A pick-up ball intelligent training robot based on AI and cloud cooperation and an interactive training method thereof

Abstract

This invention provides a pickleball intelligent training robot and its interactive training method, belonging to the technical field of intelligent sports training equipment. The robot includes a frame, a ball storage and serving system, a multi-degree-of-freedom robotic arm, a visual acquisition system, a visual prompting system, a central processing unit, a human-computer interaction module, a cloud communication unit, a health and safety monitoring module, an autonomous mobile chassis, and a solar-powered auxiliary power supply module. Its core functionality lies in capturing player movements, ball trajectories, and facial expressions in real time through the visual acquisition system. The central processing unit, based on a deep learning model, integrates multi-dimensional analysis of the player's hitting habits, biomechanical posture, emotional state, and physiological indicators, and calls upon cloud data to dynamically generate and adjust personalized training strategies in real time. This invention achieves highly human-like, personalized, interactive, and safe intelligent pickleball training, suitable for training needs across all scenarios from amateur enthusiasts to professional athletes, significantly improving the scientific rigor, relevance, and sustainability of training.

Description

Technical Field

[0001] This invention relates to the field of intelligent sports training equipment technology, specifically to a peak ball intelligent training robot and its interactive training method that integrates computer vision, artificial intelligence, robotics, affective computing, health monitoring, environmental perception, cloud collaboration and energy-saving power supply technologies. Technical Background

[0002] Pickleball is a racket sport that combines features of tennis, badminton, and table tennis, and has rapidly gained popularity worldwide in recent years. Its technical characteristics include fast-paced net play (kitchen zone strategy), diverse serve and return strategies, and high demands on reaction speed, footwork, and tactical anticipation. However, compared to established sports like tennis and table tennis, pickleball's high-quality, professional training system is still under development, particularly lacking intelligent training equipment that can simulate real opponents, provide personalized feedback, offer safety monitoring, and support social competition.

[0003] Pickleball has unique rules and technical challenges, such as: 1) The Non-Volley Zone (NVZ) rules require players to volley within a specific area at the net, demanding centimeter-level precision in footwork control and shot selection during training; 2) The aerodynamic characteristics of the pickle (made of porous polymer material) differ significantly from those of tennis and badminton. Its flight speed is slower, but its descent trajectory is unique, and the generation and handling of spin effects differ. Existing general-purpose ball machines or post-match analysis systems have failed to provide in-depth adaptation and real-time training intervention for these unique technical aspects of pickleball.

[0004] Currently, there are two main types of pickle training equipment on the market: one is a simple fixed-point ball-launching machine, whose function is limited to launching pickles in a fixed or simply programmed pattern. It cannot be adjusted in real time according to the trainee's performance, the training mode is monotonous and boring, and it lacks tactical focus; the other relies on human sparring partners or coaches. Although this method can provide flexible sparring and guidance, it is costly, coaching resources are scarce, and the coach's observation and analysis are inevitably subjective and delayed.

[0005] In the broader field of sports technology, while some tennis or badminton serve machines incorporating sensors, or motion posture assessment systems based on video analytics, have emerged, they are often functionally fragmented. For example, serve machines lack in-depth perception and intelligent decision-making capabilities regarding athlete performance; and posture analysis systems are mostly post-event analyses, unable to provide real-time intervention during training. How to form a real-time, efficient, safe, and environmentally robust closed loop integrating perception, decision-making, execution, feedback, safety, and adaptation, and specifically customize it for the kitchen area game strategies of pickleball, the handling of unique ball spin, and the predictive footwork that matches its flight trajectory, remains a gap in existing technology.

[0006] Furthermore, existing equipment has significant shortcomings in health and safety monitoring, environmental adaptation, interpersonal interaction, emotional adaptation, system scalability, and energy conservation: it lacks real-time safety monitoring of trainees' physiological states and cannot dynamically adjust training strategies according to changes in the outdoor environment; it struggles to simulate the dynamic tactical styles of real opponents and lacks a functional architecture to support remote combat and social training; the closed system results in poor scalability, and the absence of energy-efficient power supply design makes it difficult to adapt to outdoor training scenarios. Therefore, there is an urgent need for an intelligent training system that integrates artificial intelligence, real-time visual analysis, health monitoring, environmental perception, affective computing, cloud collaboration, modular robot control, and energy-efficient power supply technologies.

[0007] Technical issues

[0008] Based on the above technical background, the existing technology mainly has the following problems:

[0009] (1) Lack of anthropomorphic and adaptive competitive training: Existing ball machines cannot simulate the tactical habits and styles of real opponents, nor can they dynamically evolve tactics and adjust difficulty according to the trainee's real-time performance. The training is not targeted and cannot be adapted to the specific technical points of the pick ball.

[0010] (2) Lack of real-time emotional adaptation and humanized incentives in the training process: Existing equipment cannot sense the trainee's emotional state (such as fatigue or frustration), and cannot adjust the training rhythm or provide emotional support according to the emotional dynamics, which affects the training experience and sustainability.

[0011] (3) Lack of health and safety monitoring: Existing training equipment generally lacks real-time, non-intrusive monitoring of trainees' physiological indicators (such as heart rate and respiration), and cannot provide safety warnings and automatic interventions during high-intensity training, posing safety hazards.

[0012] (4) The system is closed and has poor scalability: the existing equipment has fixed functions and cannot flexibly add new sensors or training modules according to future needs, resulting in a short technology life cycle.

[0013] (5) Insufficient environmental robustness and social features: Existing equipment is difficult to adapt to complex outdoor environments (such as wind and light changes), and lacks social competition functions that support remote battles and data synchronization, and the training mode is isolated.

[0014] (6) Lack of energy-saving and environmental protection design: The existing equipment does not have an energy-saving power supply solution suitable for outdoor scenarios, and the standby power consumption is high, which does not conform to the development trend of green technology.

[0015] The content of this invention

[0016] The present invention aims to overcome the shortcomings of the prior art and provide a highly intelligent, personalized, social, safe, environmentally adaptive, scalable, energy-saving and environmentally friendly intelligent training robot for pickleballs and its interactive training method.

[0017] This invention constructs a closed-loop training ecosystem centered on "real-time perception, intelligent analysis, dynamic decision-making, precise execution, instant feedback, health monitoring, environmental adaptation, emotional motivation, and energy-saving management." The system's perception layer comprises a visual acquisition system, a health and safety monitoring module, and a cloud communication unit, responsible for collecting player movements, ball trajectories, physiological indicators, environmental information, and cloud data. The central processing unit constitutes the analysis and decision-making layer, integrating deep learning, health models, affective computing, and environmental perception algorithms to formulate personalized, safe, and environmentally adaptable training strategies. The execution layer consists of a ball storage and serving system, a multi-degree-of-freedom robotic arm, a visual cueing system, and an autonomous mobile chassis, realizing the physical output of strategies, environmental shaping, and court management. The human-computer interaction module runs through the feedback, motivation, and display layers, providing real-time voice guidance, emotional interaction, health prompts, skill visualization, and social connections within the training loop. A solar-powered auxiliary power supply module provides energy-saving power support for the entire system, ensuring the continuity of outdoor training. Under the intelligent scheduling of the central processing unit and cloud collaboration, all functional modules work together to achieve a highly human-like, adaptive, safe, reliable, and environmentally robust comprehensive training experience.

[0018] This invention is implemented as follows:

[0019] This invention discloses an AI- and cloud-based intelligent training robot for pickleball, characterized by comprising: a frame, a ball storage and launching system, a multi-degree-of-freedom robotic arm, a visual acquisition system, a visual cueing system, a central processing unit, a human-computer interaction module, a cloud communication unit, a health and safety monitoring module, an autonomous mobile chassis, and a solar-assisted power supply module; the ball storage and launching system is mounted on the frame and is used to store and launch pickleballs; the multi-degree-of-freedom robotic arm is connected to the ball storage and launching system and is used to adjust the launching direction, angle, and posture; the visual acquisition system includes at least two cameras for real-time acquisition of visual information including player hitting actions, pickleball trajectory, and player facial expressions; the visual cueing system is used to project or display visual cues or prompts onto the training field. The system includes a central processing unit (CPU) electrically connected to the ball storage and launching system, the multi-degree-of-freedom robotic arm, the visual acquisition system, and the visual prompting system, serving as the core of system control and computation. A human-machine interaction module, electrically connected to the CPU, is used to set training modes, view and analyze data, and engage in social interaction. A cloud communication unit, also electrically connected to the CPU, is used for data interaction with a cloud server. A health and safety monitoring module, also electrically connected to the CPU, is used for non-contact monitoring of the trainee's physiological indicators. An autonomous mobile chassis, located at the bottom of the frame, is used to automatically retrieve scattered pickles during training breaks. A solar-powered auxiliary power supply module, electrically connected to the CPU, provides energy-saving power support for the robot.

[0020] Furthermore, the multi-degree-of-freedom robotic arm adopts a serial six-degree-of-freedom articulated structure, which includes, in sequence, a base rotation joint, a shoulder pitch joint, an elbow pitch joint, a wrist pitch joint, a wrist rotation joint, and an end flange rotation joint. Each joint is driven by a servo motor with an integrated absolute encoder and equipped with a harmonic reducer. It receives motion commands from the central processing unit via an EtherCAT bus, and the repeatability is better than ±0.5mm. The ball storage and launching system includes a ball chamber with a capacity of no less than 200 picks and an internal inclined guide plate, a spiral lifting and ball feeding mechanism driven by a stepper motor, a dual friction wheel launching mechanism driven by dual independent speed-regulating DC brushless motors, and a pneumatically assisted directional nozzle controlled by an electromagnetic valve. The central processing unit converts the target launching position into robotic arm joint angle commands through an inverse kinematics algorithm, and coordinates the control of the robotic arm posture and friction wheel speed to achieve the launching of multiple types of balls, including topspin, backspin, sidespin, high balls, and DINK balls.

[0021] Furthermore, the camera of the visual acquisition system adopts a global shutter CMOS sensor with a frame rate of no less than 120fps and a resolution of no less than 1920×1080. It is connected to the central processing unit through a gigabit Ethernet interface, and the camera is mounted on a servo-driven adjustable pan-tilt unit to achieve binocular stereo vision measurement. The visual prompting system adopts a high-lumen laser projector or an RGB LED array, supports dynamic pattern projection, and can be used in conjunction with the ball storage and launch system to achieve dynamic interference training.

[0022] Furthermore, the central processing unit adopts a high-performance embedded industrial computer based on ARM or x86 architecture, equipped with the ROS 2 real-time operating system, and integrates multiple industrial bus interfaces such as EtherCAT, CAN, SPI, and I2C, configured to perform the following functions:

[0023] Adaptive return path learning and virtual opponent simulation: Based on players' historical shot data or cloud-based professional player match data, a deep reinforcement learning tactical simulation model is used to generate return strategies targeting players' weaknesses.

[0024] Intelligent training plan generation and dynamic adjustment: Automatically generates long-term training plans based on user training goals, and dynamically adjusts the progress and content based on real-time performance;

[0025] Biomechanical posture analysis and real-time correction: Calculate player movement parameters based on 3D motion capture data, compare with professional player biomechanical benchmark models, and output improvement suggestions;

[0026] Create a dynamic digital skills profile for each user, generate radar charts, growth curves and periodic training reports through data visualization, and support report sharing.

[0027] Furthermore, the health and safety monitoring module includes a millimeter-wave radar or infrared thermal imaging unit operating in the 60GHz frequency band, with a sampling rate of not less than 10Hz, used for non-contact monitoring of the trainee's heart rate and respiratory rate; the central processing unit acquires physiological data through the SPI interface, and when it detects that the physiological indicators exceed the safety threshold, it sends an emergency stop command through the CAN bus, automatically pausing the training, issuing an audible and visual warning, and suggesting rest.

[0028] Furthermore, the central processing unit is also configured to perform the following adaptive adjustment functions:

[0029] Adaptive environmental compensation: Based on the analysis of environmental images from the visual acquisition system, light intensity, wind speed and direction are dynamically adjusted to adjust the serve strategy parameters and the display parameters of the visual cue system;

[0030] Emotion Recognition and Motivation: Based on facial expression capture from the visual acquisition system or voice data from the human-computer interaction module, the trainee's emotional state is recognized. When fatigue or frustration is detected, the training pace is automatically adjusted, the difficulty is reduced, or encouraging prompts are played.

[0031] Multi-ball mixing and dynamic interference training: Control the ball storage and serving system and the robotic arm to launch multiple types of balls in a predetermined or random sequence, while controlling the visual cue system to project moving light spot patterns, requiring players to avoid the light spots and complete the return of the ball.

[0032] Furthermore, the robot's mechanical architecture and electrical interfaces adopt a modular design, with standardized physical expansion slots and electrical interface boards reserved at the rear of the frame; the interface boards provide USB-C, RJ45, 12V / 24V DC power outputs and CAN bus interfaces for connecting detachable functional modules such as ball and net tension monitoring modules, sound field analysis microphone arrays or professional biomechanical sensors.

[0033] Furthermore, the cloud communication unit supports dual-mode communication of 5G and Wi-Fi 6, and has a built-in MQTT protocol stack; it supports multi-machine linkage and remote battle functions, enabling real-time data synchronization and collaborative training of two or more robots, and allowing players in different locations to conduct real-time remote pickleball battles through their respective robots, and record and analyze the game data; the human-computer interaction module includes a touch screen, a voice input / output unit and a wireless communication unit, and can receive remote control commands.

[0034] Furthermore, the solar-assisted power supply module includes a flexible photovoltaic panel with a peak power of not less than 200W, a lithium iron phosphate energy storage unit, and an MPPT maximum power point tracking charging controller; the central processing unit is configured with a low-power mode based on dynamic voltage and frequency adjustment, which automatically reduces the system main frequency and shuts down unnecessary peripherals when in standby or performing simple tasks; the autonomous mobile chassis adopts a four-wheel omnidirectional mobile structure, integrates a lidar and visual SLAM module, and also includes a ball recognition module; the central processing unit plans paths based on the ROS navigation stack, controls the autonomous mobile chassis to cruise and pick up balls, and the ball recognition module analyzes the surface state of the pickles and classifies them into usable balls and scrapped balls.

[0035] The interactive training method for the intelligent training robot of the pickle ball of the present invention is characterized by comprising the following steps:

[0036] Step 1: Visual Acquisition: Acquire real-time video streams of the training scene through a visual acquisition system, and identify and track key points of the player's body, the pickle, and the player's facial expressions;

[0037] Step 2: Serving Strategy Generation: Based on visual recognition results, user-preset training objectives, and cloud-based opponent data, generate serving control commands that include ball type, landing point, ball speed, and spin instructions.

[0038] Step 3: Execution of the serve: The serve control command is sent to the driver of the serve storage system and the multi-degree-of-freedom robotic arm through the real-time motion control bus, driving them to execute the serve action in coordination.

[0039] Step 4: Adaptive Adjustment: This includes emotion recognition and adaptive motivation, health and safety monitoring, and environmental adaptive compensation. It involves real-time analysis of player emotions and physiological indicators, environmental parameters, and dynamic adjustment of training intensity, serving parameters, and visual cue parameters.

[0040] Step 5 Feedback and Updates: Provide players with biomechanical posture improvement suggestions through the human-computer interaction module, update users' digital skill profiles in real time, generate data reports after training or matches, and support sharing.

[0041] Core System Structure and Hardware Implementation of the Invention

[0042] The robot consists of a hardware subsystem and an intelligent control software subsystem working together. The hardware subsystem adopts a modular and highly reliable industrial design, and the frame is constructed using aluminum alloy profiles, combining lightweight and high rigidity.

[0043] 1. Multi-degree-of-freedom robotic arm and ball storage / serving system

[0044] The multi-degree-of-freedom robotic arm adopts a serial six-axis articulated design. Each joint is driven by a servo motor with an integrated absolute encoder and a harmonic reducer. The arm span can reach 1.5 meters, the end effector load is ≥2kg, and the repeatability is ≤±0.5mm. It receives motion commands via an EtherCAT bus. The core of the ball storage and launching system is the ball tank and launching mechanism: the ball tank has a capacity of ≥200 balls and a stepper motor-driven spiral lifting mechanism at the bottom to achieve orderly ball feeding; the launching mechanism uses dual independent speed-regulating friction wheels, driven by a DC brushless motor, supplemented by an air pressure nozzle controlled by an electromagnetic valve for ball orientation and initial velocity fine-tuning, enabling precise launching of various types of pickets.

[0045] 2. Visual Acquisition System and Visual Cueing System

[0046] The visual acquisition system consists of two global shutter CMOS cameras (frame rate ≥ 120fps), mounted on a servo gimbal, and transmits data via gigabit Ethernet to achieve high-precision stereo vision and 3D motion capture; the visual cueing system uses a high-lumen laser projector or RGB LED array, supports dynamic pattern projection, and can realize dynamic interference training.

[0047] 3. Central Processing Unit

[0048] It adopts a high-performance embedded industrial control computer based on ARM or x86 architecture (such as NVIDIA Jetson AGX Orin), equipped with a real-time operating system (such as ROS2), and integrates multiple industrial bus interfaces (EtherCAT, CAN, SPI, I2C) as the control and computing core of the system, realizing multi-source data fusion, intelligent decision-making and module scheduling.

[0049] 4. Sensing and Interaction Module (Health and Safety Monitoring Module)

[0050] The health and safety monitoring module integrates a 60GHz millimeter-wave radar and outputs heart rate and respiratory waveforms via an SPI interface; the emotion recognition module includes a directional microphone array and an infrared camera to achieve emotion perception; the human-computer interaction module includes a touch screen, a voice input / output unit, and a wireless communication unit, supporting training settings, data viewing, and remote control; the cloud communication unit is a 5G / Wi-Fi 6 dual-mode communication module with a built-in MQTT protocol stack to achieve cloud data interaction and multi-machine linkage.

[0051] 5. Mobile and power supply systems

[0052] The autonomous mobile chassis is a four-wheel omnidirectional chassis that integrates LiDAR and IMU, supporting SLAM positioning and navigation as well as intelligent ball pickup; the solar auxiliary power supply module includes flexible photovoltaic panels, MPPT controller and lithium iron phosphate battery pack, supports intelligent energy management, and works with the low power mode of the central processing unit to achieve energy saving and environmental protection.

[0053] The intelligent control software subsystem is integrated into the central processing unit and built on the ROS2 framework. It adopts a node-based, loosely coupled architecture to achieve synchronous scheduling and data fusion of various functional modules, ensuring the real-time performance and stability of the system.

[0054] The core functions and innovations of this invention are as follows:

[0055] 1. Adaptive adversarial training and virtual opponent simulation

[0056] The robot learns the trainee's technical characteristics in real time through its vision system and can also access historical match data from professional players via the cloud to accurately simulate their hitting style, tactical habits, and rhythm changes. The system's built-in deep reinforcement learning algorithm model evaluates the trainee's adaptation and dynamically adjusts the simulation strategy to achieve the tactical evolution of the intelligent opponent, accurately adapting to specific technical aspects such as pickleball kitchen area games.

[0057] 2. Integrated health and safety monitoring and intelligent early warning

[0058] Integrating millimeter-wave radar or infrared thermal imaging modules, the system continuously monitors key physiological indicators such as heart rate and respiratory rate in a non-contact and imperceptible manner during training. The central processing unit has a built-in safety threshold model; once excessive fatigue or health risks are detected, the system immediately and automatically pauses training and issues an audible and visual safety alarm, ensuring training safety from a hardware perspective.

[0059] 3. Adaptive environmental compensation and outdoor adaptation

[0060] A high-precision visual acquisition system analyzes environmental images in real time, and algorithms indirectly perceive ambient light intensity, wind direction, and wind speed. Based on these environmental parameters, the central processing unit automatically fine-tunes the serve strategy and visual cue parameters, ensuring the stability and consistency of training in different outdoor environments and improving the system's environmental robustness.

[0061] 4. Dynamic digital skills profiles and growth visualization

[0062] Each registered user has a dynamically updated digital personal skills profile, recording historical performance data across multiple dimensions, including backhand, net play, footwork, and endurance. The profile uses data visualization technology to visually display the user's ability values, progress, and areas for improvement. It supports the automatic generation of periodic text and image training reports and can be shared to social media platforms via a human-computer interaction module or the cloud.

[0063] 5. Modular and scalable architecture

[0064] The robot adopts a forward-looking modular concept in its mechanical structure and electrical design. The rear of the robot has a reserved standardized physical expansion slot and a multi-functional electrical interface board, which provides USB-C, RJ45, power output and CAN bus interfaces, allowing users or third parties to flexibly plug and play various functional modules according to future needs, thus extending the technology life cycle.

[0065] 6. Cloud-based collaborative social competition and multi-machine interaction

[0066] It supports real-time data synchronization between multiple robots via the cloud, enabling remote pickleball battles between players in different locations, and provides post-match data analysis and tactical suggestions; it also supports multi-robot collaborative training, breaking the spatial limitations of training and enhancing the interactivity and competitive fun of training.

[0067] 7. Intelligent training plan generation and dynamic adjustment

[0068] Users can set training goals through the human-computer interface. The system automatically generates long-term training plans in stages and with varying levels of difficulty based on the goals and the user's skill profile. During the training process, the system dynamically adjusts the progress and content of the plan based on real-time performance data to achieve personalized training.

[0069] 8. Emotion Recognition and Adaptive Reinforcement

[0070] By using facial expression recognition and voice emotion analysis, the system can determine the trainee's emotional state (fatigue, frustration, irritability, etc.) in real time and adaptively adjust the training intensity, insert rest periods, or play encouraging voice messages to enhance the humanized experience of the training process and strengthen the continuity of training.

[0071] 9. Energy-saving, environmentally friendly, and outdoor-compatible design

[0072] The robot integrates a high-efficiency solar panel on its top, along with a lithium iron phosphate energy storage unit and an MPPT charging controller, to achieve solar-assisted power supply. The system features a low-power standby mode based on dynamic voltage and frequency adjustment, which automatically reduces the main frequency and shuts down unnecessary peripherals during standby or simple tasks, in line with the trend of green technology development and perfectly suited for outdoor training scenarios.

[0073] 10. Immersive composite training and real-time biomechanical correction

[0074] By combining the ball-serving mechanism with a visual cueing system, multi-dimensional training tasks such as mixed ball training and dynamic interference training are designed to enhance the fun of training and the intensity of training for concentration and coordination. High-precision 3D motion capture is achieved using binocular stereo vision, and specific improvement instructions are given in real time through voice synthesis technology during the return of the ball, realizing an efficient training mode of correcting errors while playing.

[0075] 11. Intelligent court management and ball sorting

[0076] The robot has a "field management" function. When training is paused, it can automatically cruise the entire field based on SLAM map and retrieve scattered peaks through the bottom ball suction mechanism. At the same time, the ball recognition module analyzes the surface condition, wear degree and deformation of the balls and automatically classifies the balls into usable balls and unusable balls, improving training efficiency and reducing consumable consumption.

[0077] Beneficial effects of the present invention

[0078] This invention provides a comprehensive pickleball training solution that integrates an intelligent opponent, AI coach, health guardian, environmental adapter, social sparring partner, emotional companion, growth archivist, court manager, and scalable platform, and has the following significant beneficial effects:

[0079] 1. Deeply customized for the key technical points of peakball, enabling human-like adaptive combat training, greatly improving the relevance and practicality of training;

[0080] 2. Integrates non-contact health and safety monitoring and intelligent early warning, ensuring the personal safety of trainees from the hardware level and eliminating safety hazards during high-intensity training;

[0081] 3. Integrating emotion recognition and adaptive incentives enables personalized training guidance, enhancing the training experience and its sustainability;

[0082] 4. It has environmental adaptive compensation capabilities, which can adapt to complex outdoor training environments and improve the environmental robustness of the system;

[0083] 5. Supports multi-machine collaboration and remote battles via cloud, breaking the limitations of training space and enhancing the social and competitive aspects of training;

[0084] 6. It adopts a modular and scalable architecture, which supports flexible expansion of functional modules, extends the equipment's technical life cycle, and reduces upgrade costs;

[0085] 7. Designed with solar-assisted power supply and low-power mode to achieve energy saving and environmental protection, perfectly adapted to outdoor training scenarios;

[0086] 8. Enables real-time biomechanical posture correction and growth data visualization, making training more scientific and efficient;

[0087] 9. Equipped with intelligent court management and ball sorting functions, improving training efficiency and reducing consumable consumption;

[0088] 10. Suitable for all training scenarios from amateur enthusiasts to professional athletes, and can be widely used in sports venues, fitness centers, professional training bases and other places. Attached Figure Description

[0089] Figure 1 This is a data transmission path diagram of the intelligent training robot for pickleball based on AI and cloud collaboration according to the present invention.

[0090] Figure 2 This is a system block diagram of the intelligent training robot for pickleball based on AI and cloud collaboration of the present invention;

[0091] Figure 3 This is a flowchart of the method for developing a pickle ball intelligent training robot based on AI and cloud collaboration according to the present invention.

[0092] Among them, 1-frame, 2-ball storage and launching system, 3-multi-degree-of-freedom robotic arm, 4-visual acquisition system, 5-visual prompting system, 6-central processing unit, 7-human-machine interaction module, 8-cloud communication unit, 9-health and safety monitoring module, 10-autonomous mobile chassis, 11-solar auxiliary power supply module. Detailed Implementation

[0093] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the following examples provide a more detailed description of the invention. It should be noted that the specific embodiments described herein are merely illustrative and not intended to limit the invention. The following are specific training scenario embodiments of this invention, and the specific working process is as follows:

[0094] Scenario: An intermediate player wants to improve his ability to handle sidespin shots on his backhand and his footwork at the net, and hopes to enhance his combat experience by playing against a friend in another location remotely. The training venue is a Peak court converted from an outdoor tennis court, with slight changes in wind and lighting on the right side.

[0095] 1. Powering on the equipment and initializing the solar auxiliary power supply module 11

[0096] After the robot is powered on, the solar auxiliary power supply module 11 immediately enters the working state: the top flexible photovoltaic panel converts light energy into electrical energy, which is then charged by the MPPT maximum power point tracking controller to charge the lithium iron phosphate energy storage unit; the power and energy management module monitors the battery level and light intensity in real time, and automatically switches to the solar priority power supply mode to provide stable power supply for all components such as the central processing unit, the multi-degree-of-freedom robotic arm, and the human-machine interaction module; the central processing unit simultaneously completes the self-test of each module, and enters the standby state after confirming that the solar auxiliary power supply is normal.

[0097] 2. Initial setup and identity recognition (operation of human-computer interaction module 7)

[0098] Users log in and select training modes via the touchscreen of the robot's human-computer interaction module 7. The touchscreen displays an interactive interface showing virtual opponents, training goals, and difficulty levels. Users select data simulating the hitting style of a professional peakball player and confirm their training goals. The robot's vision acquisition system quickly initiates identity recognition, and the central processing unit loads the user's personal digital skill profile. The human-computer interaction module 7 visually displays a radar chart comparing the user's various abilities, clearly showing issues such as a high error rate when receiving sidespin shots on the backhand and the need to improve lateral movement speed at the net. Based on the user's training goals and skill shortcomings, the system automatically generates a four-week tiered intensive training plan and displays the training stages and core training content through the human-computer interaction module 7. After user confirmation, the system issues a start command.

[0099] 3. Environmental perception and adaptive preparation

[0100] The robot vision acquisition system 4 performs a full-area scan of the training field and analyzes the image to determine that there is a slight right-side wind and the field lighting is of medium intensity. The central processing unit 6 immediately initiates environmental adaptive compensation, adds left-angle compensation to the initial serving strategy, and appropriately increases the display brightness of the visual cue system 5 to ensure that the visual cue is clearly visible and to guarantee the stability of the training.

[0101] 4. Pre-positioning of multi-degree-of-freedom robotic arms

[0102] The central processing unit 6 sends attitude pre-positioning commands to the multi-degree-of-freedom robotic arm 3 via the EtherCAT bus. The multi-degree-of-freedom robotic arm 3 is a six-degree-of-freedom serial joint structure. Driven by servo motors and harmonic reducers, the six joints of base rotation, shoulder pitch, elbow pitch, wrist pitch, wrist rotation, and end flange rotation work together to quickly adjust to the preset initial serving posture. The repeatability is better than ±0.5mm, which prepares for subsequent accurate serving.

[0103] 5. Adaptive training initiation (multi-degree-of-freedom robotic arm + ball storage and serving system in synergy)

[0104] According to the training plan, the robot first launches a sidespin ball with medium speed and spin intensity towards the user's backhand area, accurately simulating the backhand pressure tactics of the selected professional player. The central processing unit 6 uses inverse kinematics algorithms to convert the target landing point and spin type into angle commands for each joint of the multi-degree-of-freedom robotic arm and speed commands for the friction wheel motor of the ball storage and serving system 2, which are then transmitted via the EtherCAT bus. The multi-degree-of-freedom robotic arm 3 adjusts the serving direction, pitch angle, and launching posture in real time, working in conjunction with the ball storage and serving system 2 to complete a precise serve. During training, the visual acquisition system 4, health and safety monitoring module 9, and emotion recognition module continuously work synchronously.

[0105] a) Track the ball's flight trajectory in real time using a stereo vision algorithm, predict whether the final landing point is within the boundaries, and record the quality of the returned ball;

[0106] b) Based on 3D skeleton tracking technology, the user's swing angle, center of gravity transfer speed and other parameters are calculated in real time to evaluate the footwork efficiency of moving to the hitting point;

[0107] c) Real-time identification of users' emotional states through facial expression capture and micro-expression analysis;

[0108] d) The user's real-time heart rate and respiratory rate are monitored non-contactly via a 60GHz millimeter-wave radar with a sampling rate of 10Hz, and transmitted to the central processing unit 6 for security analysis.

[0109] The entire training session is continuously powered by a solar-assisted power module, ensuring long-term outdoor training endurance.

[0110] 6. Real-time analysis and multi-dimensional adaptive adjustment

[0111] The central processing unit 6 integrates and analyzes the aforementioned multi-source data to achieve dynamic and precise adjustments to the training strategy: If the user successfully returns a backhand sidespin ball three times consecutively, and their physiological indicators (heart rate, respiration) and emotional state are both in good condition, the system will gradually increase the ball speed and spin intensity, control the multi-degree-of-freedom robotic arm to dynamically adjust the ball's landing point and angle, and activate the visual cue system 5 to project a moving light spot, increasing the difficulty of dynamic interference training. If the health monitoring module detects that the user's heart rate continuously exceeds 150 beats / minute (safe threshold), the system immediately sends an emergency stop command via the CAN bus. The multi-degree-of-freedom robotic arm immediately locks all joints, stops movement, pauses all moving parts, and simultaneously issues an audible and visual warning. The human-computer interaction module provides a voice prompt: "High heart rate detected. It is recommended to pause and rest. Please pay attention to safety." If the emotion recognition module detects signs of fatigue / frustration such as frowning or sighing, the system automatically inserts a 30-second rest period and plays an encouraging prompt via the human-computer interaction module: "Keep the rhythm. Your backhand return angle is becoming more and more accurate. Keep adjusting!"

[0112] 7. Cloud-based collaborative social battles

[0113] When the training session reaches 30 minutes, the user receives a remote battle invitation from a friend in another location via the touchscreen of the human-computer interaction module. After the user confirms, the system automatically switches to remote battle mode. The two robots synchronize the real-time ball hitting data and player status through the cloud MQTT protocol. The user and friend engage in a remote pickleball battle using their respective robots, simulating real match rules. The multi-degree-of-freedom robotic arms adjust their posture in real time according to the return ball instructions from the cloud, and work with the ball storage and serving system to complete the simulated return ball. The robots record data such as the ball landing point, ball speed, and return ball quality in real time during the battle.

[0114] After the match ends, the system automatically generates a match data comparison report, which displays visual analysis of both sides' mistakes, advantageous ball paths, footwork efficiency, etc. through the human-computer interaction module, and synchronizes the match data to both sides' digital skill profiles.

[0115] 8. Real-time biomechanical correction and humanized feedback

[0116] During a backhand return, the robot's 3D motion capture data showed that the user's center of gravity leaned backward and the hitting point was too far back, causing the return to go out of bounds. In the gap when the robot was preparing for the next serve, the voice module of the human-computer interaction module immediately provided real-time correction prompts: "Press your center of gravity forward, and move the hitting point forward by 5-10 centimeters." At the same time, the touch screen of the human-computer interaction module briefly displayed an animation comparison between the key posture of this hit and the standard model of a professional player, allowing the user to intuitively see the problem.

[0117] The emotion recognition module detected that the user was briefly frustrated by the mistake, and the system then played encouraging words through the human-computer interaction module: "Your footwork was very fast just now. If you just make a slight adjustment to your hitting posture, it will be even more perfect. Keep it up!" This approach combines training correction with emotional motivation.

[0118] During training, users can view the real-time data panel of the training session at any time through the touch screen of the human-computer interaction module, including the success rate of returning the ball, ball speed distribution, footwork efficiency, and personal skill growth curve, so as to grasp the training effect in real time.

[0119] 9. Training Cycle and Tactical Evolution

[0120] The above steps form a real-time closed-loop training process that continues continuously: the robot dynamically adjusts the training difficulty and tactics based on the user's real-time performance; the multi-degree-of-freedom robotic arm continuously responds to the instructions of the central processing unit, completing various types of serve actions such as topspin, backspin, sidespin, high lob, and DINK; the health monitoring module ensures safety throughout the process; the emotion recognition module provides humanized incentives in real time; and the environmental perception module continuously adapts to changes in the outdoor environment; the training strategy, like a real intelligent opponent, is constantly upgraded as the user's performance improves, achieving precise training that "trains weaknesses and fills gaps in skills."

[0121] The solar-assisted power supply module provides continuous and stable output, while the low-power management unit works synchronously to maintain optimal energy consumption for the entire unit.

[0122] 10. Training End and Intelligent Court Management

[0123] After the user manually triggers the training end command through the human-computer interaction module, the robot immediately stops serving training, the multi-degree-of-freedom robotic arm performs a reset action, returns to the initial safe posture, and automatically enters the ball collection and court management mode: the autonomous mobile chassis 10 plans a full-coverage cruising path based on the constructed SLAM map, and automatically collects the scattered peaks in the court through the bottom ball suction mechanism; the ball recognition module performs image analysis on the surface condition, wear degree, and deformation of the collected peaks, and automatically sorts the balls into the "usable ball bin" and "discarded ball bin"; the power supply and energy management module links with the solar auxiliary power supply module to intelligently allocate the charging power of the solar photovoltaic panel according to the current battery power and light conditions, and replenishes the power of the energy storage unit through the MPPT controller.

[0124] 11. Low-power standby and data synchronization

[0125] After completing all ball collection and sorting tasks, the robot automatically returns to the training base station. The central processing unit 6 detects that there has been no user operation for an extended period of time, immediately activates the low-power mode, reduces the system's main frequency, and shuts down unnecessary peripherals such as the visual cue system 5 and the lidar. The multi-degree-of-freedom robotic arm enters a torque-locked energy-saving state. At the same time, the cloud communication unit 8 synchronously uploads all training data to the cloud server, updates the user's digital skill profile, and generates a training summary report for the session. This report is then pushed to the user's mobile phone, tablet, and other terminals through the human-computer interaction module, allowing for viewing and sharing.

[0126] In standby mode, the solar-assisted power supply module 11 maintains low-power charging to keep the battery charged and wait for the next wake-up.

[0127] Through the above embodiments, the various functions and specific implementation methods of the present invention are demonstrated in a coordinated manner, forming a complete, intelligent, safe, interactive, adaptive, user-friendly and environmentally friendly pickleball interactive training process, which fully reflects the technical advantages and practical value of the present invention.

[0128] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A pickle ball intelligent training robot based on AI and cloud collaboration, characterized in that, include: Frame (1), ball storage and launching system (2), multi-degree-of-freedom robotic arm (3), vision acquisition system (4), vision prompting system (5), central processing unit (6), human-computer interaction module (7), cloud communication unit (8), health and safety monitoring module (9), autonomous mobile chassis (10), solar auxiliary power supply module (11). The ball storage and serving system (2) is mounted on the frame (1) and is used to store and launch pickles; the multi-degree-of-freedom robotic arm (3) is connected to the ball storage and serving system (2) and is used to adjust the serving direction, angle and posture; the visual acquisition system (4) includes at least two cameras and is used to collect visual information in real time, including the player's hitting action, the trajectory of the pickle and the player's facial expression; the visual cueing system (5) is used to project or display visual cueing or interference information onto the training field; the central processing unit (6) is electrically connected to the ball storage and serving system (2), the multi-degree-of-freedom robotic arm (3), the visual acquisition system (4) and the visual cueing system (5), and is the core of system control and calculation; the human-machine interface... The human-computer interaction module (7) is electrically connected to the central processing unit (6), and the human-computer interaction module (7) is used to set training modes, view and analyze data, and conduct social interactions; the cloud communication unit (8) is electrically connected to the central processing unit (6), and the cloud communication unit (8) is used to interact with the cloud server; the health and safety monitoring module (9) is electrically connected to the central processing unit (6), and the health and safety monitoring module (9) is used to monitor the physiological indicators of trainees in a non-contact manner; the autonomous mobile chassis (10) is located at the bottom of the frame (1), and the autonomous mobile chassis (10) is used to automatically retrieve scattered pickles in the training interval mode; the solar auxiliary power supply module (11) is electrically connected to the central processing unit (6), and the solar auxiliary power supply module (11) provides energy-saving power support for the robot.

2. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The multi-degree-of-freedom robotic arm (3) adopts a serial six-degree-of-freedom joint structure, which includes a base rotation joint, a shoulder pitch joint, an elbow pitch joint, a wrist pitch joint, a wrist rotation joint, and an end flange rotation joint in sequence. Each joint is driven by a servo motor with an integrated absolute encoder and equipped with a harmonic reducer. It receives motion commands from the central processing unit via an EtherCAT bus, and the repeatability is better than ±0.5mm. The ball storage and launching system (2) includes a ball chamber with a capacity of not less than 200 picks and an inclined guide plate, a spiral lifting ball feeding mechanism driven by a stepper motor, a dual friction wheel launching mechanism driven by dual independent speed-regulating DC brushless motors, and a pneumatically assisted directional nozzle controlled by a solenoid valve. The central processing unit (6) converts the target serving position into a robotic arm joint angle command through an inverse kinematics algorithm, and coordinates the robotic arm posture and friction wheel speed to achieve multi-ball launches such as topspin, backspin, sidespin, high lob, and DINK.

3. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The camera of the visual acquisition system (4) adopts a global shutter CMOS sensor with a frame rate of not less than 120fps and a resolution of not less than 1920×1080. It is connected to the central processing unit through a gigabit Ethernet interface, and the camera is installed on a servo-driven adjustable gimbal to achieve binocular stereo vision measurement. The visual prompting system adopts a high-lumen laser projector or RGB LED array, supports dynamic pattern projection, and can be used in conjunction with the ball storage and launch system to achieve dynamic interference training.

4. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The central processing unit (6) adopts a high-performance embedded industrial control computer with ARM or x86 architecture, equipped with the ROS2 real-time operating system, and integrates multiple industrial bus interfaces such as EtherCAT, CAN, SPI, and I2C. It is configured to perform the following functions: Adaptive return path learning and virtual opponent simulation: Based on players' historical shot data or cloud-based professional player match data, a deep reinforcement learning tactical simulation model is used to generate return strategies targeting players' weaknesses. Intelligent training plan generation and dynamic adjustment: Automatically generates long-term training plans based on user training goals, and dynamically adjusts the progress and content based on real-time performance; Biomechanical posture analysis and real-time correction: Calculate player movement parameters based on 3D motion capture data, compare with professional player biomechanical benchmark models, and output improvement suggestions; Create a dynamic digital skills profile for each user, generate radar charts, growth curves and periodic training reports through data visualization, and support report sharing.

5. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The central processing unit (6) is also configured to perform the following adaptive adjustment functions: Adaptive environmental compensation: Based on the analysis of environmental images from the visual acquisition system, light intensity, wind speed and direction are dynamically adjusted to adjust the serve strategy parameters and the display parameters of the visual cue system; Emotion Recognition and Motivation: The emotion recognition module uses facial expression capture from the visual acquisition system or voice data from the human-computer interaction module to identify the trainee's emotional state. When fatigue or frustration is detected, the training pace is automatically adjusted, the difficulty is reduced, or encouraging prompts are played. Multi-ball mixing and dynamic interference training: Control the ball storage and serving system and the robotic arm to launch multiple types of balls in a predetermined or random sequence, while controlling the visual cue system to project moving light spot patterns, requiring players to avoid the light spots and complete the return of the ball.

6. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The health and safety monitoring module (9) includes a millimeter-wave radar or infrared thermal imaging unit operating in the 60GHz band with a sampling rate of not less than 10Hz, used for non-contact monitoring of the trainee's heart rate and respiratory rate; the central processing unit (6) acquires physiological data through the SPI interface, and when it detects that the physiological indicators exceed the safety threshold, it sends an emergency stop command through the CAN bus, automatically suspends training, issues an audible and visual warning and suggests rest.

7. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The robot's mechanical structure and electrical interfaces adopt a modular design, with standardized physical expansion slots and electrical interface boards reserved at the rear of the frame. The interface boards provide USB-C, RJ45, 12V / 24V DC power outputs and CAN bus interfaces for connecting detachable functional modules such as ball and net tension monitoring modules, sound field analysis microphone arrays, or professional biomechanical sensors.

8. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The cloud communication unit (8) supports dual-mode communication of 5G and Wi-Fi 6 and has a built-in MQTT protocol stack; it supports multi-machine linkage and remote battle functions, realizes real-time data synchronization and collaborative training of two or more robots, supports remote players to conduct real-time remote pickle battles through their respective robots, and records and analyzes the competition data; the human-computer interaction module includes a touch screen, a voice input / output unit and a wireless communication unit, which can receive remote control commands.

9. The intelligent training robot for pickleball based on AI and cloud collaboration according to claim 1, characterized in that, The solar auxiliary power supply module (11) includes a flexible photovoltaic panel with a peak power of not less than 200W, a lithium iron phosphate energy storage unit, and an MPPT maximum power point tracking charging controller. The central processing unit (6) is configured with a low-power mode based on dynamic voltage and frequency adjustment, which automatically reduces the system frequency and shuts down unnecessary peripherals when in standby or performing simple tasks. The autonomous mobile chassis (10) adopts a four-wheel omnidirectional mobile structure and integrates a lidar, a visual SLAM module, and a ball recognition module. The central processing unit (6) plans the path based on the ROS navigation stack and controls the autonomous mobile chassis to cruise and pick up balls. After analyzing the surface state of the pick balls, the ball recognition module classifies them into usable balls and scrapped balls.

10. An interactive training method for the intelligent training robot based on any one of claims 1-9, characterized in that, Includes the following steps: Step 1, Visual Acquisition: The real-time video stream of the training scene is acquired through the visual acquisition system (4), and the key points of the player's body, the pickle ball and the player's facial expressions are identified and tracked. Step 2: Serving Strategy Generation: Based on visual recognition results, user-preset training objectives, and cloud-based opponent data, generate serving control commands that include ball type, landing point, ball speed, and spin instructions. Step 3: Execution of the serve: The serve control command is sent to the driver of the ball storage system (2) and the multi-degree-of-freedom robotic arm (3) through the real-time motion control bus, so as to drive them to perform the serve action in coordination. Step 4: Adaptive Adjustment: This includes emotion recognition and adaptive motivation, health and safety monitoring, and environmental adaptive compensation. It involves real-time analysis of player emotions and physiological indicators, environmental parameters, and dynamic adjustment of training intensity, serving parameters, and visual cue parameters. Step 5 Feedback and Updates: Provide players with biomechanical posture improvement suggestions through the human-computer interaction module, update users' digital skill profiles in real time, generate data reports after training or matches, and support sharing.

Images