Intelligent tennis partner training robot with body and working method thereof

Abstract

The invention relates to an intelligent tennis training robot with a body and a working method thereof. The robot comprises a wheel foot module, a sensing module, a ball receiving and sending module, a feedback module and an electric control module. The ball receiving and serving module can control the ball serving pitch angle and the ball serving speed of the robot in a ball serving mode; when the tennis ball collecting mode is switched, the tennis ball can be collected; the feedback module visually feeds back the ball hitting action of the trainee through the lamp strip and the display screen; the electric control module comprises a main control board, an analysis board and a power supply assembly and is responsible for motion control and sensing feedback data processing; the main control board coordinates the work of the ball receiving and serving module and the wheel foot module; the analysis plate receives and processes data from the sensing module, and performs man-machine interaction with a trainee through the feedback module; the labor cost of tennis partner training is saved; the problem that all-directional ball feeding is difficult to realize by traditional manual coaches is solved; the defect of intelligence of an existing action analysis method and a training robot is overcome, and the training efficiency is remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of robotics, and relates to an embodied intelligent tennis training robot and a working method thereof, and specifically to a tennis training robot integrating tennis serving and receiving functions, and a method for realizing intelligent human-computer interaction by means of embodied intelligent technology. Background Art

[0002] Tennis is a sport that is loved by many sports enthusiasts for its competitiveness, fun, sociability, and safety, and has a wide range of participants. However, tennis hitting techniques involve complex and delicate movements, and participants will inevitably make mistakes in learning and practice, which makes the entry threshold of tennis high, the learning cost high, and the training cycle long. In order to master the correct tennis movements, trainees need to form accurate movement memory through multi-ball exercises, and coaches use objective data analysis methods to guide and adjust the trainees' movements to ensure that they can return the ball at high speed and high accuracy. In order to provide a large number of stable incoming balls, smart ball machines are used to feed balls; in order to reduce the tedious task of picking up balls, manual or automatic ball picking tools are usually equipped. Therefore, modern tennis training is a complex interactive process involving people, machines, and balls.

[0003] However, current tennis training equipment has the problems of single function, low integration and weak human-computer interaction. Existing training equipment can only complete simple tasks such as fixed-point serving and automatic tennis ball recovery according to the goals set in advance by players or coaches. It lacks real-time perception of players' performance and it is difficult to generate scientific and effective guidance programs based on players' practice performance. In existing tennis training, the understanding and learning between humans and machines are one-way. Only by effectively enhancing the mutual understanding and cognition between players and auxiliary equipment, and giving training equipment the ability to guide players' actions and behaviors, can the best training program be formulated dynamically according to players' practice performance, shortening the tennis learning cycle and consolidating correct movements.

[0004] Today's technology is developing rapidly, especially the rapid progress of a series of cutting-edge technologies such as artificial intelligence, machine learning, and sensor technology. Compared with traditional artificial intelligence, embodied intelligence focuses more on obtaining information through real-time interaction with the environment, and thinking, making decisions, and acting based on this information. Combining embodied intelligence methods with tennis training robots, the training robots are expected to achieve greater breakthroughs in perception analysis and interactive decision-making, improve the scientificity and efficiency of tennis training, and benefit a large number of players. Summary of the invention

[0005] Based on the design of a tennis training robot with basic interactive functions, the present invention proposes a method for using embodied intelligence to drive the tennis training robot to work efficiently. The present invention adopts the following technical solutions, which are described in detail in conjunction with the accompanying drawings.

[0006] The present invention proposes a tennis training robot, which not only has the basic functions of launching tennis balls in all directions in space and automatically picking up scattered tennis balls in the whole field, but also further integrates the ability to perceive and analyze the trainer's batting posture and tennis ball trajectory. Through visualization, the robot can give the trainer real-time feedback on the problems in his movements, thereby helping the trainer to understand and correct wrong movements more scientifically and efficiently, and ultimately providing the trainer with an advanced training auxiliary tool.

[0007] The embodied intelligent tennis training robot comprises a wheel and foot module 1, a perception module 3; and is characterized in that it also comprises a ball receiving and serving module 2, a feedback module 4 and an electric control module 5;

[0008] The serving and receiving module is located above the wheel-foot module. As a core component, it is responsible for performing the serving and receiving tasks and can switch freely between the two modes. In the serving mode, it can control the pitch angle and ball speed of the robot. When switching to the receiving mode, it can recycle the tennis ball.

[0009] The perception module is located on the top of the robot and is equipped with a binocular camera and an event camera to capture the trainer's movements, the robot's position on the court, and the landing point of the tennis ball.

[0010] The feedback module is arranged on the front of the robot, and provides visual feedback on the trainee's batting action through a light strip and a display screen.

[0011] The electronic control module is placed inside the robot, including components such as the main control board, analysis board, and power supply, providing hardware support for the robot's motion drive, visual data analysis, and control logic. It is responsible for motion control and sensory feedback data processing; the main control board coordinates the work of the ball receiving and sending module and the wheel and foot module; the analysis board receives and processes data from the sensory module, generates scientific decisions, and interacts with the trainer through the feedback module.

[0012] The wheel-foot module mainly includes a pair of motor wheels, a pair of calves, front legs, front calves, thighs and four joint motors. When the motor wheels rotate, the robot can move at different speeds and directions; the four joint motors are responsible for driving the flexion and extension of the thighs and calves, thereby realizing the height adjustment function of the receiving and serving module.

[0013] Furthermore, the ball-receiving and serving module is mainly composed of a friction wheel mechanism, a ball-shooting mechanism, a ball-storing mechanism, and a ball-receiving and serving switching mechanism, which are specifically described as follows:

[0014] The friction wheel mechanism drives the friction wheel to rotate at different speeds and directions through a pair of AC brushless motors, realizing the outward launch (serve mode) and inward recovery (collection mode) of the tennis ball. The pitch adjustment motor controls the angle of the friction wheel fixing bracket to achieve the shooting angle adjustment in the serving state and the control of the friction wheel angle when switching between the serving and collecting states;

[0015] The shooting mechanism includes a three-way shooting cylinder. Under the action of the electric push rod, the tennis ball in the shooting cylinder can be pushed to contact the friction wheel and shoot out. When the push rod is retracted, the next tennis ball to be launched automatically falls under the action of gravity and enters the shooting preparation state;

[0016] The ball storage mechanism is located on the top of the robot, and the tennis balls fall into the ball storage bin after being collected.

[0017] The serve-and-receive switching mechanism gives the friction wheel the ability to move back and forth and swing, making the overall mechanism design compact while ensuring the robot's serve-and-receive function. In the serve mode, the friction wheel mechanism is stored in the robot's abdomen; in the receive mode, the friction wheel mechanism is driven by the push rod and the servo, extending from the robot's abdomen and swinging to a height close to the ground to pick up the tennis ball on the ground.

[0018] Furthermore, under the action of the stirring arm inside the ball storage bin, the bottom layer of tennis balls keeps moving slowly, thereby effectively preventing tennis balls from getting blocked;

[0019] Furthermore, the right swing arm servo 2-4-1 of the transmitting and receiving switching mechanism 2-4 is connected to the right swing arm 2-1-5 in the friction wheel mechanism 2-1 by bolts, and the left swing arm servo 2-4-2 is connected to the left swing arm 2-1-6 in the friction wheel mechanism 2-1 by bolts; under the drive of the servo, the friction wheel mechanism 2-1 can rotate around the servo shaft; the right swing arm servo 2-4-1 is fixed to the right servo bracket 2-4-5 by bolts, and the right servo bracket 2-4-5 is fixed to the right guide rail 2-4-3 by bolts. The left servo motor 2-4-2 is fixed to the left servo motor bracket 2-4-6 by bolts, and the left servo motor bracket 2-4-6 is fixed to the servo motor slider of the left guide rail 2-4-4 by bolts; both ends of the connecting block 2-4-7 are connected to the right servo motor bracket 2-4-5 and the left servo motor bracket 2-4-6 by bolts at the same time, and the middle part of the connecting block 2-4-7 is fixed to the output end of the electric push rod 2-4-8 by bolts; the rear end of the electric push rod 2-4-8 is fixed to the chassis 1-11 by bolts.

[0020] Furthermore, the wheel-foot module is located at the bottom of the robot and adopts a double-wheel-foot mechanism to realize the robot movement and height adjustment functions;

[0021] The perception module 3 is composed of a binocular camera facing forward and an event camera facing backward, which is used to capture the trainer's movements, the robot's position on the court, and the landing point of the tennis ball;

[0022] The binocular camera in the perception module uses the existing human skeleton key point recognition algorithm to accurately identify the trainer's position and posture. Combined with the target detection algorithm, the binocular camera can calculate the distance between the camera and the court boundary, thereby determining the robot's position coordinates in the court coordinate system. When the binocular camera detects that the trainer is swinging the racket to hit the ball, it triggers the event camera, which can collect the changing pixel points in the scene at high speed and accurately identify the spatial position coordinates of the tennis ball by analyzing the parallax of the two event camera data.

[0023] The feedback module 4 includes a display screen 4-1 located at the top of the front of the robot and light strips distributed throughout the body of the robot, which are leg light strips 4-2, chest light strips 4-3, upper arm light strips 4-4 and forearm light strips 4-5; the arm servo 4-6 is used to control the extension and tightening of the arm 4-7.

[0024] The feedback module is located on the front of the robot, and includes an arm, a light strip, and a display screen. The light strip is installed on the surface of the feet, body, and arms to provide visual feedback on the coordination of the batting action; the display screen is used to display the geometric posture of the batting action.

[0025] The present invention proposes that the working modes of the embodied intelligent tennis training robot include a serving training mode (referred to as the serving mode) and an autonomous tennis ball recovery mode (referred to as the collecting mode).

[0026] 1. Serving mode

[0027] 1) The robot calculates the temporal and spatial position of the human-machine-ball. The court coordinate system is established; the binocular camera installed in front of the robot captures the position of the court sideline in real time, and calculates the distance between the camera and the two end points of the bottom line to determine the real-time position of the robot in the court; the binocular camera captures the position coordinates of the trainer relative to the camera and converts them to the court coordinate system; the binocular event camera installed behind the robot captures the landing position and speed of the tennis ball hit by the trainer.

[0028] 2) The robot serves the ball according to the set rules. The joint motor in the wheel and foot module of the robot is responsible for adjusting the ball position, and the motor wheel controls the forward and backward movement and steering angle of the robot; in the ball receiving and serving module, the ball storage mechanism releases the tennis ball into the ball shooting barrel, and the pitch adjustment mechanism controls the pitch angle of the friction wheel. The speed of the two friction wheels determines the speed and rotation of the ball. The electric push rod in the ball shooting mechanism pushes the tennis ball to contact the friction wheel, thereby launching the tennis ball according to the predetermined parameters, realizing the basic ball feeding function.

[0029] 3) The robot analyzes the trainer's posture standardization and provides visual feedback to the trainer. The binocular camera captures the spatial position information of the human skeleton in real time, defines the key posture characteristics of the human body's hitting action, and calculates the deviation value between the trainer's posture and the standard characteristics in real time. The robot feeds back the deviation to the trainer in the form of numbers and visual icons on the display screen to remind the trainer to correct the action.

[0030] 4) The robot analyzes the trainee's coordination of movements and gives visual feedback to the trainee. The binocular camera captures the spatial position and angular velocity information of the hips, chest, shoulders and elbows in the human skeleton in real time; calculates the speed of each part's contribution to the racket, compares it with the database of professional athletes, and evaluates the degree and timing of each part's force; through the color changes and flow speed of the light strips on the corresponding parts of the robot, it intuitively displays the force situation and timing deviation of each part, reminding the trainee to correct the movement.

[0031] 2. Ball collection mode

[0032] 1) When there are no more tennis balls in the ball storage bin, the robot automatically enters the ball collection mode. The robot moves around the field autonomously, identifies and records the position coordinates of each scattered tennis ball through the binocular camera, and records the serial number of each ball in order.

[0033] 2) After the sending and receiving switching mechanism is activated, the wheel-foot module controls the robot to squat to the lowest height. The electric push rod and the servo work together to move the friction wheel forward and downward, accurately controlling the speed and angle of the ball collection. The robot moves to the back of each tennis ball in sequence according to the serial number, adjusts its posture so that the center of the friction wheel mechanism is aligned with the center of the tennis ball, and projects the tennis ball into the ball storage bin through the friction wheel mechanism.

[0034] 3) After all the balls with serial numbers are collected, the robot automatically returns to the vicinity of the trainer and waits for instructions to enter the next round of training.

[0035] A working method of an embodied intelligent tennis training robot, characterized in that:

[0036] The electronic control module is responsible for motion control and sensory feedback data processing; the main control board coordinates the work of the ball receiving and sending module and the wheel and foot module; the analysis board receives and processes data from the sensory module, generates scientific decisions, and interacts with the trainer through the feedback module;

[0037] The wheel-foot module realizes the height adjustment, movement and direction control of the robot as a whole; the four leg motors coordinately control the bending degree of the upper and lower legs to adjust the ball position and ensure the squatting posture when receiving the ball; the speed difference between the left and right motor wheels determines the steering angle of the robot; when the left and right motor wheels have the same speed, the robot moves in a straight line; when there is a speed difference, the robot turns;

[0038] The serving and receiving module has two working states: serving mode and receiving mode. In serving mode, the tennis ball in the ball storage mechanism is released into the ball shooting tube, and the electric push rod pushes the tennis ball to make it contact with the friction wheel. Driven by the AC brushless motor, it generates a preset speed and rotation, the servo controls the pitch angle, and the wheel-foot module controls the serving height and direction, and finally the tennis ball is thrown to the trainee according to the preset parameters. In receiving mode, the receiving and serving switching mechanism is started, and the electric push rod and the servo work together to move the friction wheel mechanism to the front and lower part of the robot, control the receiving speed, and the pitch adjustment mechanism adjusts the receiving angle. The wheel-foot module controls the robot to squat and adjust the direction, and finally the tennis balls scattered on the court are recovered into the ball storage mechanism.

[0039] The perception module consists of a binocular camera and an event camera. It uses computer vision technology to solve the robot's position coordinates on the court, the spatiotemporal position of the trainer's human skeleton, the landing point information of the hit tennis balls, and the position information of the scattered tennis balls, providing effective data support for the analysis board of the electronic control part; the analysis board has built-in tennis motion analysis and visual motion feedback prompt algorithms, which drive the light strips and display screens in the feedback module according to the analysis results to achieve real-time guidance for the trainers; at this point, the tennis training robot has realized embodied intelligence from perception to action.

[0040] Furthermore, when the serving and receiving module switches from the serving mode to the receiving mode, the entire friction wheel mechanism 2-1 is moved and extended out of the robot body under the drive of the electric push rod 2-4-8; the pitch adjustment motor 2-1-4 is used to adjust the appropriate angle; the right swing arm 2-1-5 and the left swing arm 2-1-6 rotate counterclockwise under the drive of the right swing arm servo 2-4-1 and the left swing arm servo 2-4-2, so that the right friction wheel 2-1-1 and the left friction wheel 2-1-2 in the friction wheel mechanism 2-1 can approach and contact the tennis ball 6 on the ground, thereby throwing the tennis ball 6 upward and causing it to fall into the ball basin 2-3-1 in the ball storage mechanism 2-3.

[0041] Furthermore, when the tennis training robot is in the ball receiving mode, the wheel and foot module controls the robot to squat, while the serving and receiving module switches to the ball receiving mode and extends the friction wheel mechanism; the binocular camera in the perception module can accurately identify the position of the tennis ball in the venue and send its coordinates to the electronic control module; then, the electronic control module directs the robot to move towards the tennis ball and aligns the center of the friction wheel mechanism with the tennis ball; under the action of friction, the tennis ball is thrown upward and falls smoothly into the ball basin of the ball storage mechanism, thereby completing the ball receiving process.

[0042] Furthermore, the tennis training robot can sense and feedback the kinematic performance of the tennis trainer's movements; after processing the data from the binocular camera in the analysis board using the existing human skeleton recognition algorithm, the position and angular velocity data of the trainer's hips, chest, shoulders, elbows, and hands are obtained; let i represent a body part, i=1 represents the hips, i=2 represents the chest, i=3 represents the shoulders, i=4 represents the elbows, and i=5 represents the hands; ω i (t) represents the angular velocity vector of part i at time t, p i (t) represents the spatial position of part i at time t, v5(t) represents the spatial velocity of the hand at time t, and the contribution velocity of each part to the hand is calculated and accumulated to the observed part i according to the following formula;

[0043]

[0044] Where n is the cumulative variable, n = (1, ..., i), and the final cumulative contribution speed v Ai (t) is a scalar;

[0045] The kinematic chain analysis of tennis hitting movements mainly focuses on the degree of force and the timing of force; the degree of force CV i The calculation is done using the following formula:

[0046]

[0047] Assume that the time when the cumulative contribution speed of part i reaches its peak is t i , which is calculated as

[0048]

[0049] The timing of force application is calculated as follows: the chest lags behind the hips by Δt1=t2-t1, the shoulder lags behind the chest by Δt2=t3-t2, and the elbow lags behind the shoulder by Δt3=t4-t3.

[0050] Furthermore, the tennis training robot visualizes the trainer's motion chain feedback as follows: the leg light belt 4-2 is used to feedback the hip information, the chest light belt 4-3 is used to feedback the chest information, the upper arm light belt 4-4 is used to feedback the shoulder information, and the forearm light belt 4-5 is used to feedback the elbow information; the color of the light belt is used to indicate the degree of force, which is green for good RGB=(0,255,0), orange for better RGB=(255,255,0), yellow for bad RGB=(255,50,0), and red for severe bad RGB=(255,0,0), and the calculation method is as follows:

[0051]

[0052] In the formula, CV iRefIndicates the reference value of the force level, R iC , R iC , R iC It is a manually set fault tolerance value; if CV1 is greater than the reference value CV 1Ref If CV4 is higher than the reference value CV 4Ref Low, RGB4 is set to (0,255,0);

[0053] The light strips of each part are used to flow to provide feedback on the coordination of the timing of each part’s force application Δt i ; If there is an action delay, the first light in the light strip at that part will light up i Seconds later, the second light strip lights up, 2f i After seconds, the third light strip lights up; then the light strip flow time interval f of the i-th part of the entire light strip system i Calculated as

[0054]

[0055] In the formula, Δt i represents the lag time of site i, Δt iRef Indicates the standard reference value, T iA , T iB , T iC Indicates the preset tolerance value; when f i =0 means the LED in this area is always on; if Δt i Below the reference value Δt iRef , the direction of the light strip is from the far end to the near end of the body, prompting the trainee to slow down the timing of this part; on the contrary, if Δt i Higher than the reference value Δt iRef , the light strip flows from the proximal end to the distal end of the body, prompting the trainee to speed up the timing of this part.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The tennis training robot of the present invention combines the tasks that were originally completed by the serving robot and the ball picking robot into one, and realizes the smooth switching of the sending and receiving functions through a sophisticatedly designed mechanism. The robot can accurately control the trajectory of tennis ball launch and automatically find and recover scattered tennis balls, thereby ensuring the full automation of the tennis training task, saving the manpower cost of tennis training, and improving the stability and accuracy of the training.

[0058] 2. The tennis training robot of the present invention has a flexible mechanical structure and can accurately simulate balls coming from different directions, speeds, ball heights and pitch angles. With its rapid lifting function, the robot can quickly lower its height after serving the ball to the trainee to prevent being hit and damaged, effectively solving the problem of all-round ball feeding that is difficult to achieve with traditional manual coaches.

[0059] 3. The working principle of the embodied intelligent tennis training robot described in the present invention gives the robot the ability to perceive, analyze and provide feedback, enabling it to conduct human-machine interaction scientifically and efficiently during training, meeting the needs of tennis teaching and training in the new era. This working mode overcomes the limitations of traditional teaching and coach guidance, makes up for the shortcomings of existing motion analysis methods and intelligent training robots, and significantly improves training efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to better describe and explain the technical solution of the invention and clearly show the content and structure of the invention, the present invention is further described below in conjunction with the accompanying drawings:

[0061] Figure 1 The isometric view of the tennis training robot in the serving mode of the present invention;

[0062] Figure 2 This is an axonometric view of the tennis training robot in the ball receiving mode of the present invention;

[0063] Figure 3 This is an axonometric diagram of the wheel and foot module of the tennis training robot according to the present invention;

[0064] Figure 4 It is an axonometric diagram of the serving and receiving module of the tennis training robot in the serving mode of the present invention;

[0065] Figure 5 A top view of the serving and receiving module of the tennis training robot according to the present invention in the serving mode;

[0066] Figure 6 It is an axonometric diagram of the serving and receiving module of the tennis training robot in the present invention in the receiving mode;

[0067] Figure 7 This is an axonometric diagram of the launching mechanism in the serving and receiving module of the tennis training robot of the present invention;

[0068] Figure 8 This is an axonometric diagram of the ball storage mechanism in the serving and receiving module of the tennis training robot of the present invention;

[0069] Fig. 9 This is an axonometric diagram of the electric control module of the tennis training robot according to the present invention;

[0070] Fig.10This is a schematic diagram of the relationship between the modules of the tennis training robot of the present invention;

[0071] Fig.11 This is the principle diagram of the geometric posture analysis of the present invention;

[0072] Description of labels:

[0073] Wheel and foot module 1; ball receiving and serving module 2; perception module 3; feedback module 4; electronic control module 5; tennis ball 6;

[0074] Motor wheel 1-1; calf 1-2; front leg 1-3; front calf 1-4; thigh 1-5; front leg motor 1-6; thigh motor 1-7; housing 1-8; support 1-9; bearing 1-10; bottom plate 1-11;

[0075] Friction wheel mechanism 2-1; ball shooting mechanism 2-2; ball storage mechanism 2-3; ball receiving and sending switching mechanism 2-4;

[0076] Right friction wheel 2-1-1; left friction wheel 2-1-2; friction wheel bracket 2-1-3; pitch adjustment motor 2-1-4; right swing arm 2-1-5; left swing arm 2-1-6;

[0077] Ball shooting tube 2-2-1; electric push rod for pushing ball 2-2-2; ball pushing block 2-2-3; push rod fixing block 2-2-4; push rod fixing frame 2-2-5;

[0078] Ball basin 2-3-1; stirring arm 2-3-2; stirring servo 2-3-3;

[0079] Right swing arm servo 2-4-1; left swing arm servo 2-4-2; right guide rail 2-4-3; left guide rail 2-4-4; right servo bracket 2-4-5; left servo bracket 2-4-6; connecting block 2-4-7; electric push rod 2-4-8;

[0080] Binocular camera 3-1; event camera 3-2;

[0081] Display screen 4-1; leg light strip 4-2; chest light strip 4-3; upper arm light strip 4-4; lower arm light strip 4-5; arm servo 4-6; arm 4-7;

[0082] Battery 5-1; control board 5-2; analysis board 5-3. DETAILED DESCRIPTION

[0083] In order to make the objectives, technical solutions and advantages of the embodiments of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings.

[0084] In order to make the drawings concise and easy to understand, the figure numbers in the patent of this invention adopt the secondary numbering rule. Not all parts of the robot are marked, but only the relevant parts of the important contents such as the structure and working principle described in this article are marked in detail.

[0085] In the invention described herein, unless otherwise specified or limited, the terms "installed", "set", "connected" and the like should be understood in a broad sense. For example, they may include fixed connection, detachable connection, integral connection; they may be mechanical connection, electrical connection; they may be direct connection between two elements, or indirect connection through an intermediate medium, or even internal connection of an element. For those skilled in the art, the specific meanings of the above terms in the invention may be understood according to the specific context.

[0086] The present invention provides an embodiment of an embodied intelligent tennis training robot, such as Figure 1 As shown, the axonometric view of the serving mode of the tennis training robot of the present invention shows its structural characteristics. The tennis training robot adopts a modular design, and the wheel foot module 1 is located at the bottom of the robot, which is responsible for support, movement, steering and height adjustment. The serving and receiving module 2 is located in the middle of the robot, which can control the pitch angle and ball speed of the robot in the serving mode; when switched to the receiving mode, the tennis ball on the ground can be recovered into the ball basin 2-3-1 on the top of the robot. In addition, the robot is equipped with a perception module 3, which is composed of a binocular camera facing forward and an event camera facing backward, aiming to achieve the time and space capture of the positions of people, robots and tennis players in the training scene. The feedback module 4 includes a display screen 4-1 located at the top of the front of the robot and light strips distributed throughout the body of the machine, which are leg light strips 4-2, chest light strips 4-3, upper arm light strips 4-4 and lower arm light strips 4-5. At the same time, the arm servo 4-6 is used to control the extension and tightening of the arm 4-7. The electric control module 5 is arranged inside the robot to ensure stable and efficient operation of the system.

[0087] like Figure 2 As shown, in the ball receiving mode, the wheel-foot module 1 controls the robot to squat, and at the same time, the ball receiving and serving module 2 switches to the ball receiving mode and extends the friction wheel mechanism 2-1. The binocular camera 3-1 in the perception module 3 can accurately identify the position of the tennis ball 6 in the field and send its coordinates to the electronic control module 5. Subsequently, the electronic control module directs the robot to move toward the tennis ball and aligns the center of the friction wheel mechanism 2-1 with the tennis ball 6. Under the action of friction, the tennis ball is thrown upward and smoothly falls into the ball basin 2-3-1 of the ball storage mechanism 2-3, thereby completing the ball receiving process.

[0088] like Figure 3As shown, the axonometric diagram of the wheel-foot module of the tennis training robot of the present invention shows its structure. The wheel-foot module consists of a pair of left and right wheel-feet. The bottom is equipped with a motor wheel 1-1, which can rotate after power is turned on. The wheels on the left and right sides can achieve all-round movement of the robot by adjusting the speed difference. The motor wheel 1-1 is connected to the shank 1-2 by bolts, and the middle part of the shank 1-2 is connected to the bottom of the front leg 1-3 by bearings and pins to ensure relative rotation; the bottom of the front shank 1-4 is also connected to the front leg 1-3 by bearings and pins, and the top of the front shank 1-4 is connected to the output shaft of the front leg motor 1-6 by bolts. In addition, the upper part of the shank 1-2 is also connected to the lower part of the thigh 1-5 by bearings and pins, and the upper part of the thigh 1-5 is fixed to the output shaft of the thigh motor 1-7 by bolts. The fixed ends of the front leg motor 1-6 and the thigh motor 1-7 are fixed to the housing 1-8 by bolts, and the housing 1-8 is connected to the support 1-9 by bolts, and the support 1-9 is installed on the bottom plate 1-11. The motors of the front legs and thighs can use the mature wheel-foot mechanism control rules to achieve height adjustment, thereby changing the overall height of the robot.

[0089] like Figures 4 to 8 As shown, the serving and receiving module 2 of the tennis training robot of the present invention mainly includes a friction wheel mechanism 2-1, a ball shooting mechanism 2-2, a ball storage mechanism 2-3 and a serving and receiving switching mechanism 2-4.

[0090] like Figure 4 This is an axonometric diagram of the tennis training robot's serving and receiving module in the serving mode, wherein the friction wheel mechanism 2-1 is at the front, the shooting mechanism 2-2 is located in the middle, and the front of the shooting mechanism 2-2 is close to the right friction wheel 2-1-1 and the left friction wheel 2-1-2 in the friction wheel mechanism. The receiving and sending switching mechanism 2-4 is located at the rear and bottom, and is responsible for pushing the friction wheel mechanism 2-1 outward.

[0091] Figure 5 The working principle of the friction wheel mechanism 2-1 is shown. In the friction wheel mechanism 2-1, the right friction wheel 2-1-1 and the left friction wheel 2-1-2 are both wheel motors that can achieve high-speed rotation after being powered on. Their motor ends are fixed to the friction wheel bracket 2-1-3 by bolts. When the tennis ball 6 touches the friction wheel, it will be ejected quickly. In order to adjust the ejection angle of the tennis ball 6, one end of the friction wheel bracket 2-1-3 is connected to the output shaft of the pitch adjustment motor 2-1-4 by bolts, and the other end of the bracket is connected to the right swing arm 2-1-5 through a bearing. At the same time, the pitch adjustment motor 2-1-4 is also fixed to the left swing arm 2-1-6 by bolts.

[0092] Figure 5The working principle of the transmitting and receiving switching mechanism 2-4 is described in detail. The right swing arm servo 2-4-1 in the mechanism is connected to the right swing arm 2-1-5 in the friction wheel mechanism 2-1 by bolts, and the left swing arm servo 2-4-2 is connected to the left swing arm 2-1-6 in the friction wheel mechanism 2-1 by bolts. Driven by the servo, the friction wheel mechanism 2-1 can rotate around the servo shaft. The right swing arm servo 2-4-1 is fixed to the right servo bracket 2-4-5 by bolts, and the right servo bracket 2-4-5 is fixed to the slider of the right guide rail 2-4-3 by bolts to achieve left and right movement; similarly, the left swing arm servo 2-4-2 is fixed to the left servo bracket 2-4-6 by bolts, and the left servo bracket 2-4-6 is fixed to the slider of the left guide rail 2-4-4 by bolts. The two ends of the connection block 2-4-7 are connected to the right servo support 2-4-5 and the left servo support 2-4-6 by bolts, and the middle of the connection block 2-4-7 is fixed to the output end of the electric push rod 2-4-8 by bolts. The rear end of the electric push rod 2-4-8 is fixed to the chassis 1-11 by bolts.

[0093] like Figure 6 As shown, the isometric diagram of the tennis training robot serving and receiving module of the present invention in the serving mode. Figure 4 ) Switch to ball collection mode (see Figure 6 ), the entire friction wheel mechanism 2-1 moves and extends out of the robot body under the drive of the electric push rod 2-4-8; the pitch adjustment motor 2-1-4 is used to adjust the appropriate angle; the right swing arm 2-1-5 and the left swing arm 2-1-6 rotate counterclockwise under the drive of the right swing arm servo 2-4-1 and the left swing arm servo 2-4-2, so that the right friction wheel 2-1-1 and the left friction wheel 2-1-2 in the friction wheel mechanism 2-1 can approach and contact the tennis ball 6 on the ground, thereby throwing the tennis ball 6 upward and making it fall into the ball basin 2-3-1 in the ball storage mechanism 2-3.

[0094] Figure 7 This is an axonometric diagram of the launching mechanism 2-2 in the ball serving and receiving module of the tennis training robot described in the present invention. The top of the ball shooting barrel 2-2-1 in the ball shooting mechanism 2-3 and the bottom of the ball basin 2-3-1 in the ball storage mechanism 2-3 are connected by corresponding bolt holes. The front end of the ball pushing electric push rod 2-2-2 is connected to the ball pushing block 2-2-3 by bolts. After the tennis ball 6 in the storage mechanism 2-3 falls into the ball shooting barrel 2-2-1, it moves forward under the push of the ball pushing electric push rod 2-2-2, and is ejected after contacting the friction wheel. The outside of the ball pushing electric push rod 2-2-2 is interference fit with the hole in the push rod fixing block 2-2-4, and the push rod fixing block 2-2-4 is connected to the ball shooting barrel 2-2-1 by bolts.

[0095] Figure 8This is an axonometric diagram of the ball storage mechanism in the ball receiving and sending module of the tennis training robot of the present invention. The ball basin 2-3-1 is installed on the top of the entire robot, in which a large number of tennis balls 6 can be stored. A hole for only one tennis ball to fall is provided at the bottom of the ball basin 2-3-1. In order to prevent a large number of tennis balls from accumulating and blocking the ball drop hole, a stirring arm 2-3-2 is installed on the top of the hole. Under the action of the stirring servo 2-3-3, the stirring arm 2-3-2 drives the tennis balls in the ball basin 2-3-1 to move at a low speed, so that they can freely fall when there is a vacant position in the ball shooting cylinder 2-2-1 when the previous tennis ball is launched from the ball shooting cylinder 2-2-1, so as to fill the vacant position.

[0096] Fig. 9 The isometric view of the electric control module of the tennis training robot of the present invention is shown. The electric control module 5 is installed on the upper surface of the bottom plate 1-11, located at the rear of the robot. Two batteries 5-1 are placed above the bottom plate 1-11, wherein the top of the right battery is equipped with a control board 5-2, and the top of the left battery is equipped with an analysis board 5-3.

[0097] The present invention provides another embodiment, a working method of an embodied intelligent tennis training robot, such as Fig.10As shown, the hardware composition and functional relationship diagram of the tennis training robot described in the present invention clearly shows the collaborative work of each module. The electronic control module located in the center of the figure is the control core of the entire system, responsible for motion control and perception feedback data processing. Among them, the main control board coordinates the work of the serving and receiving module and the wheel and foot module; the analysis board receives and processes data from the perception module, generates scientific decisions, and conducts human-computer interaction with the trainer through the feedback module. The wheel and foot module realizes the overall height adjustment, movement and direction control of the robot. The four leg motors coordinately control the bending degree of the upper and lower legs to adjust the ball position and ensure the squatting posture when receiving the ball; the speed difference between the left and right motor wheels determines the steering angle of the robot. When the speeds of the left and right motor wheels are the same, the robot moves in a straight line; when there is a speed difference, the robot turns. The serving and receiving module has two working states: serving mode and receiving mode. In the serving mode, the tennis ball in the ball storage mechanism is released into the ball shooting tube, and the electric push rod pushes the tennis ball to make it contact with the friction wheel. Driven by the AC brushless motor, the preset speed and rotation are generated. The servo controls the pitch angle, and the wheel-foot module controls the serving height and direction, and finally the tennis ball is thrown to the trainee according to the preset parameters. In the receiving mode, the sending and receiving switching mechanism is activated, and the electric push rod and the servo work together to move the friction wheel mechanism to the front and bottom of the robot to control the receiving speed. The pitch adjustment mechanism adjusts the receiving angle, and the wheel-foot module controls the robot to squat and adjust the direction, and finally the tennis balls scattered on the court are recovered into the ball storage mechanism. The perception module consists of a binocular camera and an event camera. It uses computer vision technology to solve the position coordinates of the robot on the court, the spatiotemporal position of the trainee's human skeleton, and the landing point information of the hit tennis ball and the position information of the scattered tennis balls, etc., to provide effective data support for the analysis board of the electronic control part. The analysis board has built-in tennis action analysis and visual action feedback prompt algorithms. According to the analysis results, the light strip and display screen in the feedback module are driven to work, realizing real-time guidance for the trainee. At this point, the tennis training robot has achieved embodied intelligence from perception to action.

[0098] like Fig.11 The figure shows the robot's perception and feedback method of the tennis trainer's posture. In the analysis board 5-3, the data from the binocular camera 3-1 is processed using the existing human skeleton recognition algorithm to obtain the trainer's posture. The entire tennis hitting action is divided into four stages: preparation, racket lead, racket swing, and follow-through. In the preparation stage, the hip height ratio δ is used. h (ratio of hip height in preparation to standing position), and the ratio of knee distance δ L (the ratio of the distance between the two knees when preparing to stand) two characteristics describe the rationality of the action. In the racket-drawing phase, the angle α1 of the chest relative to the direction of the ball and the distance d1 of the hand relative to the chest in the direction of the ball are used to describe whether the racket-drawing action is sufficient. In the racket-swinging phase, the angle α between the chest and the upper arm at the moment of hitting the ball is us And the angle α between the upper arm and the lower armua Describe whether the posture at the moment of hitting the ball is scientific and reasonable. In the follow-up phase, it is only necessary to release the residual speed naturally, so no action posture analysis is performed. The display screen 4-1 in the feedback module 4 of the robot will display the characteristic values ​​of each swing action of the trainee, and combine Fig.11 The visual images shown provide intuitive feedback to the trainer.

[0099] The robot can sense and feedback the kinematic performance of the tennis trainer's movements. In the analysis board 5-3, the data from the binocular camera 3-1 is processed using the existing human skeleton recognition algorithm to obtain the position and angular velocity data of the trainer's hips, chest, shoulders (racquet hand), elbows (racquet hand), and hands (racquet hand). Let i represent body parts, i=1 represents hips, i=2 represents chest, i=3 represents shoulders, i=4 represents elbows, and i=5 represents hands. ω i (t) represents the angular velocity vector of part i at time t, p i (t) represents the spatial position of part i at time t, v5(t) represents the spatial velocity of the hand at time t, and the contribution velocity of each part to the hand is calculated according to the following formula and accumulated to the observed part i

[0100]

[0101] Where n is the cumulative variable, n = (1, ..., i), and the final cumulative contribution speed v Ai (t) is a scalar.

[0102] The kinematic chain analysis of tennis hitting movements mainly focuses on the degree of force and the timing of force. i The calculation is done using the following formula:

[0103]

[0104] Assume that the time when the cumulative contribution speed of part i reaches its peak is t i , which is calculated as

[0105]

[0106] The timing of force application is calculated as follows: the chest lags behind the hips by Δt1=t2-t1, the shoulder lags behind the chest by Δt2=t3-t2, and the elbow lags behind the shoulder by Δt3=t4-t3.

[0107] like Figure 1As shown, it is a visual feedback of the trainer's kinematic chain. The leg light belt 4-2 is used to feedback the hip information, the chest light belt 4-3 is used to feedback the chest information, the upper arm light belt 4-4 is used to feedback the shoulder information, and the forearm light belt 4-5 is used to feedback the elbow information. The color of the light belt is used to indicate the degree of force, which is green for good RGB=(0,255,0), orange for better RGB=(255,255,0), yellow for bad RGB=(255,50,0), and red for severe bad RGB=(255,0,0). The calculation method is as follows

[0108]

[0109] In the formula, CV iRef Indicates the reference value of the force level, R iA , R iB , R iC It is a human-set error tolerance value. According to tennis experience, a higher hip CV is conducive to improving movement stability and speed, while a lower elbow CV is conducive to improving hitting stability and avoiding injuries. Therefore, if CV1 is higher than the reference value CV 1Ref If CV4 is higher than the reference value CV 4Ref If low, RGB4 is set to (0,255,0).

[0110] The light strips of each part are used to flow to provide feedback on the coordination of the timing of each part’s force application Δt i If there is an action delay, the first light in the light strip at that location will light up. i Seconds later, the second light strip lights up, 2f i After 1 second, the third light strip lights up. Then the light strip flow time interval f of the i-th part of the entire light strip system is i It can be calculated as

[0111]

[0112] In the formula, Δt i represents the lag time of site i, Δt iRef Indicates the standard reference value, T iA , T iB , T iC Indicates the preset tolerance value. i =0 means the LED in this area is always on. i Below the reference value Δt iRef , then the direction of the light strip is from the far end of the body to the near end (hip to foot, chest to hip, elbow to shoulder, hand to elbow), prompting the trainee to slow down the timing of this part. i Higher than the reference value Δt iRef, the light strip flows from the proximal end to the distal end of the body (feet to hips, hips to chest, shoulders to elbows, elbows to hands), prompting the trainee to speed up the timing of this part.

[0113] The above is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent replacement and improvement made by any person skilled in the art within the technical scope disclosed by the present invention and within the spirit and principle of the present invention shall be covered within the protection scope of the present invention. At the same time, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. An embodied intelligent tennis training robot, comprising a wheel and foot module and a perception module; characterized in that: It also includes a ball receiving and sending module, a feedback module and an electronic control module; The serving and receiving module is located above the wheel-foot module, and can control the serving pitch angle and ball delivery speed of the robot in the serving mode; when switched to the receiving mode, the tennis ball can be recovered; The feedback module is arranged on the front of the robot, and provides visual feedback on the trainee's batting action through a light strip and a display screen; The electronic control module includes a main control board, an analysis board and a power supply component, which are responsible for motion control and perception feedback data processing; the main control board coordinates the work of the ball receiving and sending module and the wheel and foot module; the analysis board receives and processes data from the perception module, generates scientific decisions, and conducts human-computer interaction with the trainer through the feedback module.

2. The embodied intelligent tennis training robot according to claim 1, characterized in that: The ball-receiving and serving module is mainly composed of a friction wheel mechanism, a ball-shooting mechanism, a ball-storing mechanism and a ball-receiving and serving switching mechanism; The friction wheel mechanism drives the friction wheel to rotate at different speeds and directions through a pair of AC brushless motors, realizing the outward launch mode and the inward collection mode of the tennis ball; the pitch adjustment motor controls the angle of the friction wheel fixing bracket to realize the shooting angle adjustment in the serving state and the control of the friction wheel angle when switching between the serving and collection states; The shooting mechanism includes a three-way shooting barrel. Under the action of the electric push rod, the tennis ball in the shooting barrel can be pushed to contact the friction wheel and shot out. When the push rod is retracted, the next tennis ball to be launched automatically falls down under the action of gravity and enters the shooting preparation state. The ball storage mechanism is located on the top of the robot, and the retrieved tennis balls fall into the ball storage bin; The serve-and-receive switching mechanism gives the friction wheel the ability to move forward and backward and swing. In the serve mode, the friction wheel mechanism is stored in the robot's abdomen; In the ball collecting mode, the friction wheel mechanism, driven by the push rod and the servo, extends from the robot's abdomen and swings to a height close to the ground to pick up the tennis ball on the ground.

3. The embodied intelligent tennis training robot according to claim 2, characterized in that: A stirring arm is arranged inside the ball storage bin. Under the action of the stirring arm inside the ball storage bin, the tennis balls at the bottom layer are kept moving slowly to prevent the tennis balls from being blocked.

4. The embodied intelligent tennis training robot according to claim 1, characterized in that: The wheel-foot module is located at the bottom of the robot and is responsible for support, movement, steering and height adjustment; The perception module consists of a forward-facing binocular camera and a rear-facing event camera, which are used to capture the trainer's movements, the robot's position on the court, and the landing point of the tennis ball; The feedback module includes a display screen located at the top of the front of the robot and light strips distributed throughout the body of the robot, which are leg light strips, chest light strips, upper arm light strips and forearm light strips; the arm servo is used to control the extension and tightening of the arm.

5. The embodied intelligent tennis training robot according to claim 2, characterized in that: The right swing arm servo of the transmitting and receiving switching mechanism is connected to the right swing arm in the friction wheel mechanism by bolts, and the left swing arm servo is connected to the left swing arm in the friction wheel mechanism by bolts; under the drive of the servo, the friction wheel mechanism can rotate around the servo shaft; the right swing arm servo is fixed to the right servo bracket by bolts, and the right servo bracket is fixed to the slider of the right guide rail by bolts to achieve left and right movement; similarly, the left swing arm servo is fixed to the left servo bracket by bolts, and the left servo bracket is fixed to the slider of the left guide rail by bolts; both ends of the connecting block are connected to the right servo bracket and the left servo bracket by bolts at the same time, and the middle part of the connecting block is fixed to the output end of the electric push rod by bolts; the rear end of the electric push rod is fixed to the chassis by bolts.

6. A working method of the embodied intelligent tennis training robot according to claim 1, characterized in that: The electronic control module is responsible for motion control and sensory feedback data processing; the main control board coordinates the work of the ball receiving and sending module and the wheel and foot module; the analysis board receives and processes data from the sensory module, generates scientific decisions, and interacts with the trainer through the feedback module; The wheel-foot module realizes the height adjustment, movement and direction control of the robot as a whole; the four leg motors coordinately control the bending degree of the upper and lower legs to adjust the ball position and ensure the squatting posture when receiving the ball; the speed difference between the left and right motor wheels determines the steering angle of the robot; when the left and right motor wheels have the same speed, the robot moves in a straight line; when there is a speed difference, the robot turns; The serving and receiving module has two working states: serving mode and receiving mode. In serving mode, the tennis ball in the ball storage mechanism is released into the ball shooting tube, and the electric push rod pushes the tennis ball to make it contact with the friction wheel. Driven by the AC brushless motor, it generates a preset speed and rotation, the servo controls the pitch angle, and the wheel-foot module controls the serving height and direction, and finally the tennis ball is thrown to the trainee according to the preset parameters. In receiving mode, the receiving and serving switching mechanism is started, and the electric push rod and the servo work together to move the friction wheel mechanism to the front and lower part of the robot, control the receiving speed, and the pitch adjustment mechanism adjusts the receiving angle. The wheel-foot module controls the robot to squat and adjust the direction, and finally the tennis balls scattered on the court are recovered into the ball storage mechanism. The perception module consists of a binocular camera and an event camera. It uses computer vision technology to solve the robot's position coordinates on the court, the spatiotemporal position of the trainer's human skeleton, the landing point information of the hit tennis balls, and the position information of the scattered tennis balls, providing effective data support for the analysis board of the electronic control part; the analysis board has built-in tennis motion analysis and visual motion feedback prompt algorithms, which drive the light strips and display screens in the feedback module according to the analysis results to achieve real-time guidance for the trainers; at this point, the tennis training robot has realized embodied intelligence from perception to action.

7. The working method of the embodied intelligent tennis training robot according to claim 6, characterized in that: When the ball serving and receiving module switches from the serving mode to the receiving mode, the entire friction wheel mechanism moves and extends out of the robot body under the drive of the electric push rod; the pitch adjustment motor is used to adjust the appropriate angle; the right swing arm and the left swing arm rotate counterclockwise under the drive of the right swing arm servo and the left swing arm servo, so that the right friction wheel and the left friction wheel in the friction wheel mechanism can approach and contact the tennis ball on the ground, thereby throwing the tennis ball upward and causing it to fall into the ball basin in the ball storage mechanism.

8. The working method of the embodied intelligent tennis training robot according to claim 7, characterized in that: When the tennis training robot is in the ball receiving mode, the wheel and foot module controls the robot to squat, while the ball receiving and serving module switches to the ball receiving mode and extends the friction wheel mechanism; the binocular camera in the perception module can accurately identify the position of the tennis ball in the field and send its coordinates to the electronic control module; Subsequently, the electronic control module directs the robot to move toward the tennis ball and aligns the center of the friction wheel mechanism with the tennis ball; under the action of friction, the tennis ball is thrown upward and falls smoothly into the ball basin of the ball storage mechanism, thus completing the ball collecting process.

9. The working method of the embodied intelligent tennis training robot according to claim 6, characterized in that: The tennis training robot can sense and feedback the kinematic performance of the tennis trainer's movements; after processing the data from the binocular camera in the analysis board using the existing human skeleton recognition algorithm, the position and angular velocity data of the trainer's hips, chest, shoulders, elbows, and hands are obtained; let i represent body parts, i=1 represents hips, i=2 represents chest, i=3 represents shoulders, i=4 represents elbows, and i=5 represents hands; ω i (t) represents the angular velocity vector of part i at time t, p i (t) represents the spatial position of part i at time t, v5(t) represents the spatial velocity of the hand at time t, and the contribution velocity of each part to the hand is calculated and accumulated to the observed part i according to the following formula; Where n is the cumulative variable, n = (1,...,i), and the final cumulative contribution speed v Ai (t) is a scalar; The kinematic chain analysis of tennis hitting movements mainly focuses on the degree of force and the timing of force; the degree of force CV i The calculation is done using the following formula: Assume that the time when the cumulative contribution speed of part i reaches its peak is t i , which is calculated as The timing of force application is calculated as follows: the chest lags behind the hips by Δt1=t2-t1, the shoulder lags behind the chest by Δt2=t3-t2, and the elbow lags behind the shoulder by Δt3=t4-t3.

10. The working method of the embodied intelligent tennis training robot according to claim 9, characterized in that: The tennis training robot visualizes the trainer's motion chain: the leg light belt is used to feedback the hip information, the chest light belt is used to feedback the chest information, the upper arm light belt is used to feedback the shoulder information, and the forearm light belt is used to feedback the elbow information; the color of the light belt is used to indicate the degree of force, which is green for good RGB=(0,255,0), orange for better RGB=(255,255,0), yellow for bad RGB=(255,50,0), and red for severe bad RGB=(255,0,0). The calculation method is as follows In the formula, CV iRef Indicates the reference value of the force level, R iA , R iB , R iC It is a manually set fault tolerance value; if CV1 is greater than the reference value CV 1Ref If CV4 is higher than the reference value CV 4Ref Low, RGB4 is set to (0,255,0); The light strips of each part are used to flow to provide feedback on the coordination of the timing of each part’s force application Δt i ; If there is an action delay, the first light in the light strip at that part will light up i Seconds later, the second light strip lights up, 2f i After seconds, the third light strip lights up; then the light strip flow time interval f of the i-th part of the entire light strip system i Calculated as In the formula, Δt i represents the lag time of site i, Δt iRef Indicates the standard reference value, T iA , T iB , T iC Indicates the preset tolerance value; When f i =0 means the LED in this area is always on; if Δt i Lower than the reference value Δt iRef , the direction of the light strip is from the far end to the near end of the body, prompting the trainee to slow down the timing of this part; on the contrary, if Δt i Higher than the reference value Δt iRef , the light strip flows from the proximal end to the distal end of the body, prompting the trainee to speed up the timing of this part.

Images