing target after a predetermined time (after movement) is predicted, and the moving body 10 is controlled to move based on the prediction result.

  However, the position (predicted position) after the predetermined time of the passing object this time is a relative position (relative position) determined by a lateral distance d (t + 1) described later with respect to the predicted position of the moving body 10. . The same applies hereinafter.

  Next, a method for generating a database based on feature values and a method for controlling movement of the moving object 10 using the database will be described in order.

  FIG. 7 shows a feature amount when a state in which the moving body 10 passes by a pedestrian as a passing target at time t is described. In the example illustrated in FIG. 7, the passing object moves from left to right, and the moving body 10 moves from right to left. Further, as shown in FIG. 7, at time t, the relative distance between the moving object 10 and the passing object is represented by D (t). Further, at time t, the straight line distance (hereinafter referred to as “side distance”) between the line passing through the current position of the moving body 10 and the destination of the moving body 10 and the passing object is represented by d (t). Is done. At time t, the moving speed of the passing object is represented by v (t).

  In this embodiment, the measurement apparatus 100 acquires the above-described feature amount for various passing objects. However, the feature amount is detected after the passing target (passing target) is detected by the measurement apparatus 100 until the passing target is passed. That is, the measuring apparatus 100 measures (calculates) the relative distance D (t), the moving speed v (t), and the lateral distance d (t) for each passing candidate at predetermined time intervals.

  Feature quantities for a plurality of passing objects are acquired, and feature vectors are generated from temporal changes of the feature quantities. The state (how to) of passing differs greatly depending on the relative distance D (t) between the passing object and the moving body 10. Therefore, a database is generated for each relative distance D (t). In this embodiment, for each of the relative distances D (t) of 7, 6, 5, 4, 3, 2 (m), the method of passing the previous predetermined time (3 seconds in this embodiment) is described. It is conceivable to express in a time series of the lateral distance d (t) and the moving speed v (t). When the relative distance D (t) reaches a certain value, a certain passing method can be expressed by a feature vector F (t) shown in Formula 2.

[Equation 2]
F (t) = (d (t), d (t-Δ),…, d (t- (N-1) Δ),
v (t), v (t-Δ),…, v (t- (N-1) Δ)]
Note that Δ is an interval (unit time) for expressing the time series, and is set to 500 msec, for example. N is determined to be a value obtained by dividing a predetermined time by Δ. Therefore, in this embodiment, since the predetermined time is 3 seconds, N = 6.

  Thus, in this embodiment, the feature vector F is represented by the lateral distance d (t) and the moving speed v (t). FIG. 8A shows a feature vector for a certain passing object at a certain relative distance D. FIG. However, in FIG. 8A, for the sake of simplicity, only a temporal change in the lateral distance d is shown, and the moving speed v (t) is omitted. The same applies to FIGS. 8B and 9A to 9C.

  Such feature vectors are generated for all measurement results, and all feature vectors are clustered. In this embodiment, a set of feature vectors is generated for each relative distance D (t). For example, as shown in FIG. 8B, a set of feature vectors at a certain relative distance D (t) is generated. A set of such feature vectors is clustered.

  In order to classify the ways of passing, clustering is performed by defining a distance function of two ways of passing. The distance function distance (Fa, Fb) of the feature vectors Fa and Fb of the two passing methods can be defined as, for example, Equation 3 and Equation 4.

  Here, d '(t) and v' (t) represent amounts obtained by normalizing d (t) and v (t), respectively. Ni is a weight, and the importance of an element close to the current time is increased.

  As a clustering method, for example, a k-average method or an x-average method for automatically determining the number of clusters is effective. In this way, a database is created for each relative distance D (t).

  Therefore, for example, by clustering the set of feature vectors shown in FIG. 8B, as shown in FIG. 9A, FIG. 9B, and FIG. Categorized as a category).

  For example, the class shown in FIG. 9A is a class of feature vectors when the passing target takes a large avoidance action. The class shown in FIG. 9B is a class of feature vectors when the passing object takes a moderate avoidance action. The class shown in FIG. 9C is a class of feature vectors when the passing object takes a small avoidance action.

  Note that the number of classes (types of avoidance actions) need not be limited to this. For example, it is considered that the avoidance behavior differs between a pedestrian who does not use a walker or the like, and a pedestrian or wheelchair who uses a walker or the like. In addition, the avoidance behavior is considered to vary depending on the age of the pedestrian.

  It should be noted that it is not necessary to specify the type of avoidance action, and it is only necessary to classify different avoidance actions.

  As described above, a set of feature vectors at a certain relative distance D (t) is clustered, and an average vector for a plurality of feature vectors of each class is calculated. Although illustration is omitted, feature vectors are similarly generated for other relative distances D (t), and after being clustered, an average vector of each class is calculated. The plurality of feature vectors and the average vector are registered in the computer 12 (the moving body 10) so that the class (type of avoidance action) can be identified for each relative distance D (t).

  Therefore, when the moving body 10 passes the passing object, the relative distance D (t) is determined from the distance between the moving body 10 and the passing object, and the characteristics of the passing object from each class at the determined relative distance D (t). Find a feature vector that approximates the vector. Then, the movement of the passing object is predicted using the average vector of the class including the obtained feature vector. That is, for each relative distance D (t), a database based on the acquired feature amount is generated, and movement of the moving body 10 is controlled using the generated database.

  When the movement of the passing object is newly measured, the measuring apparatus 100 calculates (generates) the feature vector F according to Equation 2 from the position measured for the passing object and the moving speed v (t). Then, the measuring apparatus 100 transmits the passing target information to the moving body 10.

  In response to this, the moving body 10 selects a database corresponding to the value of the relative distance D (t), and obtains a feature vector that is most similar (approximated) to the feature vector F from the selected database. A type (class) of avoidance action including a feature vector is selected. The average vector of the selected class is extracted and the average vector is used as a future prediction result.

  In this embodiment, when obtaining a feature vector that is most similar (approximate) to the feature vector F from the selected database, the distance between the feature vector F and each of all the feature vectors included in the selected database. And the feature vector with the shortest distance is obtained as an approximate feature vector. However, this distance can be calculated according to the above equations 3 and 4. In addition, when there are two or more closest feature vectors, any one feature vector is arbitrarily selected. That is, it is only necessary to select one feature vector that approximates the feature vector F.

  As a prediction result, a time series of the lateral distance d (t) with reference to the direction of the destination of the moving body 10 is obtained. However, since the current lateral distance d (t) is different from the average vector, if the average vector is directly used as a future prediction result, the movement becomes discontinuous. Therefore, the average vector translated so that both are the same is used. That is, the average vector is translated so that the calculated side distance d (t) of the feature vector F at the present time matches the side distance d (t) of the average vector. Then, the lateral distance d (t + 1) at time t + 1 is acquired from the average vector.

  As described above, the prediction result is a time series of the lateral distance d (t) with reference to the direction of the destination of the moving body 10 and is a relative positional relationship between the two, so that the actual passing The series of the lateral distance d (t) changes according to the movement (movement) of the moving body 10. Therefore, the moving body 10 is controlled to move so that the passing object can realize the predicted lateral distance d (t). As shown in FIG. 10, when the passing object moves to the position x (t + 1), a path along which the moving body 10 moves to the position y (t + 1) so as to realize the predicted lateral distance d (t + 1). A plan is made. This realizes the passing as intended of the passing object, and there is an advantage that there is little uncomfortable feeling given to a pedestrian or the like as the passing object with respect to the movement of the moving body 10, and a dangerous impression is not given.

  However, if the moving body 10 is controlled to follow the predicted lateral distance d (t), there is a possibility that the moving body 10 can make a movement plan that cannot be realized, such as a sudden speed change or direction change.

  Therefore, in this embodiment, among the positions where the moving body 10 can be realized at the time t + 1, the moving body 10 performs movement control so as to perform the movement closest to the predicted lateral distance d (t + 1). That is, a feasible movement plan is made within the constraints such as the maximum acceleration and maximum rotational angular velocity of movement of the moving body 10. Therefore, the moving body 10 also stores the same data as the map data stored in the measuring apparatus 100.

  The following method is mentioned as an example of such a method of movement control of the moving body 10. When the current time is t, the position (target position) after movement of the moving body 10 at time t + 1 is obtained in the following three stages.

<First stage>
Assuming that the moving body 10 moves while maintaining the current moving speed vm, the passing object moves so as to maintain a distance LD = d (t + 1) × α with respect to the predicted lateral distance d (t + 1). That is expected. Here, α is a parameter, and 0 ≦ α ≦ 1. The parameter α indicates the rate at which the passing object avoids the moving body 10 because the side distance d (t + 1) results.

  As shown in FIG. 11, the predicted position of the moving body 10 at time t + 1 is a point (position) E, and the destination of the moving body 10 is a point G. A straight line that is a distance LD away from the direction of the destination of the moving body 10 is defined as a straight line L1. However, the direction of the destination of the moving body 10 is a direction in which a straight line L2 connecting the point E of the predicted position of the moving body 10 and the point G of the destination extends. Further, it is assumed that the straight line L1 passes through the point A as the point closest to the point E. Similarly, a point on the straight line L1 closest to the point G is defined as a point B. If the passing object exists at a position corresponding to the point M at the current time t and can be moved by the distance w by the time t + 1, the reachable position of the passing object is on the circle C1 with the radius from the point M as the distance w. A position corresponding to a point. Therefore, by obtaining the intersection of the straight line L1 and the circle C1, the point N is obtained as the predicted position of the passing object. However, the distance w is predicted from the current moving speed v (t) of the passing object. This moving speed v (t) is included in the passing object information.

  Specifically, the point A and the point B are expressed by Equation 5 and Equation 6, respectively. However, (dx, dy) is a vector from the point E to the point A, which is perpendicular to the straight line L2 connecting the point E and the point G, and can be obtained because the length is the distance LD.

[Equation 5]

[Equation 6]

  Therefore, the straight line L1 and the circle C1 are expressed by Equation 7 and Equation 8, respectively. The solutions satisfying Expressions 7 and 8 correspond to the intersection of the straight line L1 and the circle C1 in FIG. 11, and generally two are obtained. Therefore, in this embodiment, the passing object is considered to move to a position close to the extension line in the moving direction up to the current time t, and one point (solution) corresponding to the position close to the extension line is selected. . The point (solution) is the predicted position (here, point N) at the time t + 1 to be passed. However, the moving direction of the passing target up to the current time t is included in the passing target information.

[Equation 7]

[Equation 8]
(x-p _t ) ² + (y-q _t ) ² = w ²
<Second stage>
Next, in order to move the moving body 10 to the position where the lateral distance d (t + 1) is maintained with respect to the predicted position (point N) to be passed, as shown in FIG. The moving body 10 may be moved on a tangent line (straight line L3) drawn from the point G, which is the destination of the moving body 10, to the circle C2, considering a circle C2 having a radius d (t + 1) centered on. In FIG. 12, a point of contact between a circle C2 and a straight line L3 is a point H. In this case, the circle C2 is expressed by Equation 9, and the condition that the point H is a contact is expressed by Equation 10. A point H as a contact point is obtained as a solution of Equations 9 and 10.

[Equation 9]

[Equation 10]
(x-p _{t + 1} ) (x-x _g ) + (y-q _{t + 1} ) (y-y _g ) = 0
<Third stage>
Also, assuming that the distance that the moving body 10 can move to time t + 1 is r, the reachable position of the moving body 10 at time t + 1 is a point on the circle C3 having a radius from the point J at the current time t. Corresponding position. However, the distance r is predicted from the moving speed vm of the moving body 10 at the current time t. An intersection of the straight line L3 and the circle C3 is obtained, and this is set as a target position at a point K of the moving body 10 at time t + 1. However, the straight line L3 is represented by Equation 11, the circle C3 is represented by Equation 12, and the intersection is obtained as a solution of Equations 11 and 12. Here, two solutions are generally obtained, but the solution closer to the point G of the destination of the moving body 10 is adopted as the target position.

[Equation 11]

[Equation 12]
(x-x _t ) ² + (y-y _t ) ² = r ²
This forecast is updated regularly. That is, when the moving object 10 approaches or separates from the passing object, the database is changed with respect to the changed relative distance D (t), and the prediction is performed again, thereby making a more accurate prediction. it can. In this embodiment, since the database is created by changing the relative distance D (t) at 1 m intervals, the prediction is updated every time the relative distance D (t) changes by 1 m.

  Since the moving object 10 and the passing object pass each other, the relative distance D (t) is considered to be gradually shortened. However, the relative distance D (t) may be increased depending on how the passing object moves. For this reason, every time the relative distance D (t) changes by a predetermined distance (1 m), the database is changed and predicted again.

  Further, when the moving object 10 passes by the passing object, it recalculates the moving route from the current position and the destination. And it moves toward the destination along the recalculated movement route.

  However, whether or not the moving body 10 has passed the passing object is determined by the measuring apparatus 100, and when it is determined that the moving body 10 has passed, this is notified to the moving body 10. The measuring apparatus 100 determines whether or not the passing object exists behind the left and right sides of the moving body 10.

  In this embodiment, the moving body 10 may be moved from the current position to the closest point (position) on the moving path so that the moving body 10 is moved along the moving path calculated first. .

  FIG. 13 shows an example of the memory map 300 of the RAM 12b shown in FIG. As shown in FIG. 13, the RAM 12 b includes a program storage area 302 and a data storage area 304. The program storage area 302 stores a movement control program for controlling movement of the moving body 10. The movement control program includes a movement route calculation program 302a, a lateral distance acquisition program 302b, a target position calculation program 302c, a drive control program 302d, a state calculation program 302e, and the like.

  The movement route calculation program 302a is a program for calculating a movement route. When the movement control is started, the movement route is calculated from the start point and the destination, or when the passing target is passed, Recalculate the travel route from the current position and destination.

  The lateral distance acquisition program 302b obtains a feature vector that is closest to the feature vector F included in the passing target information obtained from the measurement apparatus 100, selects (detects) the average vector of the class including the obtained feature vector, and selects This is a program for obtaining the predicted lateral distance d (t + 1) using the average vector.

  The target position calculation program 302c is a program for calculating a target position at which the moving body 10 should move at time t + 1 using the side distance d (t + 1) acquired according to the side distance acquisition program 302b.

  The drive control program 302d moves the moving body 10 according to the movement path calculated according to the movement path calculation program 302a, or moves the moving body 10 toward the target position calculated according to the target position calculation program 302c. This is a program for controlling the driving of the motors 22a and 22b.

  The state calculation program 302e is a program for calculating the position and direction of the moving body 10 based on the measurement values obtained by odometry and storing (updating) state data 304f described later.

  Although illustration and detailed description are omitted, the program storage area 302 also stores programs for communicating with other computers.

  The data storage area 304 includes map data 304a, all feature vector data 304b, start position data 304c, destination data 304d, movement route data 304e, state data 304f, passing target information data 304g, side distance data 304h, predicted position. Data 304i, straight line data 304j, target position data 304k, and the like are stored.

  The map data 304a is data about a two-dimensional map in which the environment in which the mobile object 10 is traveled is viewed from above. This map is a map in the world coordinate system. For example, the map data 304a is generated in advance and input to the computer 12, or stored in advance in the HDD 12c in the computer 12, and stored in the RAM 12b when executing the movement control process.

  The total feature vector data 304b is data about all the feature vectors registered in advance, and data of a plurality of feature vectors generated and classified for each relative distance D and an average vector data of the plurality of feature vectors. Including. As shown in FIG. 14, the total feature vector data 304b includes a database for each relative distance D. Specifically, the total feature vector data 304b includes a first distance database 310 for the relative distance D = 7 (m), a second distance database 320 for the relative distance D = 6 (m), and a relative distance D = 5 ( a third distance database 330 for m), a fourth distance database 340 for relative distance D = 4 (m), a fifth distance database 350 for relative distance D = 3 (m), and a relative distance D = 2 (m) A sixth distance database 360 for.

  The first distance database 310 further includes first type feature vector data 310a, second type feature vector data 310b,. That is, it includes data on a plurality of feature vectors and their average vectors for each type (class) of avoidance action.

  Although illustration is omitted, the same applies to the other distance databases (320-360).

  Returning to FIG. 13, the start position data 304 c is data (two-dimensional coordinate data) regarding the start position of the moving body 10. The destination data 304d is two-dimensional coordinate data regarding the destination which is the final target position of the mobile object 10. For example, the start position and the destination are input to the computer 12 from another computer (host computer).

  The movement route data 304e is data about a movement route calculated according to the movement route calculation program 302a, and includes, for example, a plurality of points (positions) arranged in order from the start position or the current position of the moving body 10 toward the destination. ) Of each two-dimensional coordinate data (data group).

  The state data 304f is two-dimensional coordinate data about the current position of the moving body 10 calculated according to the state calculation program 302e and data about the current direction of the moving body 10 (orientation data). However, the two-dimensional coordinate data and the orientation data included in the state data 304f are both data converted into the world coordinate system.

  The passing object information data 304g is data on passing object information acquired from the measuring apparatus 100. The side distance data 304h is data on the side distance d (t + 1) acquired according to the side distance acquisition program 302b.

  The predicted position data 304i is two-dimensional coordinate data regarding the position of the passing target predicted at time t + 1. The straight line data 304j is data about a straight line (L3 in FIG. 12) including a point (position) to which the moving body 10 should move at time t + 1. The target position data 304k is two-dimensional coordinate data of a position after movement (target position) where the moving body 10 should exist at time t + 1.

  Although not shown, the data storage area 304 stores other data necessary for executing the movement control program, and is provided with a counter (timer) and a flag.

  15 and 16 are flowcharts of the movement control process of the CPU 12a shown in FIG. When the start position and the destination are input, as shown in FIG. 15, the CPU 12a starts the movement control process, and calculates the movement route in step S1. Here, the CPU 12a refers to the map data 304a, calculates a travel route connecting the input start position and destination according to a general route plan, and stores the corresponding travel route data 304e in the RAM 12b.

  In the next step S3, the movement of the moving body 10 is started so as to move according to the movement route calculated in step S1. That is, the CPU 12a controls the driving of the motors 22a and 22b so as to move the moving body 10 along the movement path. The same applies to the case where the moving body 10 is moved.

  Although detailed description is omitted, the position and orientation of the moving body 10 are tracked by a flow different from the movement control process, and the state data 304f is updated each time the moving body 10 is moved.

  Subsequently, in step S <b> 5, it is determined whether or not passing target information has been acquired from the measurement apparatus 100. If “NO” in the step S5, that is, if the passing target information is not acquired, it is determined whether or not the destination is reached in a step S7. Here, the CPU 12a determines whether or not the two-dimensional coordinate data of the current position of the moving object 10 included in the destination data 304d and the state data 304f match or substantially match.

  If “NO” in the step S7, that is, if the destination is not reached, the process returns to the step S5 as it is. Accordingly, the moving body 10 is moved along the moving path. On the other hand, if “YES” in the step S7, that is, if the destination has been reached, the movement is terminated in a step S9, and the movement control process is ended. In step S9, the CPU 12a controls the driving of the motors 22a and 22b so as to decelerate and stop, for example.

  If “YES” in the step S5, that is, if the passing object information is acquired, the database is selected in a step S11 according to the relative distance D included in the passing object information. That is, one database corresponding to the relative distance D is selected from a plurality of databases (310-360) included in all feature vector data 304b.

  In the subsequent step S13, a feature vector that most closely approximates the feature vector F of the passing target included in the passing target information is obtained from the selected database. In step S14, a class including the feature vector obtained in step S13 is obtained.

  Subsequently, as shown in FIG. 16, a target position calculation process (see FIG. 17) described later is executed in step S15. In the next step S17, the moving body 10 is moved toward the calculated target position. In step S19, it is determined whether or not the destination has been reached. This determination process is the same as step S7 described above.

  If “YES” in this step S19, the process proceeds to the step S9 shown in FIG. 15, but if “NO”, it is determined whether or not a passing object is passed in a step S21. Here, the CPU 12a determines whether or not it has been notified from the measuring apparatus 100 that it has passed as a passing target.

  If “NO” in the step S21, that is, if the passing object is not passed, it is determined whether or not the passing object information is acquired from the measuring apparatus 100 in a step S23. That is, here, the CPU 12a determines whether or not the updated passing target information has been acquired from the measuring device 100 because the relative distance D has changed by a predetermined distance.

  If “NO” in the step S23, the process returns to the step S15. On the other hand, if “YES” in the step S23, the process returns to the step S11 shown in FIG. Therefore, a feature vector that is closest to the updated feature vector F is obtained from the database corresponding to the relative distance D that has changed by a predetermined distance, and a class that includes the feature vector is selected.

  If “YES” in the step S21, that is, if a passing object is passed, the moving route is recalculated in a step S25. That is, the CPU 12a refers to the map data 304a and calculates a movement route from the current position of the moving body 10 to the destination. In step S27, the movement is continued toward the destination along the movement route recalculated in step S25, and the process returns to step S5 shown in FIG.

  FIG. 17 is a flowchart showing the target position calculation process shown in step S15 of FIG. As shown in FIG. 17, when starting the target position calculation process, the CPU 12a acquires the side distance (t + 1) in step S51 using the class average vector obtained in step S14 as described above.

  In the next step S53, the predicted position (point N in FIGS. 11 and 12) of the passing time t + 1 is calculated. Predicted position data 304 i for the calculated predicted position is stored in the data storage area 304. However, the current time is t. The calculation of the predicted position is as described in the first stage.

  In the subsequent step S55, a straight line (L3 in FIG. 12) including the position where the moving body 10 should move is calculated from the predicted position to be passed. The calculation of the straight line L3 is as described in the second stage.

  In step S57, the position on the straight line where the moving body 10 can reach time t + 1 is calculated as the target position (point K in FIG. 12), and the process returns to the movement control process. The calculation of the target position in step S57 is as described in the third stage.

  According to this embodiment, the feature vector of the avoidance action of the passing object is acquired in advance, the relative position after the movement of the passing object is predicted based on the feature vector, and the target position of the moving object is calculated from the predicted relative position and the destination. Therefore, the moving body can be moved so as to cause the passing object to perform a normal avoidance action. That is, appropriate movement control can be performed. Therefore, it is safe and can give a sense of security to the person who is passing. Moreover, since the moving body is not rotated more than necessary, the moving body can be moved efficiently.

  In addition, as described in this embodiment, there are various passing objects in an environment to which a moving body that moves autonomously is applied. As described above, since the passing object includes other wheelchairs and pedestrians, there are various ways of passing when the passing object and the moving object pass. As described above, since a cluster of feature vectors corresponding to each avoidance action is generated for various passing objects, movement at the time of passing can be accurately predicted. For this reason, it is particularly effective in an environment where various passing objects exist.

  In this embodiment, the case has been described where the object is a passing object that moves so as to face the moving body. However, the present invention is not limited to this. As described above, the measurement device detects a passing candidate from a result measured by a distance sensor installed in the environment, and determines a passing target. For this reason, even if a pedestrian as a passing target (passing candidate) cannot visually recognize (recognize) a moving body, such as a corner, the measurement device predicts that the pedestrian and the moving body pass each other. Can do. Therefore, a safe passing can be realized by controlling the moving body so as to avoid a pedestrian from the previous stage where the moving body faces the passing object.

  For example, in FIG. 18, in a T-shaped road with poor visibility, the moving body 10 goes straight from the lower direction (Y direction), and a passing object such as a pedestrian is from the right direction (Z direction) of the T-shaped road. The case of going straight will be described. In FIG. 18 (FIGS. 19A and 19B are the same), the sensor (distance sensor 104) provided in the environment is indicated by a circle.

  In a place where there is such a corner, a pedestrian as a passing object cannot recognize the presence of the moving object 10 because the field of view is blocked by the wall at an early stage of movement. Therefore, even if the distance (relative distance) between the moving object 10 and the passing candidate is shortened to some extent, the pedestrian as the passing candidate cannot recognize the moving object 10, so the passing object is not determined and is based on the lateral distance. Movement control is not executed.

  However, a plurality of distance sensors 104 are arranged in an environment where the moving body 10 autonomously travels, and a pedestrian or the like existing in the environment can be detected. That is, the measurement apparatus 100 can detect a passing candidate or determine a passing object.

  Therefore, in such a case, even if a passing object such as a pedestrian cannot recognize the moving body 10, by performing movement control based on the lateral distance, a collision with a passing object such as a pedestrian can be avoided safely. You may be able to pass each other.

  For example, when the moving body 10 moves in the Z direction on a T-shaped road as shown in FIG. 18, if the pedestrian as a passing target takes two different types of avoidance actions depending on the moving direction, Conceivable.

  As shown in FIG. 19A, when a pedestrian as a passing object moves in the X direction, it is predicted that the pedestrian takes a course to the right so as to avoid the moving body 10. This is because there are few cases to avoid on the left side in the same situation, and a prediction result reflecting it is obtained. Therefore, based on the movement control of the above-described embodiment, the moving body 10 can move in the Z direction while taking a course toward the right from the stage before recognizing each other.

  Further, when a pedestrian as a passing object moves in the Y direction, it is predicted that the pedestrian takes a course toward the right side or the left side in order to avoid the moving body 10. Here, FIG. 19B shows a case where the path is taken to the left. Even in such a case, by acquiring the feature vector in advance, it is possible to appropriately move the moving body 10 with respect to the passing target that takes a course to the right or left according to the movement control of the above-described embodiment. it can.

  In this embodiment, a computer that also functions as a movement control device is mounted on the moving body. However, the present invention is not limited to this. The computer may be provided outside the moving body, and only control data for movement control may be transmitted to the moving body. However, the control data means the calculated movement route data, the movement start and stop instructions, and the target position data in the case of passing each other.

  In this embodiment, an electric wheelchair is shown as an example of a moving body, but an autonomously traveling robot can also be used. As an example of this autonomously traveling robot, Robbie (registered trademark) developed by the present applicant can be used.

  Furthermore, in this embodiment, a measurement device including a distance sensor provided in the environment is provided so that the measurement device can detect and track a passing candidate and a passing target, and generate a passing target information to generate a moving object. However, the present invention is not limited to this. The distance data may be acquired from the measurement device, and the passing candidate and the passing target may be detected and tracked on the moving body side, and the passing target information may be generated.

  Furthermore, in this embodiment, the movement control process and the target position calculation process shown in FIGS. 15 to 17 are executed by the computer (CPU) of the moving body, but the present invention is not limited to this. These processes may be executed by a computer of the measuring device, and control data for movement control of the moving body may be transmitted to the moving body. In this case, the computer of the measuring device functions as a movement control device and a route calculation device for the moving body. However, as described above, the movement data means calculated movement route data, movement start and stop instructions, and target position data when passing each other. Accordingly, the moving body starts moving according to the movement start instruction from the computer of the measuring device, moves along the acquired moving route, and when the target position is given from the computer of the measuring device, the target moves regardless of the moving route. It moves toward the position and ends the movement according to the movement stop instruction from the computer of the measuring device. In addition, when the moving body acquires updated (recalculated) movement path data from the computer of the measuring device during movement, the moving body moves along the newly acquired movement path.

  In this embodiment, a plurality of classified feature vectors and their average vectors are stored, a feature vector that approximates the current feature vector of the passing object is obtained, and an average of the classes including the obtained feature vectors is obtained. Although the vector is used to determine the predicted position of the passing object, it is not necessary to be limited to this. For example, only the average vector of a plurality of feature vectors may be stored by classification, and the predicted position of the passing target may be obtained using an average vector that approximates the current passing target feature vector.

  It should be noted that the specific numerical values shown in this embodiment are merely examples, and should not be limited, and can be appropriately changed according to the product to be implemented.

DESCRIPTION OF SYMBOLS 1 ... Mobile body control system 10 ... Mobile body 12, 102 ... Computer 12a ... CPU
12b RAM
12c HDD
16 ... Operation levers 18a, 18b, 104 ... Distance sensors 20a and 20b ... Motor drivers 22a and 22b ... Motors 24a and 24b ... Encoders 32a and 32b ... Front wheels 34a and 34b ... Rear wheels 100 ... Measuring devices

Claims (3)

A mobile control device that controls the movement of a mobile that travels autonomously,
Wherein the feature vector for the moving object and passing each other each of avoidance behavior of an unspecified number of passing interest, clustering a set of relative distance feature vectors generated for each of the the mobile body and the unspecified number of passing object Storage means for storing in advance a database of a plurality of feature vectors of each class clustered and an average vector of the plurality of feature vectors for each relative distance;
Select a database corresponding to the relative distance between the moving object and the current passing object from the storage means, and obtain a class including a feature vector that approximates the current feature vector of the current passing object from the database, Position predicting means for predicting a relative position after the predetermined time of the current passing target with respect to the moving body at the predicted position after the predetermined time, using an average vector of the class;
Target position calculating means for calculating a target position of the moving body based on the relative position predicted by the position predicting means and the destination of the moving body; and the moving body at the target position calculated by the target position calculating means. Bei El control means for moving the mobile control unit. The feature vector for each of the avoidance behavior of the unspecified number of passing objects moving body and passing each other to autonomous, the set of feature vectors generated for each relative distance between the mobile body and the unspecified number of passing object A moving body control apparatus comprising a storage unit that performs clustering and stores in advance a database of a plurality of feature vectors of each class and an average vector of the plurality of feature vectors for each relative distance, and controls movement of the moving body A mobile control program executed by a computer,
In the processor of the computer,
Select a database corresponding to the relative distance between the moving object and the current passing object from the storage means, and obtain a class including a feature vector that approximates the current feature vector of the current passing object from the database, A position prediction step of predicting a relative position after the predetermined time of the current passing target with respect to the moving body of the predicted position after a predetermined time, using an average vector of the class;
A target position calculating step of calculating a target position of the moving body based on the relative position predicted in the position predicting step and a destination of the moving body; and moving the moving body to the target position calculated in the target position calculating step to perform the control step of the mobile control program. The feature vector for each of the avoidance behavior of the unspecified number of passing objects moving body and passing each other to autonomous, the set of feature vectors generated for each relative distance between the mobile body and the unspecified number of passing object A moving body control apparatus comprising a storage unit that performs clustering and stores in advance a database of a plurality of feature vectors of each class and an average vector of the plurality of feature vectors for each relative distance, and controls movement of the moving body A moving body control method comprising:
The computer of the mobile control device is:
(A) A class including a feature vector that approximates the current feature vector of the current passing target from the database by selecting a database corresponding to the relative distance between the moving object and the current passing target from the storage unit. And using the average vector of the class, predicting the relative position after the predetermined time of the current passing target with respect to the moving body of the predicted position after the predetermined time,
(B) calculating a target position of the moving body based on the relative position predicted in the step (a) and the destination of the moving body, and (c) adding the target position calculated in the step (b) to the target position A moving body control method including a step of moving a moving body.

Images