Pick ball partner training robot and method and system for perceiving level of trainer

Abstract

The invention relates to a Pick ball partner training robot and a method and system for perceiving the level of a trainee through the Pick ball partner training robot. The robot comprises a moving module, a ball receiving module, a ball serving module, a positioning module and a control circuit board. The level sensing method for the trainee comprises the following steps: establishing a coordinate system by using the midpoint of the bottom line of the trainee in the Pick court; adopting a Kriging agent model to construct a ball hitting expression model under the complex fixed-point ball path; establishing a training task difficulty quantification model considering cognitive complexity; calculating the performance of the trainee under the condition of determining the task difficulty so as to evaluate the comprehensive level of the return ball of the trainee; by deeply analyzing the task complexity of the players, the weak links of the players are accurately informed, targeted guidance suggestions are provided, and the improvement of the skill of the pike ball and the optimization of the training effect of the players are promoted; according to the invention, the complete Pick ball launching track control capability is realized, the pitching time, pitching speed, rotation degree and pitching angle of the Pick ball can be adjusted, and the diversity and personalization of simulated playing are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of sports robots and relates to a pickleball auxiliary training device, and more specifically, to an automated device capable of cyclically receiving and serving balls and performing sparring training with a pickleball trainee, and a method and system for sensing the hitting level of the trainee. Background Art

[0002] As an emerging sport that combines the characteristics of tennis and table tennis, pickleball has set off a global craze in recent years, and its popularity and number of participants continue to grow. Although it is relatively easy to get started with pickleball, the improvement of participants' skills is inseparable from long-term and diversified practice to improve the judgment of the ball trajectory and the fine control of various parts of the body.

[0003] Pickleball training is a human-machine interaction problem consisting of man-machine-ball. Applying intelligent training equipment and building a scientific human-machine interaction training model will greatly save training time and cost. At present, although there are a large number of table tennis, badminton, tennis and other training robots on the market, there are very few types of sparring robots for pickleball. As an immature and less widely used device, most pickleball serving machines are modified from other racket-type serving machines, and can only effectively control different feeding points, but the ball output position is fixed and the working process often requires multiple balls. After a round of training, the balls must be collected and refilled with pickleball; the existing robots lack scientific and intelligent human-machine interaction training methods and can only feed the trainee with balls according to the pre-set serving trajectory. Develop a new pickleball sparring robot, give it the ability to move quickly in all directions, effectively recover the trainee's shots, and accurately and maneuverably serve, and formulate a method for it to perceive and analyze the trainee's batting level, so that the training system can be adjusted to a suitable training difficulty, which can optimize the training efficiency. This technical equipment can help trainees improve their ball skills and save training fees, provide effective tools and new models for sports enthusiasts, and promote a new form of intelligent training that integrates man and machine. It is expected to improve training results by relying on advanced scientific and technological means and promote the intelligent development of ball sports training. Summary of the invention

[0004] The purpose of the present invention is to provide a pickle ball training robot, which has the functions of automatically and cyclically sending and receiving pickle balls and timely directional movement, and further establishes a quantitative model of the difficulty of pickle ball hitting tasks, dynamically evaluates the trainee's ball return performance, so that the robot can adapt to the trainee's performance and carry out corresponding ball feeding training, providing a scientific training tool for the trainee.

[0005] In order to solve the above technical problems, the present invention is implemented by adopting the following technical solutions, which are described as follows in conjunction with the accompanying drawings:

[0006] A pickleball training robot, comprising a moving module, a ball receiving module, a ball serving module, a positioning module, and a control circuit board;

[0007] The mobile module serves as the chassis of the robot and is arranged below the serving module, and the ball receiving module is arranged above the serving module;

[0008] The positioning module includes a binocular camera, an ultra-wideband positioning base station, and a location tag;

[0009] The ultra-wideband positioning base stations of the positioning module are set at the four corners of the pickleball court, the location tags of the positioning module are installed on the control circuit board, and the binocular camera of the positioning module is arranged on the trainer's side of the pickleball court; the control circuit board is installed on the mobile module.

[0010] Furthermore, one of the ultra-wideband positioning base stations set in the four corners is a main base station, which is connected to a computer and can read the positions of the location tags and the base stations in the four corners respectively, thereby calculating the position coordinates of the mobile module through a formula.

[0011] Furthermore, the binocular camera can collect RGB images and depth images of the scene at high speed, use a conventional neural network to identify the plane coordinates of the pickle ball in the RGB image, and combine the depth information of the binocular camera to comprehensively form the spatial position coordinates of the pickle ball; predict the landing point of the pickle ball through the pickle ball aerodynamic model, and send the result to the control circuit board via Wi-Fi; similarly, measure the position coordinates of the trainee on the court and send them to the control circuit board.

[0012] Furthermore, the mobile module uses a wheel-foot mechanism; a shell is installed above the mobile module, and the shell contains a serving module, and the serving module is connected to the ball net of the upper receiving module through a pipeline; (I marked the differences in green)

[0013] The mobile module is driven by wheels, including wheels, connecting parts and drive system. When the motor wheel rotates, the pickleball training robot can move at a variety of speeds and directions, and the wheel-foot mechanism can flexibly adjust the height and angle of the pickleball serve and receive.

[0014] The serving module launches the pickle ball in front of the trainee's coordinates, and requires the trainee to return the ball with the robot's location as the target; after the trainee hits the ball, the binocular camera captures and calculates the landing point of the pickle ball, and the control circuit board controls the motor wheel in the mobile module to rotate, so that the mobile module moves to the landing point of the pickle ball to the receiving module to receive the ball, and at the same time the control circuit board controls the serving module to launch the next pickle ball.

[0015] Furthermore, the serving module is mainly composed of a support frame and a pickle ball propulsion mechanism; the support frame is located on one side of the serving module and is connected to the chassis by bolts at the bottom; the pickle ball propulsion mechanism is composed of a support plate, a pickle ball propulsion device, a ball shooting barrel and a ball storage barrel, the support plate is fixed on the support frame, the ball shooting barrel and the ball storage barrel form a three-way pipe and are fixed by the support plate, the pickle ball propulsion device extends into the ball shooting barrel from one side opening of the three-way pipe, and the other side opening faces the gap between the two friction wheels, the upper side opening is the ball storage barrel and is connected to the ball dropping hole of the ball receiving module.

[0016] Furthermore, the ball receiving module is mainly composed of a ball receiving hole, a ball receiving net, and a bracket. The ball receiving hole is integrally connected to the ball receiving net, with a mesh hole the size of a pickle ball left in the middle, and the lower end of the mesh hole is connected to the ball shooting tube and the ball storage tube to form a three-way pipe; the ball receiving net is made of low-elasticity mesh knitted fabric, and is fixed to the ball receiving frame by four adjustable brackets; the shape of the ball net is high on all sides and low in the middle, and a ball falling hole with an aperture slightly larger than the diameter of a standard pickle ball is set in the middle, which is connected to the ball storage tube of the serving module; the bracket is installed below the ball net, the upper end supports and connects the fixed parts of the ball net, and the lower end is fixed on the chassis.

[0017] Furthermore, the control circuit board adjusts the serving speed and rotation by controlling the rotation speed of the left and right friction wheel motors in the serving module; after the trainee hits the pickle ball from the robot, the binocular camera provides the pickle ball coordinates to the control circuit board, and the ultra-wideband positioning system base station provides the robot position coordinates to the control circuit board, and the control circuit board controls the rotation of the motor wheel in the moving module to move the robot to the landing point of the pickle ball; after the ball receiving net in the ball receiving module intercepts the pickle ball hit by the trainee, the pickle ball falls into the ball receiving hole and rolls into the ball storage barrel; each time a pickle ball is launched from the shooting barrel, a pickle ball will be refilled into the shooting barrel in the ball storage barrel.

[0018] An adaptive adjustment strategy for the training difficulty of a pickleball training robot:

[0019] First, a coordinate system is established with the midpoint of the opponent's baseline in the pickleball court, the variables of the training system are defined, and the basic data for quantitative analysis of task difficulty and task performance are obtained. Secondly, the Kriging proxy model is used to construct a batting performance model under complex fixed-point ball paths, so as to expand the batting performance data of some ball paths to the batting performance under all ball paths. Third, a quantitative model of training task difficulty considering cognitive complexity is established, and the task difficulty and task complexity are scored respectively to determine the single task difficulty index and the multiple task difficulty index. Fourth, the performance of the trainee under the determined task difficulty is calculated to evaluate his comprehensive level of return. Through in-depth analysis of the complexity of the player's task, accurate insight into the player's weak links and provide targeted guidance and suggestions to promote the improvement of the player's pickleball skills and the optimization of training effects.

[0020] Furthermore, a coordinate system is established with the midpoint of the baseline of the opponent in the pickleball court to define the training system variables, including:

[0021] Let m be the number of trainings, each training contains N rounds, where i represents a round; let the task variable be in the i-th round of the m-th training This includes the speed at which the robot shoots the ball. and the distance between the landing point and the player Players in fixed-point task variables The quality of the ball return is reflected in the accuracy of the ball return. The ball return accuracy is calculated as:

[0022]

[0023] In the formula, is the deviation between the actual point where the ball lands and the target point; A is the set error resolution.

[0024] Furthermore, a batting performance model under complex fixed-point ball paths is constructed, and the specific contents are as follows:

[0025] In the entire m training, there are N different combinations of task variables; let represents the set of task variables in the mth training. The set of the player’s return quality actually measured during the human-computer training is:

[0026]

[0027] Using the Kriging prediction model, the player's performance on the task variable x is predicted based on the test results of all historical task variables. m The mass of the ball returned Y(x m ):

[0028] Y(x m )=Kriging{y(x m )}

[0029] The predicted value of the ball return performance for a certain task variable is:

[0030] Furthermore, a training task difficulty quantification model considering cognitive complexity is established, including:

[0031] For the task variable x m Difficulty of hitting the ball C s (x m ) is expressed as follows, where ε is the task difficulty correction factor;

[0032]

[0033] Set batting combination In the above example, let the joint probability distribution of ball path i be X i (d,v), the variation between this ball path and the previous ball path is

[0034]

[0035] The cognitive load of a particular shot is for

[0036]

[0037] In the mth training cycle, let the cognitive complexity of the ith shot be We believe that the cognitive differences in each shot produced by the player are related to the degree of difference in the previous shots in the same training session. Therefore, the exponentially weighted moving average method is used to calculate the cognitive complexity of this shot:

[0038]

[0039] Among them, β is the weighting coefficient;

[0040] The difficulty index of the task in the i-th round of the m-th training Indicates the difficulty of hitting the ball and cognitive complexity The product of:

[0041]

[0042] The total task difficulty index TCI of the mth training m is the average task difficulty of the entire training process:

[0043]

[0044] Furthermore, the comprehensive level of the trainees' ball return is evaluated, including the following:

[0045] In the i-th round of the m-th training, the task difficulty index The player's return accuracy is The comprehensive return score P of the mth training is m (TCI m )for:

[0046]

[0047] A pickleball training robot's horizontal perception system for a trainer, comprising:

[0048] The basic data acquisition module is used to establish a coordinate system based on the midpoint of the baseline of the opponent in the pickleball court, define the variables of the training system, and obtain basic data for quantitative analysis of task difficulty and task performance;

[0049] Constructing a batting performance model module, which is used to construct a batting performance model under complex fixed-point ball paths using a Kriging proxy model, thereby expanding the batting performance data of a part of the ball paths to the batting performance under all the ball paths;

[0050] Establish a training task difficulty quantification model module, which is used to establish a training task difficulty quantification model that takes cognitive complexity into account, and score task difficulty and task complexity respectively, so as to determine the single task difficulty index and multiple task difficulty index;

[0051] The calculation and evaluation module is used to calculate the trainee's performance under a certain task difficulty, so as to evaluate the trainee's comprehensive level of ball return; through in-depth analysis of the complexity of the player's task, it can accurately understand the player's weak links and provide targeted guidance and suggestions, so as to promote the improvement of the player's pickleball skills and optimize the training effect.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The wheel-footed pickleball training robot with integrated sending and receiving functions described in the present invention realizes complete pickleball launching trajectory control capability through a clever mechanical structure layout. The pickleball serving time, serving speed, rotation degree, and serving angle can all be adjusted, thereby realizing the diversity and personalization of simulated playing.

[0054] 2. The method for quantifying the complexity of the pickleball playing robot training task described in the present invention comprehensively considers the influence of differences in ball paths such as the distance between the incoming ball and the trainee and the incoming ball speed, and can calculate and analyze the difficulty of the training task in real time in combination with the training site data, thereby solving the problem that traditional robots lack perception and analysis capabilities, and quantifying the comprehensive hitting ability of the trainee, providing scientific decision-making for the design of targeted training programs. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] 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:

[0056] Figure 1 This is an axonometric diagram of the overall appearance of the pickleball training robot of the present invention;

[0057] Figure 2 It is an axonometric diagram of the combination of the mobile module and the serving module of the pickleball training robot of the present invention;

[0058] Figure 3 This is an axonometric diagram of a mobile module of the pickleball training robot according to the present invention;

[0059] Figure 4 This is an axonometric diagram of the ball receiving module of the pickleball training robot according to the present invention;

[0060] Figure 5 Axonometric diagram of the serving module of the pickleball training robot according to the present invention;

[0061] Figure 6 It is a left view of the serving module of the pickleball training robot according to the present invention;

[0062] Figure 7 It is a front view of the serving module of the pickleball training robot according to the present invention;

[0063] Figure 8 A detailed view of the light strip of the pickleball training robot according to the present invention;

[0064] Fig. 9 A schematic diagram of the ball collecting function of the pickleball training robot according to the present invention;

[0065] Fig.10 This is a schematic diagram of the working scene of the pickleball training robot according to the present invention;

[0066] Fig.11 This is a flow chart of the control method of the pickleball training robot according to the present invention;

[0067] Fig.12 A schematic diagram of the human-computer interaction method and variables of the present invention;

[0068] Fig.13 This is a flow chart of the trainee level perception method described in the present invention.

[0069] Description of labels:

[0070] Net bag 1; moving module 2; ball receiving module 3; serving module 4; camera group 5; strip light belt 6; pickle ball 7;

[0071] Left motor wheel 2-1; right motor wheel 2-2; wheel set 2-3; left engine set (left) 2-4-1; left engine set (right) 2-4-2; right engine set (left) 2-5-1; right engine set (right) 2-5-2;

[0072] Ball receiving net 3-1; ball receiving hole 3-2, bracket 3-3;

[0073] Chassis 4-1; right friction wheel 4-2; left friction wheel 4-3; right motor 4-4; left motor 4-5; ball storage cylinder 4-6; push rod 4-7; friction wheel bracket (two) 4-8; ball launching hole 4-9;

[0074] Left camera 5-1; Right camera 5-2;

[0075] Left light strip 6-1; upper light strip 6-2; right light strip 6-3; lower light strip 6-4;

[0076] Positioning base station 7-1; positioning tag 7-2. DETAILED DESCRIPTION

[0077] 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.

[0078] 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.

[0079] 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.

[0080] like Figure 1 The figure shows the overall appearance of the pickle ball training robot of the present invention. The dual-wheeled integrated pickle ball training robot adopts a modular design, wherein the ball receiving module 3 and the moving module 2 are respectively located above and below the robot, the net bag 1 is in the middle of the robot, and the ball serving module 4 is inside the shell.

[0081] The net bag 1 is made of low-elasticity mesh knitted fabric to buffer the speed of the hit pickle ball and make the ball roll into the ball receiving hole to prepare for the next serve.

[0082] Mobile module 2, the control circuit board controls the rotation speed of the two motor wheels respectively to realize the directional movement of the robot. At the same time, it can control the four motor groups 2-4 and 2-5 to adjust the height and angle of the wheel group, so as to maintain the balance of the robot while realizing the functions of receiving and serving the ball at different heights and angles.

[0083] The ball receiving module 3 collects the pickleballs through the ball receiving net 3-1 supported by the bracket 3-3, and delivers the balls to the serving module 4 through the ball storage cylinder 4-6 below the ball receiving hole 3-2.

[0084] The serving module 4 can push the pickle balls collected by the collecting module 3 to the right and left friction wheels 4-2 and 4-3 through the push rod 4-7. The right and left motors 4-4 and 4-5 control the rotation of the friction wheels to make the pickle balls shoot out from the ball outlet 4-9 with adjustable time, speed, rotation and direction.

[0085] The positioning module includes left and right binocular cameras 5-1 and 5-2, which can measure the spatial coordinates of the pickleball, the robot, and the trainer.

[0086] The strip light strip 6 is rectangular and fixed around the net bag. It can display different colors in real time according to the trainee's hitting effect in the form of clockwise left light strip 6-1, upper light strip 6-2, right light strip 6-3, and lower light strip 6-4 sequence flow, giving simple feedback on the hitting effect to the hitter.

[0087] like Figure 2 The figure shows the axonometric view of the combination of the mobile module 2 and the serving module 4 of the present invention, see Figure 3 and Figure 7 This is a view showing two modules presented separately. The serving module 4 is installed on the chassis of the mobile module 2 through the chassis 4-1. When the serving module adjusts the shooting angle, the chassis 4-1 supports the serving module to adjust the angle together.

[0088] like Figure 3 The figure shows an axonometric view of the mobile module of the present invention, wherein the metal chassis 4-1 and the four motor wheel sets 2-4 and 2-5 are symmetrically distributed, and the control circuit board is bonded in the middle of the lower end of the chassis.

[0089] like Figure 4 The figure shows the isometric view of the ball receiving module of the present invention. The bracket 3-3 is fixed on the four corners of the chassis 4-1 to support the net bag 1. The ball receiving net 3-1 is integrated with the ball receiving hole 3-2. The ball receiving net 3-1 is concave with the ball receiving hole 3-2 as the center, so that the pickle ball can roll smoothly into the ball receiving hole 3-2.

[0090] like Figure 5 , Figure 6 and Figure 7The figures show the axonometric view, left view and front view of the ball receiving module of the present invention. The left and right cameras 5-1 and 5-2 are symmetrically distributed and fixed on the chassis 4-1. The serving hole 4-9 faces the gap area between the two friction wheels. After the pickle ball is pushed out of the three-way pipe, it is ejected by the friction force generated by the right friction wheel 4-2 and the left friction wheel 4-3. The pickle ball 7 in the ball storage tube 4- enters the serving hole 4-9. After receiving the serving time signal, the push rod 4-7 pushes the pickle ball in the ball storage tube to the gap between the left and right friction wheels 4-2 and 4-3. The pickle ball is ejected through the friction and extrusion of the two friction wheels. The friction force and relative speed of the two friction wheels determine the ejection speed and rotation degree of the pickle ball. The initial velocity of the pickle ball ejection is adjusted by changing the rotation speed of the friction wheel; if the linear speeds of the two friction wheels are the same, the ball does not rotate, and if there is a difference in the linear speed, a rotating ball is ejected. The pitch angle of the serving mechanism is adjusted by controlling the angle of the chassis 4-1, thereby adjusting the height angle of the ejected pickle ball. Based on this, the robot can achieve accurate serve.

[0091] like Figure 8 The figure shows a detail view of the strip light belt described in the present invention, in which the light belt 6 is rectangular and fixed around the net bag 1 .

[0092] like Fig. 9 The figure shows the ball receiving function of the pickle ball training robot of the present invention, that is, the working principle of the ball receiving module. The pickle ball 7 flies from the opposite side of the court, lands and jumps, touches the net bag 1, decelerates and falls into the ball receiving net 3-1; the ball receiving net 3-1 is made of low-elasticity mesh knitted fabric elastic material, so the pickle ball is difficult to bounce out of the ball receiving net and is easy to enter the ball receiving hole 3-2. Fig. 9 By adjusting the shape of the wheel set 2-3, the chassis 4-1 is driven to pitch, and the pitch angle or ball receiving area of ​​the net bag 1 and the ball receiving module 4 can be adjusted.

[0093] Fig.10 The figure shows a working scene of the pickleball training robot. A positioning base station 7-1 is arranged at each of the four corners of the pickleball court, and a positioning tag 7-2 is installed on the control circuit board of the pickleball training robot to collect the real-time position coordinates of the pickleball training robot and the trainer.

[0094] The positioning principle of the robot is that four base stations are fixed and a location tag is installed on the robot's mobile module. Then the system can calculate the coordinates of the robot through the distance between the location tag and the four base stations.

[0095] It is a prior art that an ultra-wideband positioning base station provides the robot position coordinates to a control circuit board.

[0096] It is an existing technology that one of the four base stations is a main base station connected to a computer and can read the positions of the tags and the four base stations respectively, thereby calculating the position coordinates of the mobile module through a formula.

[0097] The left and right cameras 5-1 and 5-2 are combined to form a binocular camera, which can collect RGB images and depth images of the scene at high speed, use a conventional neural network to identify the plane coordinates of the pickleball in the RGB image, and combine the depth information of the binocular camera to comprehensively form the spatial position coordinates of the pickleball; similarly, the coordinates of the trainer are obtained. The pickleball landing point is predicted by the following pickleball aerodynamic model, and the result is sent to the pickleball robot via Wi-Fi.

[0098] Binocular cameras are used to identify the spatial position of pickleballs after being trained with a neural network, which is an existing technology;

[0099] Pickleball aerodynamic models predict where a pickleball will land, which is prior art;

[0100] It is an existing technology that a binocular camera predicts the location of a pickleball and sends the result to a control circuit board via Wi-Fi.

[0101] The tracking algorithm proposed in the linear Kalman filter model that combines prediction information is used for prediction. However, when the number of prediction frames is too large, the accuracy is low. At this time, the prediction is mainly used to assist tracking and improve tracking accuracy. Considering that the pickle ball is mainly affected by gravity and air resistance during movement, a simplified model of the pickle ball position P(t) and time t can be established to fit the pickle ball motion trajectory:

[0102]

[0103] When calculating, the number of frames is used instead of time. For each three-dimensional coordinate P(t) calculated during tracking, an equation can be established. At least three points can be used to fit the motion trajectory of the pickleball. The model will be updated each time the three-dimensional coordinates of the pickleball are calculated, and the target landing point P will be predicted based on the updated model. drop and speed v drop After successfully acquiring the coordinate position of the pickle ball in three-dimensional space, these coordinate information can be used to track the motion trajectory of the pickle ball and provide support for the control of the pickle ball training robot.

[0104] In a round of training, the serving module of the pickleball robot first launches the ball to 1 meter in front of the trainer's coordinates, and the trainer is required to return the ball with the robot's position as the target. After the trainer returns the ball, the binocular camera captures and calculates the landing point of the pickleball. Under the control of the control circuit board, the robot's mobile module moves to the landing point of the pickleball to receive the ball, and the serving module launches the next pickleball.

[0105] The binocular camera captures and calculates the landing point of the pickle ball, which is the "linear Kalman filter model predicts the landing point of the pickle ball."

[0106] The control circuit board controls the motor wheel in the mobile module 2 to rotate, so that the mobile module moves to the point where the pickle ball lands, and the ball receiving module receives the ball. At the same time, the control circuit board controls the ball serving module to launch the next pickle ball. This is the prior art.

[0107] Fig.11 The flowchart of the control method of the pickle ball training robot described in the present invention. The control circuit board adjusts the serving speed and rotation by controlling the rotation speed of the left friction wheel motor 4 and the right friction wheel motor in the serving module 4; the control circuit board controls the telescopic position of the electric push rod 4-7. After the trainee hits the pickle ball from the robot, the binocular camera provides the pickle ball coordinates to the control circuit board, and the UWB provides the robot position coordinates to the control circuit board. The control circuit board controls the rotation of the motor wheel 2-4 and the motor wheel 2-5 in the moving module 2 to move the robot to the location where the pickle ball lands. Finally, after the ball receiving net 3-1 in the ball receiving module 3 intercepts the pickle ball hit by the trainee, the pickle ball falls into the ball receiving hole 3-2 and rolls into the ball storage barrel 4-6. Each time a pickle ball is launched from the shooting barrel, a pickle ball will be refilled in the ball storage barrel 4-6 to the shooting barrel.

[0108] like Fig.13 As shown, the trainer's level perception method is specifically described as follows:

[0109] 1. Establish a coordinate system based on the midpoint of the opponent's baseline on the pickleball court and define the training system variables

[0110] During the training process, the robot's serve pattern, the relative position of the robot and the player, and the player's return performance are constantly changing. Let m be the number of training sessions, and each training session contains N rounds, where i represents a certain round. As the number of training sessions accumulates, the player's overall level will gradually improve. In many racket ball games, including pickleball, winning mainly depends on speed and accuracy. The robot's ball speed and the distance between the landing point and the trainer are the most important task variables. The trainer's return speed and accuracy under the task variables are the representation of the player's level. For example Fig.12 As shown, in the i-th round of the m-th training, the task variable is This includes the speed at which the robot shoots the ball. and the distance between the landing point and the player Players in fixed-point task variables The quality of the ball return is mainly reflected in the accuracy of the ball return. The ball return accuracy is calculated as:

[0111]

[0112] In the formula, is the deviation between the actual point where the ball lands and the target point; A is the set error resolution.

[0113] 2. Modeling the performance of complex ball paths through fixed-point ball training

[0114] In the entire m training, there are N different combinations of task variables. represents the set of task variables in the mth training. The set of the player’s return quality actually measured during the human-computer training is:

[0115]

[0116] Due to the variety of combinations of task variables, it is impossible to traverse all combinations during human-computer interaction. Therefore, this paper intends to use the Kriging model with high computational efficiency and strong nonlinear modeling capabilities to scientifically predict the player's response to all task variables x based on the test results of some task variables. m The quality of the ball return:

[0117] Y(x m )=Kriging{y(x m )}

[0118] The predicted value of the ball return performance for a certain task variable is:

[0119] 3. Calculation method of training task difficulty considering cognitive complexity The difficulty of dynamic batting training needs to consider not only the difficulty of hitting a certain ball path for the trainee, but also the cognitive complexity of the batting combination for the trainee. m The performance of the return ball is different. The same task is not consistent for trainees of different levels. m The difficulty of hitting the ball is negatively correlated with the trainee's return performance. Therefore, for the task variable x m The batting difficulty can be expressed as follows, where ε is the task difficulty correction coefficient.

[0120]

[0121] In batting combination In the above example, let the joint probability distribution of ball path i be X i (d,v), the variation between this ball path and the previous ball path is

[0122]

[0123] The cognitive load of a particular shot is for

[0124]

[0125] In the mth training cycle, let the cognitive complexity of the ith shot be We believe that the cognitive differences in each shot produced by the player are related to the degree of difference in the previous shots in the same training session. Therefore, the exponentially weighted moving average method is used to calculate the cognitive complexity of this shot:

[0126]

[0127] Among them, β is the weighting coefficient, which controls the weighted descent speed. The smaller its value, the faster it descents. This formula takes the size of the minimum quantization error as the cornerstone, combines the influence of the previous cognitive complexity on this calculation, and uses the weighting coefficient to control the influence and its range.

[0128] The difficulty index of the task in the i-th round of the m-th training It is expressed as the product of the difficulty of hitting the ball and the cognitive complexity:

[0129]

[0130] Therefore, the total task difficulty index TCI of the mth training m is the average task difficulty of the entire training process:

[0131]

[0132] IV. Evaluation of the comprehensive level of the trainee's ball return

[0133] The core of the sparring robot's autonomous perception and interaction is to effectively evaluate the comprehensive level of the players and formulate corresponding sparring plans so that the difficulty index of the training task matches the player's level. The player's return accuracy is The comprehensive return score P of the mth training is m (TCI m )for:

[0134]

[0135] like Fig.13 As shown, the specific numerical examples of the trainer's level perception method are as follows:

[0136] The mth (m=5) batting practice is performed, which consists of a set of ball paths x 5 composition:

[0137] 1) In this training, the deviation value d between the actual return ball landing point and the target point error (x 5 ) = {10,20,30,40,50}, substitute Where A = 10 cm, calculate the ball return accuracy y(x 5 ) = {1, 0.5, 0.33, 0.25, 0.2};

[0138] 2) Combine historical batting performance to build a Kriging model of player performance Substitute x 5 predict

[0139]

[0140] 3) Bring in arrive ε is 0.8, and we get C s (x 5 ) = {0.82, 1.31, 2.50, 3.48, 4.21};

[0141] The variation between the paths is C i (x 5 ) = {1.00, 0.47, 0.35, 0.61, 0.35}, substituting E(x 5 ) = {0, 0.0007, 0.025, 0.0001, 0.0018}, substituting Formula, set β = 0.5, assign initial value Get C d (x 5 ) = {13.32, 25.67, 55.14, 110.28, 220.64};

[0142] 4) According to In

[0143] 5) Bring in The final result is 71.07.

[0144] The present invention provides a pickleball training robot's level perception system for a trainer, comprising:

[0145] The basic data acquisition module is used to establish a coordinate system based on the midpoint of the baseline of the opponent in the pickleball court, define the variables of the training system, and obtain basic data for quantitative analysis of task difficulty and task performance;

[0146] Constructing a batting performance model module, which is used to construct a batting performance model under complex fixed-point ball paths using a Kriging proxy model, thereby expanding the batting performance data of a part of the ball paths to the batting performance under all the ball paths;

[0147] Establish a training task difficulty quantification model module, which is used to establish a training task difficulty quantification model that takes cognitive complexity into account, and score task difficulty and task complexity respectively, so as to determine the single task difficulty index and multiple task difficulty index;

[0148] The calculation and evaluation module is used to calculate the trainee's performance under a certain task difficulty, so as to evaluate the trainee's overall level of ball return; through in-depth analysis of the complexity of the player's task, it accurately understands the player's weak links and provides targeted guidance and suggestions.

[0149] Promote the improvement of players' pickleball skills and optimize training results.

[0150] 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. A pickleball training robot, characterized in that: It includes a moving module, a ball receiving module, a ball serving module, a positioning module, and a control circuit board; The mobile module serves as the chassis of the robot and is arranged below the serving module, and the ball receiving module is arranged above the serving module; The positioning module includes a binocular camera, an ultra-wideband positioning base station, and a location tag; The ultra-wideband positioning base stations of the positioning module are set at the four corners of the pickleball court, the location tags of the positioning module are installed on the control circuit board, and the binocular camera of the positioning module is arranged on the trainer's side of the pickleball court; the control circuit board is installed on the mobile module.

2. A pickleball training robot according to claim 1, characterized in that: The mobile module uses a wheel-foot mechanism; a housing is installed above the mobile module, and the housing contains a serving module, which is connected to the ball net of the upper ball receiving module through a pipeline; The mobile module is driven by wheels, including wheels, connecting parts and drive system. When the motor wheel rotates, the pickleball training robot can move at a variety of speeds and directions, and the wheel-foot mechanism can flexibly adjust the height and angle of the pickleball serve and receive. The serving module launches the pickle ball in front of the trainee's coordinates, and requires the trainee to return the ball with the robot's location as the target; after the trainee hits the ball, the binocular camera captures and calculates the landing point of the pickle ball, and the control circuit board controls the motor wheel in the mobile module to rotate, so that the mobile module moves to the landing point of the pickle ball to the receiving module to receive the ball, and at the same time the control circuit board controls the serving module to launch the next pickle ball.

3. A pickleball training robot according to claim 1, characterized in that: The serving module is mainly composed of a support frame and a pickle ball propulsion mechanism; the support frame is located on one side of the serving module and is connected to the chassis by bolts at the bottom; the pickle ball propulsion mechanism is composed of a support plate, a pickle ball propulsion device, a ball shooting barrel and a ball storage barrel, the support plate is fixed on the support frame, the ball shooting barrel and the ball storage barrel form a three-way pipe and are fixed by the support plate, the pickle ball propulsion device extends into the ball shooting barrel from one side opening of the three-way pipe, and the other side opening faces the gap between the two friction wheels, the upper side opening is the ball storage barrel and is connected to the ball dropping hole of the ball receiving module.

4. The pickleball training robot according to claim 1, characterized in that: The ball receiving module is mainly composed of a ball receiving hole, a ball receiving net and a bracket. The ball receiving hole is integrally connected to the ball receiving net, with a mesh hole the size of a pickle ball left in the middle, and the lower end of the mesh hole is connected to the ball shooting tube and the ball storage tube to form a three-way pipe; the ball receiving net is made of low-elasticity mesh knitted fabric and is fixed to the ball receiving frame by four adjustable brackets; the shape of the ball net is high on all sides and low in the middle, with a ball falling hole with an aperture slightly larger than the diameter of a standard pickle ball in the middle, which is connected to the ball storage tube of the serving module; the bracket is installed below the ball net, the upper end supports and connects the fixed parts of the ball net, and the lower end is fixed on the chassis.

5. The method for sensing the level of a trainee by a pickleball training robot according to claim 1, characterized in that: First, a coordinate system was established with the midpoint of the baseline of the opponent on the pickleball court to define the variables of the training system and obtain basic data for quantitative analysis of task difficulty and task performance; Secondly, the Kriging proxy model is used to construct a batting performance model under complex fixed-point ball paths, thereby expanding the batting performance data of some ball paths to the batting performance under all ball paths; Third, a training task difficulty quantification model that takes cognitive complexity into account is established, and task difficulty and task complexity are scored separately to determine the single task difficulty index and the multiple task difficulty index; Fourth, calculate the trainee's performance under a certain task difficulty, so as to evaluate the trainee's comprehensive level of ball return; through in-depth analysis of the complexity of the player's task, accurately understand the player's weak links and provide targeted guidance and suggestions, so as to promote the improvement of the player's pickleball skills and optimize the training effect.

6. The method for sensing the level of a trainee by a pickleball training robot according to claim 5, characterized in that: A coordinate system is established based on the midpoint of the opponent's baseline on the pickleball court, and the training system variables are defined, including: Let m be the number of trainings, each training contains N rounds, where i represents a round; let the task variable be in the i-th round of the m-th training This includes the speed at which the robot shoots the ball. and the distance between the landing point and the player Players in fixed-point task variables The quality of the ball return is reflected in the accuracy of the ball return. The ball return accuracy is calculated as: In the formula, is the deviation between the actual point where the ball lands and the target point; A is the set error resolution.

7. The method for sensing the level of a trainee by a pickleball training robot according to claim 6, characterized in that: Construct a batting performance model under complex fixed-point ball paths, the specific contents are as follows: In the entire m training, there are N different combinations of task variables; let represents the set of task variables in the mth training. The set of the player’s return quality actually measured during the human-computer training is: Using the Kriging prediction model, the player's response to the task variable x is predicted based on the test results of all historical task variables. m The ball mass Y(x m ): Y(x m )=Kriging{y(x m )} The predicted value of the ball return performance for a task variable is:

8. The method for sensing the level of a trainee by a pickleball training robot according to claim 7, characterized in that: Establish a training task difficulty quantification model that takes cognitive complexity into account, including: For the task variable x m Difficulty of hitting the ball C s (x m ) is expressed as follows, where ε is the task difficulty correction factor; Set batting combination In the above example, let the joint probability distribution of ball path i be X i (d,v), the variation between this ball path and the previous ball path is The cognitive load of a particular shot is for In the mth training cycle, let the cognitive complexity of the ith shot be We believe that the cognitive differences in each shot produced by the player are related to the degree of difference in the previous shots in the same training session. Therefore, the exponentially weighted moving average method is used to calculate the cognitive complexity of this shot: Among them, β is the weighting coefficient; The difficulty index of the task in the i-th round of the m-th training Indicates the difficulty of hitting the ball and cognitive complexity The product of: The total task difficulty index TCI of the mth training m is the average task difficulty of the entire training process:

9. The method for sensing the level of a trainee by a pickleball training robot according to claim 8, characterized in that: Evaluate the comprehensive level of the trainee's ball return, including the following: In the i-th round of the m-th training, the task difficulty index The player's return accuracy is The comprehensive return score P of the mth training is m (TCI m )for:

10. A pickleball training robot's horizontal perception system for a trainer, characterized in that: include: The basic data acquisition module is used to establish a coordinate system based on the midpoint of the baseline of the opponent in the pickleball court, define the variables of the training system, and obtain basic data for quantitative analysis of task difficulty and task performance; Constructing a batting performance model module, which is used to construct a batting performance model under complex fixed-point ball paths using a Kriging proxy model, thereby expanding the batting performance data of a part of the ball paths to the batting performance under all the ball paths; Establish a training task difficulty quantification model module, which is used to establish a training task difficulty quantification model that takes cognitive complexity into account, and score task difficulty and task complexity respectively, so as to determine the single task difficulty index and multiple task difficulty index; The calculation and evaluation module is used to calculate the trainee's performance under a certain task difficulty, so as to evaluate the trainee's comprehensive level of ball return; through in-depth analysis of the complexity of the player's task, it can accurately understand the player's weak links and provide targeted guidance and suggestions, so as to promote the improvement of the player's pickleball skills and optimize the training effect.

Images