CN110653819B - System and method for generating kicking action of humanoid robot - Google Patents

System and method for generating kicking action of humanoid robot Download PDF

Info

Publication number
CN110653819B
CN110653819B CN201910912950.4A CN201910912950A CN110653819B CN 110653819 B CN110653819 B CN 110653819B CN 201910912950 A CN201910912950 A CN 201910912950A CN 110653819 B CN110653819 B CN 110653819B
Authority
CN
China
Prior art keywords
football
humanoid robot
kicking
orientation
subsystem
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910912950.4A
Other languages
Chinese (zh)
Other versions
CN110653819A (en
Inventor
贾文川
王超君
马书根
袁建军
孙翊
蒲华燕
鲍晟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Shanghai for Science and Technology
Original Assignee
University of Shanghai for Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Shanghai for Science and Technology filed Critical University of Shanghai for Science and Technology
Priority to CN201910912950.4A priority Critical patent/CN110653819B/en
Publication of CN110653819A publication Critical patent/CN110653819A/en
Application granted granted Critical
Publication of CN110653819B publication Critical patent/CN110653819B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1602—Programme controls characterised by the control system, structure, architecture
    • B25J9/1605—Simulation of manipulator lay-out, design, modelling of manipulator
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J11/00—Manipulators not otherwise provided for
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
    • B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning

Landscapes

  • Engineering & Computer Science (AREA)
  • Robotics (AREA)
  • Mechanical Engineering (AREA)
  • Automation & Control Theory (AREA)
  • Manipulator (AREA)

Abstract

The invention discloses a system and a method for generating a kicking action of a humanoid robot. The kickball action generating system of the humanoid robot comprises a football orientation recognition training subsystem, a football orientation real-time measuring subsystem, a target orientation real-time obtaining subsystem, a kickball action track planning subsystem and a kickball action executing subsystem; the humanoid robot is provided with a visual system for acquiring a scene image, and the visual system has the capability of adjusting the observation angle. And the football orientation recognition training subsystem is used for constructing a football orientation mathematical model of the correlation between the relative position and the relative direction and the image information according to the image information acquired by the vision system in the humanoid robot under the condition that the relative position and the relative direction between the football and the humanoid robot are known. The invention can quickly and accurately identify the orientation of the football and generate the kicking action, has low requirement on hardware, and is suitable for being popularized and used in humanoid robot systems with different actual specifications.

Description

System and method for generating kicking action of humanoid robot
Technical Field
The invention belongs to the technical field of robots, and particularly relates to a system and a method for generating kicking actions of a humanoid robot.
Background
Football has the reputation of "world first sport", being the single sport with the most influential. Let the robot kick the football like a human and eventually defeat the human football team, as a real target of a robotic world cup (RoboCup) event. Although the role of the robot individual in the football game is changeable in real time, like the human football players, the players who are in the attack stage and in the ball control state are most concerned, and the processing of the ball has direct influence on the game situation and can effectively reflect the technical and tactical abilities of the players. For a robot in an attack phase and in a ball control state, the problem to be solved includes "where is the ball? "," where to kick? "," how to kick? ".
Therefore, it is an urgent need to solve the problem of providing a system and method for generating kicking motions of a humanoid robot.
Disclosure of Invention
In view of the above, the present invention provides a system and a method for generating a kicking motion of a humanoid robot, which aims to generate a kicking motion more quickly, quickly and accurately in response, identify a football direction and generate a kicking motion quickly and accurately, and are suitable for being popularized and used in humanoid robot systems of different actual specifications.
In order to achieve the purpose, the invention adopts the following technical scheme:
a kickball action generating system of a humanoid robot comprises a football orientation recognition training subsystem, a football orientation real-time measuring subsystem, a target orientation real-time obtaining subsystem, a kickball action track planning subsystem and a kickball action executing subsystem;
the football orientation recognition training subsystem is used for constructing a football orientation mathematical model of the incidence relation between the relative position and the relative direction and the image information according to the image information acquired by a vision system in the humanoid robot under the condition that the relative position and the relative direction between the football and the humanoid robot are known;
the football orientation real-time measuring subsystem is used for calculating in real time to obtain the relative position and the relative direction between the football and the humanoid robot according to the established football orientation mathematical model and the image information acquired by the vision system in the humanoid robot, and recording the relative position and the relative direction as the football orientation;
the target direction real-time acquisition subsystem is used for calculating on line to obtain the relative position and the relative direction between the expected football and the humanoid robot after the football is kicked by directly appointing a target direction, or automatically identifying the goal direction, or automatically planning the pass direction, and recording the relative position and the relative direction as the target direction;
the kickball action trajectory planning subsystem is used for planning kickball actions suitable for the humanoid robot according to the football direction and the target direction which are obtained currently, and recording the kickball actions as kickball action trajectories;
and the kickball action execution subsystem is used for enabling the humanoid robot to execute kickball actions according to the currently obtained kickball action track.
Preferably, the football is a spherical rollable object, and the number of the rollable object in the field is 1; the relative position between the football and the humanoid robot is the relative position between the center of the football and the humanoid robot; the relative direction between the football and the humanoid robot is the relative direction between the center of the football and the humanoid robot;
the football field consists of a football field area, a goal and a field mark, wherein the field mark comprises an entity boundary, a color marking line, a color area, an information display board and an invisible electronic mark;
the kicking motion refers to a series of body motions that the humanoid robot needs to perform in order to strike the soccer ball from the soccer ball orientation to the target orientation.
Preferably, the humanoid robot is provided with a visual system for acquiring a scene image, and the visual system has the capability of adjusting the observation angle;
the two legs of the humanoid robot respectively have at least 6 degrees of freedom of movement, and the feet at the tail ends of the legs are of hard structures with certain shapes;
the humanoid robot is provided with a robot position coordinate system, the origin of the robot position coordinate system is located at a projection point of the center position of the waist of the humanoid robot body on the ground, the x-axis of the robot position coordinate system points to the direction of the right front of the humanoid robot body in the positive direction, and the z-axis of the robot position coordinate system points to the opposite direction of gravity in the positive direction.
Preferably, the vision system realizes the adjusting capability of the observation angle including the pitch direction and the yaw direction by adjusting part of body joints of the humanoid robot and/or adjusting the setting of the vision system.
A method for generating kicking actions of a humanoid robot comprises the following steps:
step S1, carrying out recognition training on a football orientation recognition training subsystem to obtain the football orientation mathematical model;
step S2, the vision system acquires the current scene image, inputs the image information into the football orientation real-time measuring subsystem, and acquires the football orientation;
step S3, the target azimuth real-time acquisition subsystem acquires the target azimuth;
step S4, the kickball action track planning subsystem obtains the kickball action track;
step S5, the kicking action execution subsystem executes the kicking action;
the step S1 is a training stage, and is executed once in advance in a non-kicking state;
the steps S2-S5 are kicking steps, and are repeatedly executed in a kicking state.
Preferably, the process of performing recognition training on the soccer orientation recognition training subsystem in step S1 is as follows:
step S11, adjusting the gesture of the humanoid robot to enable the distance between the visual system and the ground to be h; adjusting a pitching observation angle of the vision system to enable a rotation angle between a projection of a forward direction of the vision system in an x-z plane of the robot azimuth coordinate system and an x axis to be theta _ pitch; adjusting the yaw direction observation angle of the vision system, and enabling the rotation angle between the projection of the forward direction of the vision system in the x-y plane of the robot azimuth coordinate system and the x axis to be theta _ yaw;
step S12, keeping the current position and posture of the humanoid robot unchanged, setting a group of distributed marking points in the visible range of the visual system, and measuring and recording the relative position and relative direction between each marking point and the humanoid robot;
step S13, placing the football in the positions of the marking points in the step S12 in sequence, enabling a vision system to acquire an image for each position, further processing the image to obtain the pixel coordinates of the center of the football sphere in the image, and recording the data of the current h, theta _ pitch and theta _ yaw, the position data of the marking points and the corresponding pixel coordinate data in the image;
step S14, returning to step S11, adjusting the values of h, θ _ pitch, and θ _ yaw, and repeatedly executing steps S12 to S13;
step S15, establishing a neural network including an input layer, a hidden layer and an output layer;
and step S16, dividing the data recorded in the step S13 into a training set and a testing set, training the neural network in the step S15 according to the training set, and constructing a mapping relation between the football orientation and football image pixel coordinates, namely the football orientation mathematical model, under the condition that the values of h, theta _ pitch and theta _ yaw are known.
Preferably, the forward propagation model of the input layer and the hidden layer of the neural network established in step S15 is yl＝f(ul)＝f(Wlyl-1+bl) Where f (-) is the ReLU activation function, WlAnd blRespectively the weight and bias matrix to be optimized in the l-th layer, ylIs the output of the l-th layer; the forward propagation model of its output layer is yl＝ul＝Wlyl-1+bl(ii) a Loss function using mean square error
Figure BDA0002215254180000051
The optimization of weight and bias adopts gradient descent method
Figure BDA0002215254180000052
Preferably, the step S14 is not essential, and if the step S14 is skipped, the values of h, θ _ pitch, and θ _ yaw are all unique.
Preferably, the process of obtaining the football orientation by the football orientation real-time measuring subsystem in step S2 is as follows:
processing the image information acquired by the vision system, detecting the position of the football in the image, and acquiring the pixel coordinates of the center position of the football in the image; and jointly inputting the obtained pixel coordinates and the values of the current h, theta _ pitch and theta _ yaw of the humanoid robot into the football orientation mathematical model, and solving and outputting the football orientation.
Preferably, the process of obtaining the kicking trajectory by the kicking trajectory planning subsystem in step S4 is as follows:
step S41, obtaining or measuring the boundary contour shape of the humanoid robot foot in advance, obtaining a fitting function of the foot boundary contour by using a curve fitting method, marking the fitting function as a foot contour BoundryX (y), and obtaining the foot contour BoundryX (y) by differentiating to obtain the foot contour BoundryX (y)
Figure BDA0002215254180000054
Step S42, reasonably selecting kicking legs according to the football direction, the target direction, the current positions and the supporting states of the two legs of the humanoid robot;
step S43, selecting the relative position p in the orientation of the footballb＝[pbx,pby,pbz]TAnd a relative position p in the target positionT＝[pTx,pTy,pTz]TCalculating the direction of kicking the ball
Figure BDA0002215254180000053
Based on the principle that the kicking contact point position on the football is the same as the kicking direction theta _ kick along the normal direction of the spherical surface, the kicking contact point position p on the football is calculatedk；
Step S44, making the outline normal direction of the kicking point position of the humanoid robot foot and the kicking direction theta _ kick be the same as the principle
Figure BDA0002215254180000061
Calculating the kicking point position p of the foot of the humanoid robotr；
Step S45, obtaining a kicking point p of the foot of the humanoid robot by using a polynomial fitting methodrThe kicking action track of the football needs to pass through the kicking contact point p on the footballk。
The invention has the beneficial effects that:
(1) the robot can finish the ball recognition training and the actual combat process detection under the condition that the robot only has a vision system, and the training process is simple and reliable to use.
(2) The method is easy to realize programming and is suitable for being used as a basic algorithm of main attacking behaviors in robot football games such as attack robot shooting action generation, pass action generation and the like; and the modules are clearly divided, so that the function of a single module can be optimized according to the requirement, and the algorithm is also suitable for integrating the algorithm into a higher-layer robot football control framework.
(3) The kicking action is quick in response, and a way for realizing quick response is provided for the robot player.
(4) The method has the advantages of quick calculation and low requirement on hardware, and is suitable for application and popularization in the special humanoid robot products for economy, teaching and competition.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic view of a scene that a robot plays a football provided by the invention.
Fig. 2 is a schematic diagram of a structure and a kicking track of a humanoid robot provided by the invention.
Fig. 3 is a schematic diagram of a kicking motion generating system of a humanoid robot provided by the invention.
FIG. 4 is a schematic diagram of a training scenario of a football orientation recognition training subsystem according to the present invention.
FIG. 5 is a training flow diagram of a football orientation recognition training subsystem provided by the present invention.
Fig. 6 is an operation flow chart of a kicking trajectory planning subsystem provided by the invention.
Fig. 7 is a schematic diagram of a kicking trajectory planning subsystem provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, in a scene that a robot kicks a football, the scene comprises a plurality of humanoid robots with the capability of playing football games, a football 7 and a football field; the court is composed of a football field area 81, a goal 82 and field marks 83, wherein the field marks 83 comprise entity boundaries, color marked lines, color areas, information display boards and invisible electronic marks. Among a plurality of humanoid robots having a capability of playing a soccer game, a robot in an attack stage and in a ball control state is marked as a humanoid robot 6; a soccer ball 7, which is a spherical rollable object and the number thereof in the field is 1; the relative position between the football 7 and the humanoid robot 6 is the relative position between the center of the football 7 and the humanoid robot 6; the relative direction between the soccer ball 7 and the humanoid robot 6 is the relative direction between the center of the soccer ball 7 and the humanoid robot 6.
As shown in fig. 2, the humanoid robot 6, which may be in the form of 1 NAO robot in particular, has a vision system 61 for acquiring images of a scene, and the vision system 61 has the capability of adjusting the observation angle, which is relative to a fixed geodetic coordinate system, and typically includes yaw, pitch, and roll angles. The vision system 61 realizes the capability of adjusting the observation angle including the pitch direction and the yaw direction by adjusting a part of the body joint of the humanoid robot 6 and/or by adjusting the setting of the vision system 61 itself. For example, the pitch angles of the head and the visual system 61 located on the head may be indirectly adjusted by adjusting the kinematic joint of the waist or neck, or the angle of view in the pitch direction may be adjusted by adjusting the camera parameters built in the visual system 61.
The two legs of the humanoid robot 6 are respectively provided with at least 6 degrees of freedom of movement so as to realize the flexible movement capability required by the humanoid kicking action, and the foot part 62 at the tail end of the leg part is a hard structure with a certain shape. The kicking motion of the humanoid robot 6 refers to a series of body motions that the humanoid robot 6 needs to perform in order to strike the soccer ball 7 from the soccer ball bearing 91 to the target bearing 92.
The humanoid robot 6 has a robot orientation coordinate system 97, an origin of the robot orientation coordinate system 97 is located at a projection point of a waist center position of the humanoid robot 6 on the ground, an X-axis of the robot orientation coordinate system 97 points to a direction directly in front of the humanoid robot 6, and a Z-axis of the robot orientation coordinate system 97 points to a direction opposite to gravity.
As shown in fig. 3, the system for generating kicking motions of the humanoid robot comprises a football orientation recognition training subsystem 1, a football orientation real-time measuring subsystem 2, a target orientation real-time obtaining subsystem 3, a kicking motion trajectory planning subsystem 4 and a kicking motion executing subsystem 5.
The football orientation recognition training subsystem 1 is used for constructing a football orientation mathematical model 90 of the correlation between the relative position and the relative direction and the image information according to the image information acquired by the vision system 61 in the humanoid robot 6 under the condition that the relative position and the relative direction between the football 7 and the humanoid robot 6 are known.
And the football orientation real-time measuring subsystem 2 is used for calculating in real time to obtain the relative position and the relative direction between the football 7 and the humanoid robot 6 according to the established football orientation mathematical model 90 and the image information obtained by the vision system 61 in the humanoid robot 6, and recording the relative position and the relative direction as the football orientation 91.
And the target position real-time obtaining subsystem 3 is used for obtaining the relative position and the relative direction between the expected football 7 and the humanoid robot 6 after the football 7 is kicked through online calculation by directly specifying the target position, or automatically identifying the position of the goal 82 or automatically planning the pass position, and recording the relative position and the relative direction as the target position 92.
And the kickball action trajectory planning subsystem 4 is used for planning the kickball action suitable for the humanoid robot 6 according to the currently obtained football direction 91 and the target direction 92, and recording the kickball action trajectory as a kickball action trajectory 93.
And the kickball action execution subsystem 5 is used for enabling the humanoid robot 6 to execute kickball actions according to the currently obtained kickball action track 93.
As shown in fig. 3, the working method of the kicking motion generation system of the humanoid robot 6 includes the following steps:
step S1, carrying out recognition training on the football orientation recognition training subsystem 1 to obtain a football orientation mathematical model 90;
step S2, the vision system 61 acquires the current scene image, inputs the image information into the football orientation real-time measuring subsystem 2, and acquires the football orientation 91;
step S3, the target position real-time obtaining subsystem 3 obtains a target position 92;
step S4, the kickball action track planning subsystem 4 obtains a kickball action track 93;
in step S5, the kicking action execution subsystem 5 executes the kicking action.
Step S1 is a training phase, which is performed once in advance in a non-kicking state.
Step S2-step S5 are kicking stages, which are repeatedly executed in a kicking state.
As shown in fig. 4 and 5, in step S1 of the working method, the process of performing recognition training on the soccer orientation recognition training subsystem 1 is as follows:
step S11, adjusting the posture of the humanoid robot 6 to enable the distance between the vision system 61 and the ground to be h; adjusting the elevation direction observation angle of the vision system 61 such that the forward direction of the vision system 61 is at a rotation angle θ _ pitch between the projection 611 and the X-axis in the X-Z plane of the robot orientation coordinate system 97; the yaw-direction viewing angle of the vision system 61 is adjusted such that the forward direction of the vision system 61 is rotated by θ _ yaw from the X-axis in the projection 612 of the X-Y plane of the robot orientation coordinate system 97. In FIG. 4(a), there is an X 'axis parallel to the X axis, so both θ _ pitch and θ _ yaw are labeled next to the X' axis.
Step S12, keeping the current position and posture of the humanoid robot 6 unchanged, setting a set of distributed marker points 94 within the visible range of the vision system 61, and measuring and recording the relative position 941 and relative direction between each marker point 94 and the humanoid robot 6.
Step S13 is to place the soccer ball 7 at the positions of the marker points 94 in step S12 in sequence, and for each position, to cause the vision system 61 to capture an image, and further to process the image to obtain the pixel coordinates 95 of the center of the soccer ball 7 in the image. Recording data of current h, theta _ pitch and theta _ yaw, data of a position 941 where the mark point 94 is located and corresponding pixel coordinate 95 data in the image; as shown in fig. 4(a), the data of the position 941 where the marker 94 of the soccer ball 7 is located is (x _ ball, y _ ball); fig. 4(b) is a schematic diagram of an image acquired by the vision system 61, and the pixel coordinate 95 has a value of (x _ pixel, y _ pixel).
In step S14, the process returns to step S11, the values of h, θ _ pitch, and θ _ yaw are adjusted, and steps S12 to S13 are repeated.
In step S15, a neural network 96 including an input layer, a hidden layer, and an output layer is established.
Step S16, the data recorded in step S13 are divided into a training set and a test set, the neural network 96 in step S15 is trained, and the mapping relationship between the soccer orientation 91 and the pixel coordinates 95 of the soccer image, that is, the soccer orientation mathematical model 90, is constructed by the neural network 96 obtained after training is completed, under the condition that the values of h, θ _ pitch, and θ _ yaw are known. As shown in fig. 4, the football orientation mathematical model 90 is the inter-mapping relationship between (x _ ball, y _ ball) and (x _ pixel, y _ pixel).
In the neural network 96 established in step S15, the forward propagation model of the input layer and the hidden layer is yl＝f(ul)＝f(Wlyl-1+bl) Where f (-) is the ReLU activation function, WlAnd blRespectively the weight and bias matrix to be optimized in the l-th layer, ylIs the output of the l-th layer; the forward propagation model of its output layer is yl＝ul＝Wlyl-1+bl(ii) a Loss function using mean square error
Figure BDA0002215254180000111
The optimization of weight and bias adopts gradient descent method
Figure BDA0002215254180000112
Step S14 is not essential, and if step S14 is skipped, the values of h, θ _ pitch, and θ _ yaw are all unique.
In step S2 of the working method, the process of the football orientation real-time measuring subsystem 2 obtaining the football orientation 91 is as follows: processing the image information obtained by the vision system 61, detecting the position of the football 7 in the image, and obtaining the pixel coordinates 95 of the center position of the football 7 in the image; the obtained pixel coordinates 95 and the current values of h, theta _ pitch and theta _ yaw of the humanoid robot 6 are input into the football orientation mathematical model 90 together, and the football orientation 91 is solved and output.
As shown in fig. 6 and 7, the process of obtaining the kicking trajectory 93 by the kicking trajectory planning subsystem 4 in step S4 of the working method is as follows:
step S41, the foot boundary contour of the humanoid robot 6 is obtained or measured in advance, a curve fitting method is used for obtaining a fitting function of the foot boundary contour, the fitting function is recorded as a foot contour BoundryX (y), and the foot contour BoundryX (y) is differentiated to obtain
Figure BDA0002215254180000114
As shown in fig. 7, for the humanoid robot 6 with the specific model NAO, the outline of the foot is BoundaryX (y) -7523.85y obtained by measurement and the least square method4-1504.76y3-122.64y2-4.74y+0.03。
Step S42, reasonably selecting kicking legs according to the football direction 91, the target direction 92, and the current positions and support states of the legs of the humanoid robot 6.
Step S43, from the orientation 91 of the soccer ballRelative position pb＝[pbx,pby,pbz]TAnd relative position p in target position 92T＝[pTx,pTy,pTz]TCalculating the direction of kicking the ball
Figure BDA0002215254180000113
Based on the principle that the kicking contact point position on the football 7 is the same as the kicking direction theta _ kick along the normal direction of the spherical surface, the kicking contact point position p on the football 7 is calculatedk。
Step S44, making the outline normal direction of the kicking point position of the foot of the humanoid robot 6 and the kicking direction theta _ kick be the same as the principle
Figure BDA0002215254180000121
Calculating the kicking point position p of the foot of the kicking leg of the humanoid robot 6r。
Step S45, obtaining the kicking point p of the foot of the humanoid robot 6 by using a polynomial fitting methodrThe kicking track 93 of the football, the kicking track 93 needs to pass through the kicking contact point p on the football 7k. Further, kick the contact point pkAs a boundary, the kicking trace 93 can be divided into 2 segments, q1(t)＝a10+a11t+a12t2+a13t3And q is2(t)＝a20+a21t+a22t2+a23t3And q is1(t) and q2(t) at the kick contact point pkAnd (4) intersecting.
The invention aims to realize the generation of kickball actions with quicker, quicker and more accurate reaction, can quickly and accurately identify the orientation of the football and generate the kickball actions, and is suitable for being popularized and used in humanoid robot systems with different actual specifications.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (8)

1. A kickball action generating system of a humanoid robot is characterized by comprising a football orientation recognition training subsystem, a football orientation real-time measuring subsystem, a target orientation real-time obtaining subsystem, a kickball action trajectory planning subsystem and a kickball action executing subsystem;
the football orientation recognition training subsystem is used for constructing a football orientation mathematical model of the incidence relation between the relative position and the relative direction and the image information according to the image information acquired by a vision system in the humanoid robot under the condition that the relative position and the relative direction between the football and the humanoid robot are known;
step S11, adjusting the gesture of the humanoid robot to enable the distance between the visual system and the ground to be h; adjusting a pitching observation angle of the vision system to enable a rotation angle between a projection of a forward direction of the vision system in an x-z plane of the robot azimuth coordinate system and an x axis to be theta _ pitch; adjusting the yaw direction observation angle of the vision system, and enabling the rotation angle between the projection of the forward direction of the vision system in the x-y plane of the robot azimuth coordinate system and the x axis to be theta _ yaw;
step S12, keeping the current position and posture of the humanoid robot unchanged, setting a group of distributed marking points in the visible range of the visual system, and measuring and recording the relative position and relative direction between each marking point and the humanoid robot;
step S13, placing the football in the positions of the marking points in the step S12 in sequence, enabling a vision system to acquire an image for each position, further processing the image to obtain the pixel coordinates of the center of the football sphere in the image, and recording the data of the current h, theta _ pitch and theta _ yaw, the position data of the marking points and the corresponding pixel coordinate data in the image;
step S14, returning to step S11, adjusting the values of h, θ _ pitch, and θ _ yaw, and repeatedly executing steps S12 to S13;
step S15, establishing a neural network including an input layer, a hidden layer and an output layer;
step S16, dividing the data recorded in the step S13 into a training set and a testing set, training the neural network in the step S15 according to the training set, and constructing a mapping relation between the football orientation and football image pixel coordinates, namely a football orientation mathematical model, under the condition that the values of h, theta _ pitch and theta _ yaw are known;
the football orientation real-time measuring subsystem is used for calculating in real time to obtain the relative position and the relative direction between the football and the humanoid robot according to the established football orientation mathematical model and the image information acquired by the vision system in the humanoid robot, and recording the relative position and the relative direction as the football orientation;
the target direction real-time acquisition subsystem is used for calculating on line to obtain the relative position and the relative direction between the expected football and the humanoid robot after the football is kicked by directly appointing a target direction, or automatically identifying the goal direction, or automatically planning the pass direction, and recording the relative position and the relative direction as the target direction;
the kickball action trajectory planning subsystem is used for planning kickball actions suitable for the humanoid robot according to the football direction and the target direction which are obtained currently, and recording the kickball actions as kickball action trajectories;
the kickball action execution subsystem is used for enabling the humanoid robot to execute kickball actions according to the currently obtained kickball action track;
the humanoid robot is provided with a visual system for acquiring a scene image, and the visual system has the capability of adjusting the observation angle; the pitch angle of the head and a visual system positioned on the head is indirectly adjusted by adjusting a kinematic joint of the waist or neck part, or the visual angle in the pitch direction is adjusted by adjusting the parameters of a camera arranged in the visual system;
the two legs of the humanoid robot respectively have at least 6 degrees of freedom of movement, and the feet at the tail ends of the legs are of hard structures with certain shapes;
the humanoid robot is provided with a robot position coordinate system, the origin of the robot position coordinate system is located at a projection point of the center position of the waist of the humanoid robot body on the ground, the x-axis of the robot position coordinate system points to the direction of the right front of the humanoid robot body in the positive direction, and the z-axis of the robot position coordinate system points to the opposite direction of gravity in the positive direction.
2. The system according to claim 1, wherein the robot comprises a ball kicking device,
the football is a spherical rollable object, and the number of the football in the field is 1; the relative position between the football and the humanoid robot is the relative position between the center of the football and the humanoid robot; the relative direction between the football and the humanoid robot is the relative direction between the center of the football and the humanoid robot;
the football field consists of a football field area, a goal and a field mark, wherein the field mark comprises an entity boundary, a color marking line, a color area, an information display board and an invisible electronic mark;
the kicking motion refers to a series of body motions that the humanoid robot needs to perform in order to strike the soccer ball from the soccer ball orientation to the target orientation.
3. The system of claim 1, wherein the vision system is configured to adjust the observation angle including the pitch direction and the yaw direction by adjusting a part of the joints of the body of the humanoid robot and/or by adjusting the settings of the vision system.
4. A method for generating a kicking motion of a humanoid robot using the kicking motion generation system of the humanoid robot according to any one of claims 1 to 3, comprising the steps of:
step S1, carrying out recognition training on a football orientation recognition training subsystem to obtain the football orientation mathematical model;
step S2, the vision system acquires the current scene image, inputs the image information into the football orientation real-time measuring subsystem, and acquires the football orientation;
step S3, the target azimuth real-time acquisition subsystem acquires the target azimuth;
step S4, the kickball action track planning subsystem obtains the kickball action track;
step S5, the kicking action execution subsystem executes the kicking action;
the step S1 is a training stage, and is executed once in advance in a non-kicking state; step S11, adjusting the gesture of the humanoid robot to enable the distance between the visual system and the ground to be h; adjusting a pitching observation angle of the vision system to enable a rotation angle between a projection of a forward direction of the vision system in an x-z plane of the robot azimuth coordinate system and an x axis to be theta _ pitch; adjusting the yaw direction observation angle of the vision system, and enabling the rotation angle between the projection of the forward direction of the vision system in the x-y plane of the robot azimuth coordinate system and the x axis to be theta _ yaw;
step S12, keeping the current position and posture of the humanoid robot unchanged, setting a group of distributed marking points in the visible range of the visual system, and measuring and recording the relative position and relative direction between each marking point and the humanoid robot;
step S13, placing the football in the positions of the marking points in the step S12 in sequence, enabling a vision system to acquire an image for each position, further processing the image to obtain the pixel coordinates of the center of the football sphere in the image, and recording the data of the current h, theta _ pitch and theta _ yaw, the position data of the marking points and the corresponding pixel coordinate data in the image;
step S14, returning to step S11, adjusting the values of h, θ _ pitch, and θ _ yaw, and repeatedly executing steps S12 to S13;
step S15, establishing a neural network including an input layer, a hidden layer and an output layer;
step S16, dividing the data recorded in the step S13 into a training set and a testing set, training the neural network in the step S15 according to the training set, and constructing a mapping relation between the football orientation and football image pixel coordinates, namely a football orientation mathematical model, under the condition that the values of h, theta _ pitch and theta _ yaw are known;
the steps S2-S5 are kicking steps, and are repeatedly executed in a kicking state.
5. The method as claimed in claim 4, wherein the neural network established in step S15 has a forward propagation model of the input layer and the hidden layer as yl＝f(ul)＝f(Wlyl-1+bl) Where f (-) is the ReLU activation function, WlAnd blRespectively the weight and bias matrix to be optimized in the l-th layer, ylIs the output of the l-th layer; the forward propagation model of its output layer is yl＝ul＝Wlyl-1+bl(ii) a Loss function using mean square error
Figure FDA0002803830730000051
The optimization of weight and bias adopts gradient descent method
Figure FDA0002803830730000052
6. The method according to claim 4, wherein the step S14 is not essential, and the values of h, θ _ pitch, and θ _ yaw are unique if the step S14 is skipped.
7. The method for generating a kicking motion of a humanoid robot as claimed in claim 4, wherein said football orientation real-time measuring subsystem in step S2 obtains the football orientation by:
processing the image information acquired by the vision system, detecting the position of the football in the image, and acquiring the pixel coordinates of the center position of the football in the image; and jointly inputting the obtained pixel coordinates and the values of the current h, theta _ pitch and theta _ yaw of the humanoid robot into the football orientation mathematical model, and solving and outputting the football orientation.
8. The method as claimed in claim 4, wherein the step S4 of obtaining the kicking trajectory by the kicking trajectory planning subsystem comprises:
step S41, obtaining or measuring the boundary contour shape of the humanoid robot foot in advance, obtaining a fitting function of the foot boundary contour by using a curve fitting method, marking the fitting function as a foot contour BoundryX (y), and obtaining the foot contour BoundryX (y) by differentiating to obtain the foot contour BoundryX (y)
Figure FDA0002803830730000053
Step S42, reasonably selecting kicking legs according to the football direction, the target direction, the current positions and the supporting states of the two legs of the humanoid robot;
step S43, selecting the relative position p in the orientation of the footballb＝[pbx,pby,pbz]TAnd a relative position p in the target positionT＝[pTx,pTy,pTz]TCalculating the direction of kicking the ball
Figure FDA0002803830730000054
With the kicking contact point on the footballThe normal direction along the spherical surface is the same as the kickball direction theta _ kick, and then the kickcontact point position p on the football is calculatedk；
Step S44, making the outline normal direction of the kicking point position of the humanoid robot foot and the kicking direction theta _ kick be the same as the principle
Figure FDA0002803830730000061
Calculating the kicking point position p of the foot of the humanoid robotr；
Step S45, obtaining a kicking point p of the foot of the humanoid robot by using a polynomial fitting methodrThe kicking action track of the football needs to pass through the kicking contact point p on the footballk。
CN201910912950.4A 2019-09-25 2019-09-25 System and method for generating kicking action of humanoid robot Active CN110653819B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910912950.4A CN110653819B (en) 2019-09-25 2019-09-25 System and method for generating kicking action of humanoid robot

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910912950.4A CN110653819B (en) 2019-09-25 2019-09-25 System and method for generating kicking action of humanoid robot

Publications (2)

Publication Number Publication Date
CN110653819A CN110653819A (en) 2020-01-07
CN110653819B true CN110653819B (en) 2021-02-09

Family

ID=69039121

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910912950.4A Active CN110653819B (en) 2019-09-25 2019-09-25 System and method for generating kicking action of humanoid robot

Country Status (1)

Country Link
CN (1) CN110653819B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111487972B (en) * 2020-04-24 2024-04-26 深圳市优必选科技股份有限公司 Kicking gait planning method and device, readable storage medium and robot
CN111599252B (en) * 2020-05-12 2021-09-17 桂林电子科技大学 Programmable teaching robot based on game mechanism
CN113091749B (en) * 2021-04-12 2022-08-23 上海大学 Walking navigation and repositioning method of humanoid robot in complex unknown maze environment
CN113377108B (en) * 2021-06-09 2024-05-28 乐聚(深圳)机器人技术有限公司 Control method, device and equipment of bipedal robot and storage medium
CN113580147B (en) * 2021-09-02 2023-04-14 乐聚(深圳)机器人技术有限公司 Robot control method, device, equipment and storage medium
CN114167749A (en) * 2021-11-17 2022-03-11 深兰盛视科技（苏州）有限公司 Control method of football robot and related device
CN114905527A (en) * 2022-05-31 2022-08-16 江苏经贸职业技术学院 Football robot interception method based on Markov chain and football robot

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103217903B (en) * 2013-04-07 2016-01-20 南京邮电大学 Apery Soccer robot omnidirectional based on two control balancing making mechanism plays football method
WO2016093607A1 (en) * 2014-12-12 2016-06-16 한국항공우주연구원 Apparatus and method for controlling moving object, and computer-readable recording medium in which program for implementing method in computer is recorded
CN106426171A (en) * 2016-11-01 2017-02-22 河池学院 Self-walking type intelligent soccer robot
CN106964145B (en) * 2017-03-28 2020-11-10 南京邮电大学 Humanoid football robot passing control method and team ball control method
CN108563220A (en) * 2018-01-29 2018-09-21 南京邮电大学 The motion planning of apery Soccer robot

Also Published As

Publication number Publication date
CN110653819A (en) 2020-01-07

Similar Documents

Publication Publication Date Title
CN110653819B (en) System and method for generating kicking action of humanoid robot
CN104504694B (en) A kind of method for obtaining moving sphere three-dimensional information
CN101964047B (en) Multiple trace point-based human body action recognition method
CN107646125A (en) The sensing device further and method for sensing of mobile spheroid
US11850498B2 (en) Kinematic analysis of user form
CN107545562B (en) Method, system and non-transitory computer readable recording medium for correcting brightness of ball image
EP4031256A1 (en) Method and system of replicating ball trajectories using an automated ball throwing machine
JP2023081889A (en) Systems and methods for measurement of 3d attributes using computer vision
CN112184807B (en) Golf ball floor type detection method, system and storage medium
KR102517067B1 (en) Ceiling golf simulation system using two cameras
KR102543653B1 (en) Method for Constructing Virtual Space Movement Platform Using Cross-covariance 3D Coordinate Estimation
WO2018207388A1 (en) Program, device and method relating to motion capture
Ziegler et al. A multi-modal table tennis robot system
TWI799184B (en) Device and method for calculating spin of golf ball moved by hitting
CN115414648B (en) Football evaluation method and football evaluation system based on motion capture technology
Lopez et al. Offside detection system using an infrared camera tracking system
CN117994425A (en) Hybrid intelligent method for acquiring three-dimensional information of human skeletal joints from two-dimensional image
Fan et al. Omnidirectional kick in RoboCup3D simulation
Shih et al. A vision based interactive billiard ball entertainment system
Chuang et al. Vision-based batting training system
Song et al. Research and analysis of table tennis movement trajectory prediction model based on deep learning
JP2017113112A (en) Golf swing analysis method
Miao et al. Review on Robotic Application in the Field of Sports
CN116012415A (en) Ping-pong ball rotation estimation method and rotation estimation system based on vision measurement
Chen et al. Ball Trajectory and Landing Point Prediction Model Based on EKF Algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant