JP5638283B2 - Control device - Google Patents

Description

  The present invention relates to a robot control apparatus that causes a robot to execute a task involving interaction with an object.

  In recent years, various techniques have been proposed for causing a robot to perform a task that involves interaction with the object, such as holding the object by the robot and moving the object (for example, see Patent Document 1 below). ).

JP 2005-125462 A

  However, depending on the type of task, it takes a relatively long time to generate a target trajectory for executing the task, and there is a problem that it takes time to execute the task.

  On the other hand, in order to expedite task execution, it may be possible to divide the task into its execution stages and generate a target trajectory for each execution stage, but when the execution stage switches, the target in the next execution stage Until the trajectory is generated, the task is interrupted and the robot operation becomes discontinuous.

  Therefore, in view of the above points, the present invention has an object to provide a robot control device that can speed up the start of task execution and prevent the operation of the robot from becoming discontinuous. And

A control device according to a first aspect of the present invention performs a task divided into a plurality of execution stages involving contact between the limb and one or a plurality of objects on a robot including a base and a limb connected to the base. a control apparatus for a robot to a requested task generating means for generating a request trajectory of the robot at each of said plurality of execution stages, the request in each of the request the plurality of execution steps generated by the task generating unit based on the track, the robot is predicted and the outside world predicting means for predicting a predicted position and predicted orientation of the object in the initial future execution step of performing at least the following, by the external prediction means, the future execution phase from the predicted position and the predicted attitude of the early the object, you sequence position when the robot before Symbol future execution phase An action planning means for generating a target trajectory of the fine orientation, characterized in that it comprises a.

  According to the control device of the first aspect of the present invention, the position and orientation of the object in each execution stage of the task are predicted, and the time-series position of the robot in the execution stage to be executed by the robot based on the prediction. And generate posture. As a result, even if the task is divided into execution stages and the task execution start is accelerated, the robot action plan is created in the next execution stage when the execution stage switches, so the task It is possible to prevent the operation of the robot from being discontinuous and the operation of the robot from being discontinuous in time.

  In addition, the time-series position and orientation of the generated robot are in line with the predicted position and orientation of the target object at each execution stage of the task, so the robot position and orientation are spatially discontinuous. It can also be prevented.

  As described above, according to the control device of the first aspect of the present invention, it is possible to speed up the start of task execution and to prevent the operation of the robot from becoming discontinuous.

The control device of the second invention is the control device according to the first invention,
The behavior planning unit is configured to obtain a target position trajectory and a target posture trajectory of the robot from the predicted position and predicted posture of the object at the end of the execution stage of one of the tasks predicted by the external world predicting unit. Further, it is characterized in that a constraint condition for generating a time-series position and posture of the robot in an execution stage subsequent to the one execution stage is determined.

  According to the control device of the second aspect of the invention, the predicted position and posture of the target object at the end of each execution stage of the task, that is, the initial position and initial position of the target object in the next execution stage in which prediction has been performed are predicted. be able to. Then, a constraint condition for generating a time-series position and posture of the robot in the next execution stage is determined from the initial position and initial posture of the target object. Based on these constraints, the time-series position and orientation of the robot in the next execution stage can be generated in advance, and even when the task is divided for each execution stage, when the execution stage switches Since the robot action plan is created in the next execution stage, the task execution is not interrupted, and it is possible to prevent the robot operation from being discontinuous in time.

  Furthermore, since the generated time-series position and orientation of the robot are based on the predicted position and predicted posture of the target object at the end of each execution stage of the task, the robot position and orientation are changed when each execution stage of the task is switched. It is also possible to prevent the posture from becoming spatially discontinuous.

  Thus, according to the control device of the second aspect of the present invention, it is possible to speed up the start of task execution and prevent the robot operation from becoming discontinuous.

The control device of the third invention is the first or second invention,
The outside world predicting means is such that the predicted position and the predicted posture of the object in the execution stage of the one task predicted by the external world prediction means are different from the current position and attitude of the object in the one execution stage. when, from the position and orientation of the current of the object, characterized by predicting a predicted position and predicted orientation of the object in each execution phase of the execution stage subsequent of the 1.

  According to the control device of the third aspect of the present invention, when the predicted position and the predicted posture of the object in each execution stage of the task are different from the actual position and posture of the target object in each execution stage, the actual position of the target object The predicted position and predicted posture of the object in each subsequent execution stage are predicted based on the posture and the posture. Therefore, even if a change occurs in the predicted situation in the task execution stage, the robot executes next based on the predicted position and posture of the target object in each subsequent execution stage according to the change. A time-series position and posture of the robot in the execution stage are generated.

  As a result, even if a change occurs in the predicted situation, it takes time to generate a time-series position and posture of the robot according to the change in the situation, and the task execution of the robot is interrupted. It is possible to prevent the movement of the robot from becoming discontinuous.

  In addition, the time-series position and orientation of the generated robot are in line with the predicted position and orientation of the target object at each execution stage of the task, so the robot position and orientation are spatially discontinuous. It can also be prevented.

  As described above, according to the control device of the third aspect of the present invention, it is possible to prevent the execution of the task from being interrupted and causing the robot operation to become discontinuous even when the predicted situation changes.

FIG. 3 is a configuration explanatory diagram of the robot according to the present embodiment. FIG. 2 is a configuration explanatory diagram of a robot control device. The flowchart which shows the process in the control apparatus of a robot. Explanatory drawing which shows the outline | summary of the process in the control apparatus of a robot. Explanatory drawing which shows the processing content in the control apparatus of a robot. Explanatory drawing which shows the processing content in the control apparatus of a robot.

  Next, an embodiment of the robot control apparatus of the present invention will be described with reference to FIGS.

  First, as shown in FIG. 1, the robot control device includes a controller 100 and a support server 200 of an autonomous mobile robot R (hereinafter referred to as robot R).

  The robot R includes a base 10, a head 11 provided on the top of the base 10, left and right arms 12 extending from the left and right sides of the top of the base 10, and a hand provided at the tip of the arm 12. Part 14, left and right leg parts 13 extending downward from the lower part of base body 10, and foot part 15 attached to the tip part of leg part 13. As disclosed in Table 03/090978 and Table 03/090979, the robot R is configured so that a human shoulder joint, elbow joint, The arm portion 12 and the leg portion 13 can be bent and extended at a plurality of joint portions corresponding to a plurality of joints such as a wrist joint, a hip joint, a knee joint, and an ankle joint. The robot R can move autonomously by movement of the left and right leg portions 13 (or the foot portions 15) with repeated leaving and landing. The height of the head 11 can be adjusted by adjusting the inclination angle of the base 10 with respect to the vertical direction. Note that the moving device may be any device having a moving function, such as a wheeled mobile robot (automobile), in addition to the robot R that moves autonomously by the operation of the plurality of legs 13.

The head 11 is mounted with a pair of head cameras (CCD cameras) C _{1 that} are arranged side by side and directed to the front of the robot R. Mounted on the base 10 is a waist camera C ₂ that outputs an infrared laser beam (electromagnetic wave) to the detection area A (C ₂ ) at the lower front side of the robot R and outputs a signal corresponding to the input of the reflected light. Yes. The waist camera C ₂ measures the position of the object below the front of the robot R, measures the orientation or posture of the robot R based on recognition of the shape and posture of the mark attached to the floor surface, and a transport object such as a carriage. This is used for measuring the position or posture of the object to be transported based on the recognition result of the shape or posture of the mark attached to the object.

  The robot R includes a controller 100 constituted by an ECU or a computer (configured by a CPU, ROM, RAM, I / O, etc.) as hardware, and a communication device 140 (see FIG. 2). . A control program (software) is stored in the memory of the computer. The control program may be installed in the computer through a software recording medium such as a CD or DVD, but when the request signal is transmitted from the robot R to the server, the server downloads to the computer via the network or artificial satellite. May be.

As shown in FIG. 2, in addition to the input device 101, the controller 100 includes a head camera C ₁ , a waist camera C ₂ , a shoulder camera C ₃ , a rotary encoder 102, a gyro sensor 103, and a GPS receiver 104 mounted on the robot R. By controlling the operation of the actuator 1000 based on the output signal such as the above, the operation of the arm portion 12 and the leg portion 13 is controlled. The controller 100 controls the behavior of the robot R according to the generated trajectory transmitted from the support server 200 to the robot R.

  The input device (user terminal) 101 is a terminal composed of a personal computer including an input device such as a keyboard and a touch panel and a monitor such as a liquid crystal display, and is configured to be able to communicate with the controller 100. . The input device 101 can designate a task that can be executed by the robot R when the user operates it. As described above, the input device 101 is used as a user interface for the user to remotely start and stop the robot R by instructing the start, stop, task execution, etc. of the robot R, and to display the image of the head camera C1, etc. It can also be used to monitor its own operating status.

The head camera C ₁ , the waist camera C _{2, and the} like correspond to sensors for measuring the external state or environment of the robot R, such as the behavior state of an object or an external interference object.

  The rotary encoder 102, the gyro sensor 103, and the GPS receiver 104 correspond to sensors for measuring the internal state or behavior state of the robot R, and this includes the terminal voltage of a battery mounted on the robot R (not shown). Various sensors mounted on the robot R, such as a voltage sensor to be detected and a force sensor that outputs a signal corresponding to the floor reaction force acting on the leg 13, correspond to an internal state sensor or the like. Based on the detection values of these sensors, the controller 100 recognizes the internal state or behavior state of the robot R.

  The actuator 1000 includes a drive source such as an electric motor, a speed reducer provided between the output shaft of the drive source and the link constituting the arm portion 12 and the leg portion 13, and a flexible element such as an elastic member. Yes.

  Further, the controller 100 includes a request task generation unit 110, an image processing unit 120, a future external world prediction unit 130, a communication device 140, and a main control unit 150.

  The request task generation unit 110 has a function of recognizing a user operation performed by the input device 101, recognizes a request task specified by the user via the input device 101, and requests required to execute the request task Create a trajectory group. Here, the required trajectory group refers to a set of trajectories required at each stage when the task is roughly divided according to the execution stage. Then, the requested task generation unit 110 outputs the generated requested trajectory group to the future external world prediction unit 130 and the support server 200.

  The image processing unit 120 includes a stereo processing unit 121, an obstacle detection unit 122, and an object recognition unit 123. The image processing unit 120 processes images captured by the cameras C1 to C3, and the surroundings of the robot R from the captured images. Recognize surrounding obstacles and objects to understand the situation. The obstacle is a concept included in the object of the present invention in that it is necessary to avoid interference with the robot R as described later.

  The stereo processing unit 121 performs pattern matching on the basis of one of the two images taken by the left and right head cameras C1 and C1, calculates the parallax of each corresponding pixel in the left and right images, and generates a parallax image. The generated parallax image and the original image are output to the obstacle detection unit 122. This parallax represents the distance from the robot R to the photographed object.

  The obstacle detection unit 122 detects an obstacle in the captured image based on the data output from the stereo processing unit 101. Further, the obstacle detection unit 122 extracts an obstacle as a parallax image of only a predetermined distance range of the detected obstacle, and outputs the obstacle image to the object recognition unit 123.

  The object recognition unit 123 extracts the shape or the like of the object accompanied by the interaction when executing the task from the detected obstacle, and recognizes the object from the size, the shape, or the like.

  Note that the area, size, and shape of the obstacle detected by the obstacle detection unit 122 and the area, size, and shape of the object recognized by the object recognition unit 123 are output to the future external prediction unit 130. Is done.

  The future external world prediction unit 130 performs a task execution stage from the requested trajectory group generated by the requested task generation unit 110 and the current obstacles and objects recognized by the obstacle detection unit 122 and the object recognition unit 123. Estimate future obstacles and target positions according to the situation.

  That is, the future external world prediction unit 130 determines the position of the obstacle and the object after operating the robot R according to the trajectory required in the first stage when the task is roughly divided according to the execution stage from the present time. Etc. are estimated. Next, from this state (after execution of the first stage), the position and the like of the obstacle and the object after the robot R is operated according to the trajectory required in the second stage are estimated. Hereinafter, by repeating this estimation, the positions of obstacles and objects at each stage where the robot is operated according to the requested trajectory group generated by the required task generation unit 110 are estimated in advance.

  The main control unit 150 controls the plurality of actuators 1000 according to the generated trajectory transmitted from the support server 200 to the robot R, and causes the behavior of the robot R to follow the generated trajectory.

  A support server (comprising a CPU, ROM, RAM, I / O, etc.) 200 has a communication function with the robot R via a base station (not shown) and a communication network.

  The support server 200 includes an operation classification processing unit 210, a trajectory search management unit 220, and an interference determination unit 230.

  The motion classification processing unit 210 further classifies a series of operations in each execution stage of the requested trajectory group generated by the required task generation unit 110 as follows.

  A series of operations at each stage is performed as follows: (1) The robot R or the second object T2 integrated with the robot R transitions from the non-contact state to the contact state with respect to the first object T1 (for example, gripping the object) (1) The first motion section D1 which is the motion section, and (2) the robot R or the second object T2 integrated with the robot R transitions from the contact state to the non-contact state with respect to the first object T1 ( A second motion segment D2 that is a motion segment (releasing the object), and (3) a third motion segment D3 that is a motion segment that is continuous with one or both of the first motion segment D1 and the second motion segment D2. Divide into

  The trajectory search management unit 220 performs a trajectory search of a series of operations at each execution stage of the required trajectory group generated by the required task generation unit 110, an initial posture search unit 221, a terminal posture search unit 222, and an intermediate posture search. Part 223.

  The initial posture searching unit 221 performs a trajectory search for the second motion category D2 segmented by the motion segment processing unit 210.

  The terminal posture search unit 222 performs a trajectory search for the first motion category D1 segmented by the motion segment processing unit 210.

  The intermediate posture search unit 223 performs a trajectory search for the third motion category D3 segmented by the motion segment processing unit 210.

  The trajectory search by the trajectory search management unit 220 will be described later.

  The interference determination unit 230 is integrated with (A) the obstacle detected by the obstacle detection unit 122 of the image processing unit 120 and the object recognized by the object recognition unit 123, and (B) the robot R or the robot. Check for possible interference with the target object.

  Next, processing contents in the control device for the robot R of this embodiment will be described with reference to FIG.

  First, the controller 100 determines whether there is a task execution request from the user (FIG. 3 / STEP 10). When the user designates an execution task of the robot R via the input device 101 (YES in FIG. 3 / STEP 10), the required task generation unit 110 of the controller 100 creates a required trajectory group (FIG. 3 / STEP 20). .

  In parallel with this, processing by the main control unit 150 (FIG. 3 / STEP 30-) is executed, and details of processing by the main control means will be described later.

  Here, the main control unit 150 executes the processing from STEP 30 on in parallel with the processing from STEP 20 on. The main control unit 150 does not depend on the generation timing of the target trajectory by the processing from STEP 20 on. This is to cause the robot R to continuously execute tasks.

  Specifically, as shown in FIG. 4, the task for moving the two cups W21 and W22 on the table W1 to the tray W3 in order is designated as an execution task of the robot R. In this case, as schematically shown in FIG. 4, when the actual execution stage of the task is the i-th stage, the processing of STEP 20-by the controller 100 and the support server 200 causes the external world of the (i + 1) -th stage. Based on the predicted situation, a target trajectory of at least the (i + 1) -th stage robot R is generated in advance.

  That is, in the i-th stage, which is the actual execution stage of the task, the robot R grips the first cup W21 and moves to the tray W3. However, in the processing from STEP 20 to the controller 100 and the support server 200, The predicted position and predicted posture of the object and obstacle in the situation where the i-th stage has been completed in advance, that is, the initial position and initial posture of the object and obstacle in the (i + 1) -th stage are predicted. Since the first cup W21 is already accommodated in the tray W3 in the initial position and the initial posture of the object and obstacle in the (i + 1) th stage, from the predicted initial state, the first cup W21 is in the (i + 1) th stage. A target trajectory of the robot R is generated in advance.

  Hereinafter, the details of the processing from STEP 20 will be described in detail with reference to FIG.

  First, the requested task generation unit 110 divides a task execution stage into a first stage, a second stage,..., And an nth stage for each task designated by the user, as shown in FIG. A required trajectory group is generated (FIG. 3 / STEP 20).

  Here, the first stage, the second stage,..., The nth stage are determined in consideration of the processing unit by the trajectory search management unit 220. That is, in the trajectory search management unit 220, a series of operations from the second motion segment D2 (initial posture) to the first motion segment D1 (end posture) through the third motion segment D3 (intermediate posture) is used as a processing unit. The request task generation unit 110 divides this as one stage.

  The required trajectory group defines conditions for performing a posture search (FIG. 3 / STEPs 24 to 26) described later for each stage. That is, for each stage, the trajectory search conditions required as the target trajectory (target posture trajectory and target posture trajectory) of the robot R are defined.

  Next, the future external world prediction unit 130 determines the task from the requested trajectory group generated by the requested task generation unit 110 and the current obstacles and objects recognized by the obstacle detection unit 122 and the object recognition unit 123. A future external situation group is created in which the position of the future obstacle and the object according to the execution stage is estimated (FIG. 3 / STEP 21).

  Specifically, the future outside world prediction unit 130 creates a future outside world situation group in which the situation such as the position of the table W1, the cup W2, and the tray W3 after the first to third stages shown in FIG. 5 is estimated. For example, the future external world prediction unit 130 predicts the situation after the execution of the task from the current situation to the second stage (the situation shown in FIG. 5F in which the cup W2 is moved to the tray), and then the second The situation after the task is executed up to the third stage is predicted from the situation predicted at the end of the stage. Hereinafter, the situation up to the n-th stage is predicted in order.

More precisely, the future ambient prediction unit 130, from the position X at the end of the obstacle and object of the i stage (t _i) and orientation theta (t _i), (i + 1) th stage at the end of the obstacle And the position X (t _{i + 1} ) and posture θ (t _{i + 1} ) of the object are calculated based on the required trajectory (see STEP 20) in the (i + 1) -th stage. At this time, i is an integer of 0 ≦ i ≦ n−1, and i = 0 indicates an initial state of the robot R (initial position X (t ₀ ) and initial posture θ (t ₀ )).

  Next, the support server 200 determines whether or not there is a trajectory generation request from the controller 100 (FIG. 3 / STEP 22).

  The presence / absence of a trajectory generation request is determined based on whether or not the requested trajectory group created by the requested task generation unit 110 and the future external world situation group created by the future external world prediction unit 130 are transmitted to the support server 200.

  When there is a trajectory generation request (YES in FIG. 3 / STEP 22), the motion classification processing unit 210 divides a series of operations in each execution stage of the requested trajectory group according to the motion classification (FIG. 3 / (STEP 23). On the other hand, when there is no trajectory generation request (NO in STEP 3 in FIG. 3), this process ends.

  Specifically, in the first stage shown in FIGS. 5A and 5B, the motion classification processing unit 210 first sets the cup W2 as the first object T1 and the robot R does not move against the first object. The first operation section D1, which is an operation section that transitions from the contact state to the contact state, is divided (see FIG. 5B). Then, the motion section processing unit 210 sets the motion section that follows the first motion section D1 in the first stage as the third motion section D3 (see FIG. 5A).

  Here, the starting point of the first motion section D1 shown in FIG. 5B is defined by the initial posture and the initial position of the motion of gripping the cup that is the first object. For example, the initial posture and the initial position of the robot R that can hold the cup as the first object by moving the arm portion 12 and the hand portion 14 are the starting points of the first motion section D1. The same applies to the start point of the second operation section D2.

  Next, in the second stage shown in FIGS. 5C to 5E, the motion classification processing unit 210 first sets the cup W2 integrated with the robot R as the second object T2 and the table W1 as the second object. As the first object T1, the robot R and the cup W2 (T2) integrated with the robot R divide the second motion section D2, which is the motion section in which the table W1 (T1) transitions from the contact state to the non-contact state. (See FIG. 5C). Further, the cup W2 integrated with the robot R is defined as the second object T2, and the tray W3 is defined as the first object T1, and the robot R and the cup W2 (T2) integrated with the tray W3 ( For T1), the first motion section D1 that is the motion section that transitions from the non-contact state to the contact state is divided (see FIG. 5E). Then, in the second stage, an operation segment that is continuous with the second operation segment D2 and the first operation segment D1 is defined as a third operation segment D3 (see FIG. 5D).

  Similarly, in the third stage shown in FIG. 5F, the motion classification processing unit 210 first sets the cup W2 as the first object T1 and the robot R is not in contact with the cup W2 (T1). The second motion section D2, which is the motion section that transitions to the contact state, is divided (see FIG. 5D). Then, the motion category processing unit 210 sets the motion category (returning to the basic posture) following the second motion category D2 in the third stage as the third motion category D3 (not shown).

  Next, the trajectory search management unit 220 searches the trajectory of the robot R at each stage of the first to third stages, and generates the time-series position and orientation of the robot R at each stage of the task (FIG. 3). / STEP 24-27).

  Specifically, for each of the first to third stages, the initial posture search unit 221 performs a trajectory search for the second motion section D2 (FIG. 3 / STEP 24), and the terminal posture search unit 222 performs the first step. The trajectory search for the motion section D1 is performed (FIG. 3 / STEP 25), and the intermediate posture search unit 223 performs the trajectory search for the third motion section D3 (FIG. 3 / STEP 26).

  Here, the trajectory search by the trajectory search management unit 220 searches for a trajectory that is generated by the required task generation unit 110 and that satisfies the required trajectory required at each execution stage of the task. Specifically, for each part of the robot R, a target position trajectory as a time series of target positions of representative points representing spatial positions and a target attitude trajectory as a time series of target postures are searched. Further, when the robot R and the second object T2 are integrated, a target position trajectory as a time series of target positions of representative points representing the spatial position of the second object, and a target posture Search with the target posture trajectory as a time series.

  At this time, as shown in FIG. 6, the trajectory search management unit 220, for the first object T1 that involves interaction in the task, determines the degree of freedom of rotation and translation according to the shape of the first object T1. Either or both are given to search for a target position trajectory and a target posture trajectory.

  Specifically, as shown in FIG. 6A, when the shape of the first object T1 is a circle, the target position trajectory and the target posture trajectory are searched by giving a degree of freedom of rotation with respect to the central axis. .

  On the other hand, as shown in FIG. 6B, when there is no position dependency in the axial direction as in the case where the shape of the first object T1 is a cylindrical shape, a translational freedom is given in the axial direction, A target position trajectory and a target posture trajectory are searched.

  In the trajectory search by the trajectory search management unit 220 described above, in the trajectory search for the second motion category D2 performed by the initial posture search unit 221, the robot R and the second object T2 integrated with the robot R, the first The determination of interference with the object T1 is omitted.

  For example, in the second motion section D2 shown in FIG. 5C, the interference determination between the robot R and the cup W2 that is the second object T2 integrated with the robot R and the table W1 that is the first object T1. The interference determination by the unit 230 is omitted. Further, in the second motion section D2 shown in FIG. 5 (f), the interference determination by the interference determination unit 230 between the robot R and the cup W2 that is the first object T1 is omitted.

  Further, in the trajectory search for the first motion section D1 performed by the terminal posture search unit 222, the interference determination between the second object T2 integrated with the robot R and the robot R and the first object T1 is omitted. .

  For example, in the first motion section D1 illustrated in FIG. 5A, the interference determination by the interference determination unit 230 between the robot R and the cup W2 that is the first object T1 is omitted. Further, in the first motion section D1 shown in FIG. 5E, interference by the interference determination unit 230 between the cup W2 that is the second object integrated with the robot R and the tray W3 that is the first object T1. Judgment is omitted.

  Thereby, in 1st operation | movement division D1 and 2nd operation | movement division D2, it can suppress that the calculation processing load by the interference determination with the 1st target object T1 with the interaction in a task increases.

  In the first motion section D1 and the second motion section D2, the first object T1 is already in contact with the robot R or the second object T2 integrated with the robot R, or will come into contact therewith. Therefore, even if the interference determination with the first object T1 is omitted, the robot R can execute the task if interference with other objects other than the first object T1 can be avoided.

  On the other hand, in the trajectory search for the third motion category D3 performed by the intermediate posture search unit 223, the robot R and the second object T2 integrated with the robot R and all obstacles including the first object T1 Is determined.

  Thereby, in 3rd operation division D3 other than the 1st and 2nd operation division, by performing interference judgment with the 1st target object T1, it is with other obstacles including the 1st target object T1. A series of operations for executing a task can be realized while avoiding interference.

  Next, the trajectory search management unit 220 sets the target trajectory (target position trajectory and target posture trajectory, the same applies hereinafter) of the second motion category D2 by the initial posture search unit 221 and the first motion category D1 of the terminal posture search unit 222. It is determined whether or not all target trajectories have been generated between the target trajectory and the target trajectory of the second motion section D3 that connects these sections by the intermediate posture search unit 223 (FIG. 3 / STEP 27).

  If all target trajectories have been generated (YES in FIG. 3 / STEP 27), the trajectory search management unit 220 records the generated target trajectory in a memory (not shown) or the like (FIG. 3 / STEP 28), and STEP 23 Return to Thereby, the processing after STEP23 is repeatedly executed for the execution stage of the next task.

  On the other hand, when all the target trajectories have not been generated (NO in STEP 27 in FIG. 3), the process returns to STEP 24 and the processes after STEP 24 are executed again.

  The above is a series of processes for generating the target trajectory according to the requested task.

  Next, processing (FIG. 3 / STEP 30 to STEP 34) performed by the main control unit 150, which has been described later, will be described.

  As described above, the main control unit 150 does not control the actuator 1000 so as to follow the target trajectory at the timing when the target trajectory is generated by STEPs 24 to 26, but is recorded in the memory or the like at STEP28. If the target trajectory is accumulated, it is sequentially processed.

  Thereby, the task execution can be performed by the robot R regardless of the generation timing of the target trajectory, and the execution operation of the robot R can be given temporal continuity. Furthermore, since the generated target trajectory is in accordance with the predicted position and predicted posture of the object in each execution stage of the task, the operation of the robot R controlled based on the target trajectory is spatially discontinuous. It can also be prevented.

  First, the main control unit 150 acquires the requested trajectory group created by the requested task generation unit 110 (FIG. 3 / STEP 30).

  Next, the main control unit 150 determines whether or not there is a trajectory follow-up request (FIG. 3 / STEP 31).

  When there is no trajectory follow-up request (NO in FIG. 3 / STEP 31), the main control unit 150 ends this process, and when there is a trajectory follow-up request (YES in FIG. 3 / STEP 31), trajectory search. It is determined whether or not a trajectory has been generated by the management unit 220 (FIG. 3 / STEP 32).

  When the target trajectory generated by the trajectory search management unit 220 is recorded in a memory or the like (not shown) (YES in FIG. 3 / STEP 32), the main control unit 150 controls the actuator 1000 of the robot R. Thus, the robot R is caused to follow the generated target trajectory (FIG. 3 / STEP 33).

  In this case, when there is a pressing operation on the first target T1 in the target trajectory, the robot R is controlled to perform the pressing operation by controlling the actuator 1000 of the target portion of the robot R.

  Next, the main control unit 150 determines whether or not the robot R follows the target trajectory (FIG. 3 / STEP 34).

  When the robot R follows the target trajectory (YES in FIG. 3 / STEP 34), the main control unit 150 returns to STEP 31 and determines whether there is a next trajectory follow-up request. On the other hand, when the robot R does not follow the target trajectory (NO in FIG. 3 / STEP 34), it is determined whether or not the target trajectory cannot be followed by the actuator 1000 (FIG. 3 / STEP 35).

  If the target track can be followed (NO in STEP 3 / NO in STEP 35), the main control unit 150 returns to STEP 34 and repeats this process until the robot R follows the target track. On the other hand, if it is impossible to follow the target trajectory (YES in FIG. 3 / STEP 35), it is determined that the current task execution status is different from the target trajectory, and the process returns to STEP 21 and the current actual status. A future external world situation group is created based on (the current obstacle or object recognized by the image processing unit 120), and the target trajectory is regenerated from the current actual situation and the future external world situation group (FIG. 3). / STEP 21-).

  As a result, even when a change occurs in the predicted situation, the predicted position and predicted posture of the object in each subsequent execution stage are updated, and the robot R in each execution stage to be executed by the robot R from now on A series of target position trajectories and target attitude trajectories are regenerated. Therefore, it is possible to prevent the task execution of the robot from being interrupted because it takes time to generate the target trajectory due to a change in the situation, and it is possible to prevent the operation of the robot from becoming discontinuous.

  On the other hand, if the target trajectory generated by the trajectory search management unit 220 is not recorded in a memory (not shown) or the like (NO in STEP 31) (NO in STEP 31 in FIG. 3), the trajectory search management unit 220 waits for a predetermined time. At the timing when the trajectory is generated (FIG. 3 / NO in STEP 34), the processing after STEP 32 is executed. On the other hand, when the waiting time has elapsed and time is up (YES in FIG. 3 / STEP 34), it is assumed that the target trajectory does not exist or that the robot R has completed the tracking of all the target trajectories that have already been generated. A series of processing ends.

  The above is the processing contents in the control device for the robot R of the present embodiment. As described above, according to such a control device, the future situation in each execution stage of the task is predicted in advance, and the target trajectory is generated in advance. By doing so, when the execution stage is switched, the target trajectory of the robot R in the next execution stage is generated, so the task execution is not interrupted and the robot operation becomes discontinuous. This can be prevented.

  In the above-described embodiment, a configuration is described in which the motion classification processing unit 210, the trajectory search management unit 220, and the interference determination unit 230 are provided in the support server 200 to perform distributed processing in order to reduce the processing load on the controller 100 of the robot R. However, the present invention is not limited to this. In other words, the controller 100 may include a part or all of the processing units 110 to 130 and 150 of the controller 100 and the processing units 210 to 230 of the support server, and the support server may include the remaining processing units. Good. For example, all the processing units may be provided in the controller 100 and the support server 200 may not be provided.

  Moreover, although the said embodiment demonstrated the structure by which a task is designated by the input device 101 which is an information terminal, it is not limited to this. For example, the robot R may be configured to recognize a user instruction by a microphone or the like.

  Furthermore, in the above embodiment, the request task generation unit 110 divides a series of operations for executing a task into each execution stage (first stage, second stage,..., Nth stage) of the task. However, it is not limited to this.

  That is, in this embodiment, the task is divided into each execution stage, and the future situation realized by the operation of the robot R in each execution stage (the target at the end of each execution stage predicted from the required trajectory group in STEP 20) The target posture trajectory and target position trajectory of the robot R in the next execution stage after the predicted execution stage are generated from the object position and attitude), but the task may not be divided into execution stages. Good.

  In this case, the future external world prediction unit 130 sequentially predicts the time-series predicted position and predicted posture of the target object according to the progress of the task, and the trajectory search management unit 220 based on the predicted position and predicted posture. The robot R is configured to sequentially generate a target posture trajectory and a target position trajectory. Thereby, it is possible to prevent the operation of the robot from being discontinuous temporally and spatially as in the above embodiment.

R ... Robot, 100 ... Controller, 101 ... Input device, 110 ... Request task generation unit, 120 ... Image processing unit, 130 ... Future external world prediction unit (external world prediction means), 150 ... Main control unit, 200 ... Support server, 210 ... motion classification processing unit, 220 ... trajectory search management unit (behavior planning unit), 221 ... initial posture search unit, 222 ... end posture search unit, 223 ... intermediate posture search unit, 230 ... interference determination unit (interference determination unit), D1 ... 1st motion section, D2 ... 2nd motion section, D3 ... 3rd motion section, T1 ... 1st object, T2 ... 2nd object, W1 ... Table, W2 (W21, W22) ... Cup, W3 ... tray.

Claims (3)

A robot control apparatus that causes a robot including a base and a limb connected to the base to perform a task divided into a plurality of execution stages involving contact between the limb and one or more objects. ,
Request task generating means for generating a required trajectory of the robot in each of the plurality of execution stages;
Based on the requested trajectory in each of the plurality of execution stages generated by the requested task generation means , the robot predicts the predicted position and predicted posture of the target object at least in the initial stage of the future execution stage to be executed next. Outside world prediction means,
Predicted by the external prediction means, generating a target trajectory of the time-series position and posture of the robot from the predicted position and the predicted attitude, before Symbol future execution phase of the object in the initial of the future execution phase And a behavior planning means. The control device according to claim 1,
The behavior planning unit is configured to obtain a target position trajectory and a target posture trajectory of the robot from the predicted position and predicted posture of the object at the end of one execution stage of the task predicted by the external world predicting unit. And determining a constraint condition for generating a time-series position and posture of the robot in an execution stage subsequent to the one execution stage. The control device according to claim 1 or 2,
The outside world predicting means is such that the predicted position and posture of the object in one execution stage of the task predicted by the outside world prediction means are different from the current position and attitude of the object in the one execution stage. In this case, the control device predicts the predicted position and the predicted posture of the target object in each execution stage after the one execution stage from the current position and posture of the target object.

Images