An all-automatic moxibustion robot based on artificial intelligence and a control method thereof

Abstract

This invention provides a fully automated moxibustion robot based on artificial intelligence and its control method, relating to the field of moxibustion therapy equipment technology. The robot includes a housing, a lifting platform housed within the housing, a robotic arm mounted on the lifting platform, an end effector located at the free end of the robotic arm, a main control board, and a replaceable moxibustion box. The robotic arm has a retracted state and an extended working state; the lifting platform is used to move the robotic arm into or out of the housing when switching between the two states. The end effector is equipped with an electromagnet and an image acquisition module. The main control board carries an artificial intelligence model used to control the robotic arm to move to the target acupoint based on image information. The moxibustion box is detachably attached to the electromagnet and contains a built-in moxibustion powder and ignition mechanism. This invention achieves fully automated, unattended operation of moxibustion therapy, with advantages such as miniaturization, low cost, and localized data processing, making it suitable for home healthcare scenarios.

Description

Technical Field

[0001] This invention relates to the field of moxibustion therapy equipment technology, and in particular to a fully automatic moxibustion robot based on artificial intelligence and its control method. Background Technology

[0002] In recent years, edge AI technology has developed rapidly, with AI chips possessing edge computing power constantly emerging, enabling AI models to run independently on local devices, offering significant advantages in cost, power efficiency, and size. Simultaneously, large language models, visual language models, automatic speech recognition, text-to-speech, and human pose estimation models have undergone continuous iteration. Combined with the development of edge computing power, tasks that previously required cloud computing can now be completed offline locally. The robotics industry has also developed rapidly. The concept of embodied intelligence has accelerated the integration of robots and AI, and numerous open-source robotic arm solutions from global developers have significantly lowered the industry's entry barriers. The field of embedded electronics is also maturing, with mature technologies such as electronic cigarette lighters, electromagnets, and wide-angle distortion-free cameras providing technical references for the realization of automated moxibustion devices.

[0003] Currently, commercially available moxibustion robots primarily rely on human intervention for semi-automatic operation, often employing robotic arms used in factories or ports. Their workflow typically involves the following: the moxibustion therapist determines the acupoints and duration of moxibustion based on the client's needs. After the client adopts a fixed posture, the therapist drags the robotic arm to position the moxibustion head at the acupoint and records the location. Once all acupoints are positioned, the moxibustion head is manually ignited, and the robot performs moxibustion at the recorded position and time. In this type of solution, the robot essentially acts only as a recorder of position and time, with some advanced models possessing the ability to vibrate or move slightly around the recording point. Existing solutions suffer from drawbacks such as bulky robotic arms, high costs, and the need for professional operators, placing significant demands on personnel, space, and costs, making them difficult to promote in the home market. Summary of the Invention

[0004] In view of the above-mentioned technical problems in related technologies, the present invention proposes a fully automatic moxibustion robot based on artificial intelligence and its control method, which can overcome the above-mentioned shortcomings of the prior art.

[0005] To achieve the above-mentioned technical objectives, the technical solution of the present invention is implemented as follows: A fully automated moxibustion robot based on artificial intelligence; This fully automated moxibustion robot based on artificial intelligence includes: The box contains a storage space. A lifting platform is installed inside the housing; A robotic arm is fixedly installed on the lifting platform. The robotic arm has a retracted state and an extended working state. The lifting platform is used to move the robotic arm into or out of the housing when switching between the retracted state and the extended working state. An end effector is disposed at the free end of the robotic arm, and the end effector is equipped with an electromagnet and an image acquisition module; The main control board is located inside the housing and is electrically connected to the lifting platform, the robotic arm and the end effector. The main control board is equipped with an artificial intelligence model, which is used to control the robotic arm to move to the target position based on the image information acquired by the image acquisition module. The replaceable moxibustion box is detachably attached to the electromagnet. The moxibustion box contains a moxibustion medicine column and an ignition mechanism. The end effector is provided with an electrical contact point for triggering the ignition mechanism.

[0006] Furthermore, the housing is equipped with a touch screen, which is electrically connected to the main control board; the housing is also equipped with a new moxibustion box storage compartment and a waste moxibustion box recycling compartment; the main control board is an embedded processor with an NPU that can locally deploy LLM / VLM, and the main control board is connected to and controls the lifting platform, the robotic arm and the end effector through various buses / interfaces.

[0007] Furthermore, the robotic arm includes a base motor, a shoulder motor, an elbow motor, a wrist motor, an end effector motor, and a motor integrated control board. The motor integrated control board is electrically connected to the main control board and each of the motors. The base motor drives the robotic arm to rotate horizontally, the shoulder motor, the elbow motor, and the wrist motor drive the robotic arm to rotate vertically, and the end effector motor drives the end effector to rotate horizontally. The main control board sends control commands to the motor integrated control board via a bus / interface, and the motor integrated control board drives the corresponding motors to move according to the control commands.

[0008] Furthermore, the end effector is also equipped with a laser rangefinder and a temperature sensor. The laser rangefinder is used to detect the distance between the moxibustion box and the human skin, and the temperature sensor is located near the ventilation port of the moxibustion box to detect the internal temperature of the moxibustion box. The image acquisition module is a wide-angle distortion-free camera used to acquire images of the user's tongue coating and the user's posture on the bed. The end effector is also equipped with a microphone, and a speaker is located on the side of the housing. The microphone and the speaker are electrically connected to the main control board to acquire user voice commands and broadcast voice prompts.

[0009] Furthermore, the top of the replaceable moxibustion box is provided with a metal patch for adsorption to the electromagnet, and the outside of the moxibustion box is provided with two electrical contact points, which are electrically connected to the ignition mechanism. The end effector is provided with two corresponding electrical connection contact points, and when the moxibustion box is adsorbed onto the end effector, the electrical contact points abut against the electrical connection contact points.

[0010] According to another aspect of the present invention, a control method for a fully automated moxibustion robot based on artificial intelligence is provided; The control method for this AI-based fully automated moxibustion robot includes the following steps: Step 1: After the device is powered on and completes self-test initialization, it enters the work-ready state. Step 2: Initiate voice interaction by collecting the user's voice input about discomfort symptoms through the microphone, converting it into text information through the speech recognition model, and capturing an image of the user's tongue coating through the camera. Step 3: Input the text information and the tongue image into the local visual language model. The visual language model outputs the moxibustion acupoint plan and duration suggestion, and confirms it with the user through the display screen and voice broadcast. Step 4: After the user confirms the plan, control the robotic arm to move to a top-down position, capture the user's lying posture image through the camera, identify the user's skeletal points through the human posture estimation model, and calculate the spatial coordinates of the target acupoints according to the rules of acupoint positioning in traditional Chinese medicine. Step 5: Solve the inverse kinematics of the robotic arm based on the spatial coordinates of the target acupoint, calculate the rotation angle of each joint of the robotic arm, and control the robotic arm to move the moxibustion box attached to the end effector to the target acupoint for moxibustion. Step six: After the moxibustion is completed or when the user's termination command is received, control the robotic arm to return to the standby position.

[0011] Furthermore, in step five, when solving the inverse kinematics of the robotic arm, the end effector maintains a vertically downward posture, converts the three-dimensional spatial coordinates into a polar coordinate system, and uses trigonometric functions and the law of cosines to solve the angles of each joint in the plane where each joint of the robotic arm is located. Let the coordinates of the target point T in the robotic arm coordinate system be (Xt, Yt, Zt). The base motor J1, shoulder motor J2, elbow motor J3, and wrist motor J4 of the robotic arm are arranged sequentially. The end effector is located at the end of the wrist motor J4. Let T′ be the projection point of the target point T onto the XY plane. Then: θ1 = Arctangnet(Yt / Xt), where θ1 is the target rotation angle of the base motor J1; R= Where R is the distance from the projection point T′ to the origin of the coordinate system; In the plane defined by the Z-axis and OT′, let the distance between the shoulder motor J2 and the elbow motor J3 be the upper arm with a length of L_OA, the distance between the elbow motor J3 and the wrist motor J4 be the forearm with a length of L_AB, and the distance between the wrist motor J4 and the end effector be L_BT. Then: The coordinates B of the wrist motor J4 in the plane are (R, Zt+L_BT). Connect the origin O and point B. In right triangle OBT′, the length of OB can be obtained by the Pythagorean theorem, and ∠BOT′ and ∠OBT′ can be obtained by trigonometric functions. In triangle OAB, the lengths of the three sides OA, AB, and OB are known. Use the Law of Cosines to find ∠AOB, ∠OAB, and ∠ABO respectively. The target angle θ2 of the shoulder motor J2 is ∠BOT′+∠AOB; The target angle θ3 of the elbow motor J3 is ∠OAB; The target angle θ4 of the wrist motor J4 is ∠ABO + ∠OBT′.

[0012] Furthermore, in step four, after the human posture estimation model identifies the user's skeletal points, it determines the human body midline by connecting the midpoint of the shoulder and the midpoint of the hip. Combining the relative positional relationship between acupoints and skeletal points in traditional Chinese medicine theory, it calculates the pixel coordinates of the target acupoints in the image, and then converts the pixel coordinates into spatial coordinates in the robotic arm coordinate system through camera calibration parameters. The camera calibration parameters are obtained from multiple calibration points set in the calibration plane at the factory. These calibration points cover the camera's field of view. The pixel coordinates of each calibration point in the image correspond one-to-one with its spatial coordinates in the robotic arm coordinate system. After calibration, any image point in the camera's field of view can be converted into spatial coordinates in the robotic arm coordinate system through interpolation or coordinate transformation.

[0013] Furthermore, each artificial intelligence model is deployed and runs locally on the main control board. The visual language model and the speech recognition model are accelerated by the NPU, while the text-to-speech model and the human pose estimation model are run by the CPU. Each artificial intelligence model is managed by an independent process, and each process provides services to the outside world through a local socket. The main control software calls the corresponding model service according to the current working stage. The visual language model and the speech recognition model share the NPU computing power in a time-sharing manner.

[0014] Furthermore, it also includes: The moxibustion box replacement process is as follows: when the temperature sensor detects that the temperature of the moxibustion box is lower than the preset threshold, the robotic arm is controlled to move above the waste moxibustion box recycling bin, the power supply of the electromagnet is disconnected to release the current moxibustion box, and then the robotic arm is controlled to move to the new moxibustion box storage bin, the electromagnet is connected to attract the new moxibustion box, and the moxibustion process continues. In addition, there is a power failure protection step. When the external power supply is interrupted, the battery module sends a power failure signal to the main control board and switches to battery power supply mode. The main control board controls the robotic arm to retract to the storage posture and controls the lifting platform to lower the robotic arm into the box.

[0015] The beneficial effects of this invention are as follows: By integrating the robotic arm, lifting platform, main control board, and replaceable moxibustion box into a single housing, the entire device is miniaturized. The lifting platform enables automatic switching between working and storage states for the robotic arm, significantly reducing the space occupied when the device is not in use, thus making it suitable for the home market. By incorporating artificial intelligence models such as visual language models, speech recognition models, and human posture estimation models on the main control board and running them locally, fully automated unattended moxibustion operation without network connectivity is achieved, ensuring user data and privacy security while reducing maintenance costs in the commercial market. By limiting the robotic arm's working posture to two types—vertical downward and horizontal—and simplifying the inverse kinematics solution algorithm, the control complexity and hardware cost of the robotic arm are reduced. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the overall structure of the fully automated moxibustion robot based on artificial intelligence described in this invention; Figure 2 This is a schematic diagram of the external structure of the end effector of the fully automated moxibustion robot based on artificial intelligence described in this invention; Figure 3 This is a schematic diagram of the internal structure of the end effector of the fully automated moxibustion robot based on artificial intelligence described in this invention; Figure 4 This is a schematic diagram illustrating the working relationship between the AI-based fully automated moxibustion robot described in this invention, the bed, and the human body. Figure 5 This is a schematic diagram showing different working states of the fully automated moxibustion robot based on artificial intelligence described in this invention; Figure 6 This is a schematic diagram of the working scenario of the fully automated moxibustion robot based on artificial intelligence described in this invention. Figure 7 This is an overall flowchart of the control method for the fully automated moxibustion robot based on artificial intelligence as described in this invention; Figure 8 This is a schematic diagram of the state acquisition process of the control method for the fully automatic moxibustion robot based on artificial intelligence described in this invention; Figure 9 This is a schematic diagram of the status confirmation process of the control method for the fully automatic moxibustion robot based on artificial intelligence as described in this invention; Figure 10 This is a schematic diagram of the moxibustion implementation process of the control method for the fully automatic moxibustion robot based on artificial intelligence described in this invention; Figure 11 This is a schematic diagram of the inverse kinematics of the robotic arm in the control method of the fully automated moxibustion robot based on artificial intelligence described in this invention.

[0018] In the diagram: 101, housing; 102, robotic arm base; 104, multi-DOF robotic arm; 106, end effector; 107, replaceable moxibustion box; 108, interactive display screen; 109, new moxibustion box storage compartment; 110, waste moxibustion box recycling compartment; 501, robotic arm connection part; 502, end effector body; 503, moxibustion box mounting port; 504, filter module; 505, vision acquisition module; 506, distance detection sensor; 507, replaceable moxibustion box; 601, end effector housing; 602, replaceable filter element; 603, electrical contact point; 604, ventilation module; 605, clamping mechanism; 606, moxa stick limiting post; 607, moxa stick; 608, ignition mechanism. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0021] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0022] like Figures 1 to 6 As shown in the embodiment of the present invention, a fully automated moxibustion robot based on artificial intelligence includes: The box contains a storage space. A lifting platform is installed inside the housing; A robotic arm is fixedly installed on the lifting platform. The robotic arm has a retracted state and an extended working state. The lifting platform is used to move the robotic arm into or out of the housing when switching between the retracted state and the extended working state. An end effector is disposed at the free end of the robotic arm, and the end effector is equipped with an electromagnet and an image acquisition module; The main control board is located inside the housing and is electrically connected to the lifting platform, the robotic arm and the end effector. The main control board is equipped with an artificial intelligence model, which is used to control the robotic arm to move to the target position based on the image information acquired by the image acquisition module. The replaceable moxibustion box is detachably attached to the electromagnet. The moxibustion box contains a moxibustion medicine column and an ignition mechanism. The end effector is provided with an electrical contact point for triggering the ignition mechanism.

[0023] According to an embodiment of the present invention, a fully automated moxibustion robot based on artificial intelligence is provided. In a specific embodiment, the housing is equipped with a touch screen, which is electrically connected to the main control board. The housing is also equipped with a new moxibustion box storage compartment and a waste moxibustion box recycling compartment. The main control board is an embedded processor with an NPU that can locally deploy LLM / VLM. The main control board is connected to and controls the lifting platform, the robotic arm, and the end effector through various buses / interfaces.

[0024] According to an embodiment of the present invention, a fully automated moxibustion robot based on artificial intelligence is described. In a specific embodiment, the robotic arm includes a base motor, a shoulder motor, an elbow motor, a wrist motor, an end effector motor, and a motor integrated control board. The motor integrated control board is electrically connected to the main control board and each of the motors. The base motor drives the robotic arm to rotate horizontally, the shoulder motor, the elbow motor, and the wrist motor drive the robotic arm to rotate vertically, and the end effector motor drives the end effector to rotate horizontally. The main control board sends control commands to the motor integrated control board via a bus / interface, and the motor integrated control board drives the corresponding motors to move according to the control commands.

[0025] According to an embodiment of the present invention, a fully automatic moxibustion robot based on artificial intelligence is provided. In a specific embodiment, the end effector is further equipped with a laser rangefinder and a temperature sensor. The laser rangefinder is used to detect the distance between the moxibustion box and the human skin. The temperature sensor is located near the ventilation port of the moxibustion box and is used to detect the internal temperature of the moxibustion box. The image acquisition module is a wide-angle distortion-free camera, used to acquire images of the user's tongue coating and the user's posture on the bed. The end effector is also equipped with a microphone, and a speaker is provided on the side of the box. The microphone and the speaker are electrically connected to the main control board, respectively, for acquiring user voice commands and broadcasting voice prompts.

[0026] According to an embodiment of the present invention, a fully automatic moxibustion robot based on artificial intelligence is provided. In a specific embodiment, the top of the replaceable moxibustion box is provided with a metal patch for adsorption to the electromagnet. The outside of the moxibustion box is provided with two electrical contact points, which are electrically connected to the ignition mechanism. The end effector is provided with two corresponding electrical connection contact points. When the moxibustion box is adsorbed onto the end effector, the electrical contact points abut against the electrical connection contact points.

[0027] Secondly, such as Figures 7 to 11 As shown in the figure, a control method for a fully automated moxibustion robot based on artificial intelligence according to an embodiment of the present invention includes the following steps: Step 1: After the device is powered on and completes self-test initialization, it enters the work-ready state. Step 2: Initiate voice interaction by collecting the user's voice input about discomfort symptoms through the microphone, converting it into text information through the speech recognition model, and capturing an image of the user's tongue coating through the camera. Step 3: Input the text information and the tongue image into the local visual language model. The visual language model outputs the moxibustion acupoint plan and duration suggestion, and confirms it with the user through the display screen and voice broadcast. Step 4: After the user confirms the plan, control the robotic arm to move to a top-down position, capture the user's lying posture image through the camera, identify the user's skeletal points through the human posture estimation model, and calculate the spatial coordinates of the target acupoints according to the rules of acupoint positioning in traditional Chinese medicine. Step 5: Solve the inverse kinematics of the robotic arm based on the spatial coordinates of the target acupoint, calculate the rotation angle of each joint of the robotic arm, and control the robotic arm to move the moxibustion box attached to the end effector to the target acupoint for moxibustion. Step six: After the moxibustion is completed or when the user's termination command is received, control the robotic arm to return to the standby position.

[0028] According to an embodiment of the present invention, a control method for a fully automatic moxibustion robot based on artificial intelligence is described. In a specific embodiment, in step five, when solving the inverse kinematics of the robotic arm, the end effector maintains a vertically downward posture, converts the three-dimensional spatial coordinates into a polar coordinate system, and solves the angles of each joint in the plane where each joint of the robotic arm is located using trigonometric functions and the law of cosines. Let the coordinates of the target point T in the robotic arm coordinate system be (Xt, Yt, Zt). The base motor J1, shoulder motor J2, elbow motor J3, and wrist motor J4 of the robotic arm are arranged sequentially. The end effector is located at the end of the wrist motor J4. Let T′ be the projection point of the target point T onto the XY plane. Then: θ1 = Arctangnet(Yt / Xt), where θ1 is the target rotation angle of the base motor J1; R= Where R is the distance from the projection point T′ to the origin of the coordinate system; In the plane defined by the Z-axis and OT′, let the distance between the shoulder motor J2 and the elbow motor J3 be the upper arm with a length of L_OA, the distance between the elbow motor J3 and the wrist motor J4 be the forearm with a length of L_AB, and the distance between the wrist motor J4 and the end effector be L_BT. Then: The coordinates B of the wrist motor J4 in the plane are (R, Zt+L_BT). Connect the origin O and point B. In right triangle OBT′, the length of OB can be obtained by the Pythagorean theorem, and ∠BOT′ and ∠OBT′ can be obtained by trigonometric functions. In triangle OAB, the lengths of the three sides OA, AB, and OB are known. Use the Law of Cosines to find ∠AOB, ∠OAB, and ∠ABO respectively. The target angle θ2 of the shoulder motor J2 is ∠BOT′+∠AOB; The target angle θ3 of the elbow motor J3 is ∠OAB; The target angle θ4 of the wrist motor J4 is ∠ABO + ∠OBT′.

[0029] According to an embodiment of the present invention, a control method for a fully automatic moxibustion robot based on artificial intelligence is described. In a specific embodiment, in step four, after the human posture estimation model identifies the user's skeletal points, the midline of the human body is determined by connecting the midpoint of the shoulder and the midpoint of the hip. Combining the relative positional relationship between acupoints and skeletal points in traditional Chinese medicine theory, the pixel coordinates of the target acupoint in the image are calculated. Then, the pixel coordinates are converted into spatial coordinates in the robotic arm coordinate system through camera calibration parameters. The camera calibration parameters are obtained from multiple calibration points set in the calibration plane at the factory. These calibration points cover the camera's field of view. The pixel coordinates of each calibration point in the image correspond one-to-one with its spatial coordinates in the robotic arm coordinate system. After calibration, any image point in the camera's field of view can be converted into spatial coordinates in the robotic arm coordinate system through interpolation or coordinate transformation.

[0030] According to an embodiment of the present invention, a control method for a fully automatic moxibustion robot based on artificial intelligence is provided. In a specific implementation, each artificial intelligence model is deployed and runs locally on the main control board. The visual language model and the speech recognition model are accelerated by the NPU, while the text-to-speech model and the human posture estimation model are run by the CPU. Each artificial intelligence model is managed by an independent process, and each process provides services to the outside world through a local socket. The main control software calls the corresponding model service according to the current working stage. The visual language model and the speech recognition model share the NPU computing power in a time-sharing manner.

[0031] According to an embodiment of the present invention, a control method for a fully automated moxibustion robot based on artificial intelligence, in a specific embodiment, further includes: The moxibustion box replacement process is as follows: when the temperature sensor detects that the temperature of the moxibustion box is lower than the preset threshold, the robotic arm is controlled to move above the waste moxibustion box recycling bin, the power supply of the electromagnet is disconnected to release the current moxibustion box, and then the robotic arm is controlled to move to the new moxibustion box storage bin, the electromagnet is connected to attract the new moxibustion box, and the moxibustion process continues. In addition, there is a power failure protection step. When the external power supply is interrupted, the battery module sends a power failure signal to the main control board and switches to battery power supply mode. The main control board controls the robotic arm to retract to the storage posture and controls the lifting platform to lower the robotic arm into the box.

[0032] To facilitate understanding of the above technical solutions of the present invention, the following detailed description of the above technical solutions of the present invention is provided through specific embodiments and working principles.

[0033] Example 1: Overall structure of the equipment The AI-based fully automated moxibustion robot includes a housing, a robotic arm base, a multi-degree-of-freedom robotic arm, an end effector, replaceable moxibustion boxes, an interactive display screen, a new moxibustion box storage compartment, and a waste moxibustion box recycling compartment.

[0034] The housing forms the main structure of the robot, with its dimensions optimized for miniaturization. The interior of the housing houses the main control board, lifting platform, multi-degree-of-freedom robotic arm, as well as storage compartments for new and discarded moxibustion boxes. The housing has a lid, allowing all components except the external power supply to be stored away when not in use. The housing features a power switch and a touchscreen for user interaction and control. Two moxibustion box storage slots are also provided: one for storing unused moxibustion boxes and the other for recycling used moxibustion boxes. A speaker grille is located on the side of the housing for external sound output. By integrating all functional modules into a compact housing structure and utilizing an openable lid and lifting mechanism for the robotic arm's deployment and storage, the device occupies minimal space when not in operation, making it easy to store and move in the home. In operation, the lifting platform raises the robotic arm outside the housing, and the multi-degree-of-freedom robotic arm unfolds to cover the preset moxibustion work area.

[0035] Example 2: Power supply and battery module The external power supply uses an AC-DC power module. When using it, the user can connect the 220V input terminal of the power supply to the mains power and the DC output terminal to the power interface of the enclosure to power the whole machine.

[0036] After the battery module is connected to an external power source, it automatically charges the battery and powers the robot while the external power source is available. When the external power source fails, the battery module sends an external power failure signal to the main control board and switches to battery power mode to power the robot. At this time, the robot stops working, and the main control board controls the multi-degree-of-freedom robotic arm to retract to a safe position to prevent it from falling due to power failure. By setting the battery module as a backup power source and automatically switching to it when the external power source fails, combined with the power failure protection control logic of the main control board, the multi-degree-of-freedom robotic arm will not fall out of control in the event of an abnormal power failure, ensuring both user safety and equipment safety.

[0037] Example 3: Lifting platform The lifting platform integrates a control module to hold the multi-degree-of-freedom robotic arm, and uses a threaded rod motor to achieve lifting. When the equipment is needed, it is powered on, and the lifting platform is controlled to raise the multi-degree-of-freedom robotic arm to the outside of the housing; when not in use, the multi-degree-of-freedom robotic arm first switches to a retractable position, and then the lifting platform lowers to retract the multi-degree-of-freedom robotic arm into the housing. The main control board is connected to the lifting platform via a bus / interface to control the lifting platform's movement / stop and direction of movement. The lifting platform also integrates a laser distance detection sensor, which reports distance information via the bus / interface to assist in detecting the current position of the lifting platform.

[0038] The lifting platform serves as the connecting mechanism between the multi-degree-of-freedom robotic arm and the housing. It achieves vertical lifting of the robotic arm via the forward and reverse rotation of a threaded rod motor. Combined with the robotic arm's own posture switching, this enables the orderly transition of the entire machine between its retracted, raised, and deployed working states. A laser rangefinder provides real-time feedback on the lifting platform's position, ensuring precise control and safe limit of the lifting motion.

[0039] Example 4: Main control board The main control board uses an embedded processor with an NPU that allows for local deployment of LLM / VLM. The main control board connects to and controls the lifting platform, front-end control board, camera in the vision acquisition module, and peripheral devices such as the multi-degree-of-freedom robotic arm, microphone, and speaker via various buses / interfaces. The main control board is the core of the entire robot's control system, running the robot's main control software and various application programs.

[0040] The processor has a built-in NPU (Neural Processing Unit), which provides accelerated inference capabilities for local visual language models, speech recognition models, etc., enabling artificial intelligence models to run offline on the edge without connecting to a cloud server. The main control board uniformly schedules various peripheral devices through multiple interfaces, realizing full-process automation from voice acquisition, image recognition, solution decision-making to robotic arm motion control.

[0041] Example 5: Multi-degree-of-freedom robotic arm The multi-degree-of-freedom robotic arm includes modules such as a motor integrated control board, motor, front-end I / O board, electromagnet, sensor, camera, microphone, and robotic arm structure.

[0042] The motor integrated control board provides power supply and control interface for each motor. The main control board sends and receives control commands through the bus / interface, and the motor integrated control board converts the commands into control instructions and sends them to each motor.

[0043] The motors utilize closed-loop servo motors with a unified power supply and unified control of five motors. The five motors are: J1 base motor, J2 shoulder motor, J3 elbow motor, J4 wrist motor, and J5 terminal motor. The base motor is responsible for horizontal rotation, providing rotational functionality for the entire robotic arm. The shoulder, elbow, and wrist motors all rotate vertically, working together to provide a two-bar linkage function, allowing the moxibustion head terminal to remain vertically downward and enabling localized movement based on the wrist. The terminal motor provides rotational functionality, enabling horizontal rotation of the moxibustion head while maintaining a vertically downward orientation. Depending on the load, two types of motors from the same series are used: the base, wrist, and terminal motors use smaller load motors for lighter loads, while the shoulder and elbow motors use larger load motors for heavier loads.

[0044] The front-end I / O board is used to centrally control various functions of the moxibustion robot terminal, including the electromagnet switch, ignition device switch, distance sensor data acquisition, and temperature sensor data acquisition. The front-end I / O board adopts a microcontroller plus MOS control scheme. The microcontroller runs the front-end I / O control program, connects to the main control board through the bus interface, operates the MOS switch according to the commands of the main control board, and reports sensor information.

[0045] The multi-degree-of-freedom robotic arm integrates an electromagnet at its front end. The MOSFET on the front-end I / O board controls whether the electromagnet is energized. When the moxibustion box needs to be attracted, the MOSFET is energized, and the moxibustion box is attracted to the front end of the robotic arm; when it needs to be released, the power is turned off, and the moxibustion box falls off by gravity.

[0046] The multi-degree-of-freedom robotic arm also integrates a temperature sensor and a laser rangefinder at its front end. The temperature sensor is placed near the ventilation port of the moxibustion box to determine whether the moxibustion has been properly ignited by detecting the gas temperature at the ventilation port; the laser rangefinder is placed on the side of the terminal, with the measuring direction perpendicular to the contact surface of the moxibustion box, to assist in measuring the distance between the contact surface of the moxibustion box and the human body.

[0047] The camera, serving as the visual acquisition module, is deployed at the front end of the multi-degree-of-freedom robotic arm, oriented perpendicular to the contact surface of the moxibustion box and parallel to the direction of the laser rangefinder. The camera is a wide-angle, distortion-free model used to acquire images of the user's tongue coating, the user's posture on the bed, and to provide auxiliary positioning for the robotic arm's movement.

[0048] A stereo microphone is used, which is placed at the front end of the multi-degree-of-freedom robotic arm to collect user voice and realize voice input function.

[0049] The multi-degree-of-freedom robotic arm structure includes various structural components used to connect the aforementioned motor, robotic arm housing, and front-end structure.

[0050] Through the coordinated movement of five joint motors, the end effector can be flexibly positioned in three-dimensional space. The base motor provides horizontal rotational freedom, while the shoulder, elbow, and wrist motors form a two-bar linkage in the vertical plane, enabling the end effector to reach any position within the workspace while maintaining a vertically downward orientation. The terminal motor provides end rotational freedom, allowing the moxibustion head to rotate around a vertical axis to adjust its orientation. All motors are uniformly controlled via a serial communication network, simplifying the wiring structure. Front-end sensors provide real-time feedback of temperature, distance, and image information, providing data support for closed-loop control.

[0051] Example 6: Moxibustion box The moxibustion box is designed for recycling and reuse. The unused box contains an unlit moxibustion stick and a heating wire for ignition. The heating wire extends to the outer end of the box, forming two electrical contact points that connect to the I / O plate at the front of the robotic arm to form an ignition device. The top of the moxibustion box has a metal patch for attraction to an electromagnet. When the moxibustion box is connected to the front of the robotic arm via the electromagnet, the side electrical contact points also connect to the electrical contact points at the front of the robotic arm, ensuring a unique connection method through guide grooves and position locks.

[0052] The moxibustion box features a modular, replaceable design, enabling quick installation and removal via electromagnet attraction. Guide slots and position locks ensure accurate positioning and reliable electrical connections for each installation. The heating wire acts as the ignition element, generating heat to ignite the moxibustion sticks upon power-up, eliminating the need for manual ignition. After use, the moxibustion box is automatically transported to a recycling bin by a robotic arm, achieving fully automated replacement of the moxibustion heads.

[0053] Example 7: Equipment power on / off and status switching When the equipment is powered off, the robotic arm is located inside the housing. When external power is available and the power button is pressed, the equipment powers on, performs a self-test, and initializes. The robotic arm is then raised outside the housing by the lifting platform. Once initialization is complete and the device enters the ready-to-use state, a notification will appear on the touchscreen indicating that it is ready for use, and a voice prompt will also be given to the user.

[0054] When the equipment is in an abnormal state such as after an abnormal power outage and subsequent power restoration, the equipment will automatically return to the ready-to-work state.

[0055] When the device receives a shutdown command from the user via the touchscreen, it will initiate the shutdown process, retract the robotic arm into the housing, and finally display a message on the screen indicating that the device has been shut down. The user can then disconnect the power.

[0056] Through preset power-on self-test, initialization, status recovery, and power-off recovery processes, the equipment can start and stop safely and reliably under different operating conditions. In particular, it can automatically return to the ready state after abnormal power outage recovery, thus improving the ease of use and robustness of the equipment.

[0057] Example 8: Main Workflow After powering on, the device enters a ready-to-use state, waiting for the user to press the start button on the touchscreen before entering the status acquisition process.

[0058] The status acquisition process first collects user information through voice interaction, then acquires an image of the user's tongue coating, and finally proceeds to the solution confirmation process. Specifically: First, the text-to-speech service is activated, broadcasting a welcome message; the robotic arm turns towards the user, using the YOLO model to identify whether the user is on the left or right side of the bed, and aligns the robotic arm with the user; the prompt message is broadcast: "Hello, where do you feel unwell? Turn on the microphone and wait for the user's voice input, which is then converted into text by the speech recognition model; the message continues: "Now we will take a picture of the tongue coating. Please open your mouth as wide as possible." The camera is adjusted to face the user's head and recognizes the mouth opening posture to take the picture; the input symptoms are broadcast for user confirmation, and after confirmation, the solution confirmation process begins.

[0059] In the treatment plan confirmation process, the collected symptom text and tongue coating images are input into the local visual language model. The model analyzes the symptoms and tongue coating, and outputs the analyzed and recommended acupuncture points according to the prompts, ensuring that all output acupuncture points are those that the software can recognize and locate. The model outputs prompt text, which is then converted into speech by a text-to-speech model and broadcast to the user, including health analysis and the moxibustion plan, followed by precautions for moxibustion. Finally, the user confirms the moxibustion plan via voice or touchscreen; if approved, the moxibustion treatment process begins.

[0060] During the moxibustion process, if the robotic arm fails to grasp the moxibustion box, it first retrieves a new one from the storage compartment. Based on the required lying or prone posture for the current acupoint, the robotic arm is hovered in a panoramic overhead view position, and a voice prompt instructs the user to adopt the appropriate posture and not to move unnecessarily. After confirming the user's posture is correct via camera, a human posture model is run to obtain the positions of the main skeletal points. Based on the skeletal points and traditional Chinese medicine theory, the spatial coordinates of the target acupoint are calculated. The pixel coordinates of the acupoint in the image are converted to spatial coordinates in the robotic arm's coordinate system using camera calibration parameters. Subsequently, inverse kinematics of the robotic arm are solved to calculate the rotation angles of each joint, controlling the robotic arm to move the moxibustion box to the target acupoint for moxibustion. During the moxibustion process, a temperature sensor monitors the temperature of the moxibustion box; if the temperature is too low, the moxibustion box is replaced. After the countdown for the first acupoint ends, the next acupoint is treated. After all acupoints are treated, the system returns and provides a voice prompt to the user.

[0061] The entire workflow forms a complete closed loop of consultation, syndrome differentiation, and moxibustion application. Voice interaction and tongue image acquisition simulate the inquiry and observation techniques of traditional Chinese medicine's diagnostic methods; the visual language model comprehensively analyzes multimodal inputs, simulating the syndrome differentiation and treatment process of a TCM practitioner; the human posture estimation model, combined with TCM acupoint location rules, achieves automatic acupoint identification and location; and the robotic arm, controlled by an inverse kinematics algorithm, precisely executes the moxibustion operation. The entire process requires no human intervention, achieving fully automated, unattended operation. All AI models run locally on the device, with data not connected to the internet, ensuring user privacy and security.

[0062] Example 9: Acupoint Location and Coordinate Conversion Let's take the Guanyuan acupoint as an example to illustrate the acupoint location method. A human posture model can be used to calculate 33 skeletal points. Connecting the midpoints of the left and right shoulder positions (numbers 11 and 12) and the left and right hip positions (numbers 23 and 24) yields the human midline. In Traditional Chinese Medicine theory, the total length of the chest and abdomen is 9 inches plus 8 inches plus 5 inches, equaling 22 chest inches. The line connecting the shoulder positions (numbers 11 and 12) is approximately 4 / 9 of the upper 9 inches. The remaining distance is 5 inches plus 8 inches, equaling 18 inches. The navel is located 5 inches below, at a relative position of 13 / 18. Using the remaining 3 / 5 of this distance, the location of the Guanyuan acupoint can be determined. The Shenque and Qihai acupoints, located near the navel, can be located using the same method.

[0063] Once the coordinates in the image are obtained, OpenCV coordinate transformation is used to convert the acupoint location relative to the origin of the robotic arm's coordinate system, based on the factory calibration of the fixed camera posture. Calibration involves specifying the planar position and camera posture during equipment manufacturing, placing black and white grid points in the correct positions, allowing image acquisition and coordinate recording. The calibration covers a large area of ​​the camera's field of view, with 105 calibration points (7 rows x 15 columns). After calibration, any image point on this horizontal plane can be converted to its actual X and Y coordinates. Once the acupoint is identified and its X and Y coordinates are obtained, it can be moved accordingly.

[0064] By utilizing the skeletal coordinates provided by a human pose estimation model and combining them with the relative positional relationship between acupoints and skeletal landmarks in Traditional Chinese Medicine (TCM) theory, the pixel coordinates of acupoints are calculated in image space. Then, through pre-completed camera calibration—establishing a mapping relationship between image pixel coordinates and the spatial coordinates of the robotic arm base—the pixel coordinates of the acupoints are converted into spatial coordinates executable by the robotic arm. This composite positioning method, combining skeletal point localization, TCM theory conversion, and camera calibration mapping, achieves automatic acupoint recognition and localization without the need for external positioning aids.

[0065] Example 10: Solving the inverse kinematics of a robotic arm. The inverse kinematics solution process for the robotic arm is as follows: Define the coordinates of the target point T in the robot arm coordinate system as (Xt, Yt, Zt). The base motor J1, shoulder motor J2, elbow motor J3, and wrist motor J4 of the robot arm are set in sequence, and the end effector is located at the end of the wrist motor J4.

[0066] Let T′ be the projection of the target point T onto the XY plane. Then: θ1 = Arctangnet(Yt / Xt), where θ1 is the target rotation angle of the base motor J1; R= , where R is the distance from the projection point T′ to the origin of the coordinate system.

[0067] In the plane defined by the Z-axis and OT′, let the distance between shoulder motor J2 and elbow motor J3 be the upper arm with length L_OA, the distance between elbow motor J3 and wrist motor J4 be the forearm with length L_AB, and the distance between wrist motor J4 and the end effector be L_BT. Then the coordinates B of wrist motor J4 in this plane are (R, Zt + L_BT). Connect the origin O and point B. In right triangle OBT′, the length of OB is obtained using the Pythagorean theorem, and ∠BOT′ and ∠OBT′ are obtained using trigonometric functions. In triangle OAB, the lengths of the three sides OA, AB, and OB are known. ∠AOB, ∠OAB, and ∠ABO are obtained using the law of cosines. Then: The target angle θ2 of the shoulder motor J2 is ∠BOT + ∠AOB; The target angle θ3 of the elbow motor J3 is ∠OAB; The target angle θ4 of the wrist motor J4 is ∠ABO + ∠OBT′.

[0068] The inverse kinematics solution is now complete, and the angles of the four joints are converted and sent to the respective motors for execution.

[0069] Because the working posture of the robotic arm in this invention is limited to two modes—either the end effector maintaining a vertically downward moxibustion posture or a horizontal posture for picking up and placing moxibustion boxes—the inverse kinematics problem is significantly simplified. By reducing the three-dimensional spatial problem to a planar geometric problem, the angles of each joint can be analytically solved using the Pythagorean theorem and the law of cosines, without iterative calculations. This simplified algorithm significantly reduces the computational burden on the main control board, improves the response speed and control accuracy of the robotic arm, and simultaneously reduces the hardware requirements for the motor and control system.

[0070] Example 11: Local AI computing power scheduling This product uses four artificial intelligence-based model algorithms: a visual language model combining a large language model and image recognition, a speech recognition model, a text-to-speech model, and a human pose estimation model.

[0071] Because of the limited computing resources on the edge, which operate locally, it is impossible to run all models simultaneously. Therefore, a dynamic service-oriented implementation is adopted. Each algorithm is managed by a fixed process, and each process provides services to the outside world via a local socket. The algorithm service can be stopped or started according to external commands.

[0072] Based on the algorithm characteristics and edge computing power, the computing power allocation is as follows: the visual language model and speech recognition model are accelerated using the NPU, while the text-to-speech model and human pose estimation model are run using the CPU. Different NPU services are activated according to the current state of the device, with the visual language model and speech recognition model sharing NPU computing power in a time-sharing manner, but not simultaneously. Throughout the different stages of the entire workflow, only the NPU and CPU work in parallel; there is no situation where different algorithms are executed in parallel on the NPU. The visual language model is implemented using the open-source VLM model with fine-tuning, while the speech recognition model, human pose estimation model, and text-to-speech model directly use the open-source model.

[0073] By service-oriented architecture for four AI models and dynamically scheduling them on demand, efficient collaborative operation of multiple models is achieved under limited edge computing power. The visual language model and speech recognition model time-sharing multiplex the NPU to avoid resource contention; the text-to-speech and pose estimation models run on the CPU, executing in parallel with NPU tasks, thus improving overall processing efficiency. This scheduling strategy enables all AI functions to run smoothly on the edge chip alone, without relying on cloud computing power.

[0074] In summary, by utilizing the technical solutions described above, the entire device is miniaturized by integrating the robotic arm, lifting platform, main control board, and replaceable moxibustion box into a single enclosure. The lifting platform enables automatic switching between the robotic arm's working and stowed states, significantly reducing the space occupied when not in use, thus making it suitable for the home market. Furthermore, by integrating artificial intelligence models such as visual language models, speech recognition models, and human posture estimation models onto the main control board and running them locally, fully automated unattended moxibustion operation without network connectivity is achieved, ensuring user data and privacy security while reducing maintenance costs in the commercial market. Finally, by limiting the robotic arm's working posture to two modes—vertical downward and horizontal—and simplifying the inverse kinematics solution algorithm, the control complexity and hardware cost of the robotic arm are reduced.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fully automated moxibustion robot based on artificial intelligence, characterized in that, include: The box contains a storage space. A lifting platform is installed inside the housing; A robotic arm is fixedly installed on the lifting platform. The robotic arm has a retracted state and an extended working state. The lifting platform is used to move the robotic arm into or out of the housing when switching between the retracted state and the extended working state. An end effector is disposed at the free end of the robotic arm, and the end effector is equipped with an electromagnet and an image acquisition module; The main control board is located inside the housing and is electrically connected to the lifting platform, the robotic arm and the end effector. The main control board is equipped with an artificial intelligence model, which is used to control the robotic arm to move to the target position based on the image information acquired by the image acquisition module. The replaceable moxibustion box is detachably attached to the electromagnet. The moxibustion box contains a moxibustion medicine column and an ignition mechanism. The end effector is provided with an electrical contact point for triggering the ignition mechanism.

2. The fully automated moxibustion robot based on artificial intelligence according to claim 1, characterized in that, The housing is equipped with a touch screen, which is electrically connected to the main control board. The housing also has a storage compartment for new moxibustion boxes and a recycling compartment for discarded moxibustion boxes. The main control board is an embedded processor with an NPU that can locally deploy LLM / VLM. The main control board is connected to and controls the lifting platform, the robotic arm, and the end effector through various buses / interfaces.

3. The fully automated moxibustion robot based on artificial intelligence according to claim 1, characterized in that, The robotic arm includes a base motor, a shoulder motor, an elbow motor, a wrist motor, an end effector motor, and a motor integrated control board. The motor integrated control board is electrically connected to the main control board and each of the motors. The base motor drives the robotic arm to rotate horizontally, the shoulder motor, the elbow motor, and the wrist motor drive the robotic arm to rotate vertically, and the end effector motor drives the end effector to rotate horizontally. The main control board sends control commands to the motor integrated control board via a bus / interface, and the motor integrated control board drives the corresponding motors to move according to the control commands.

4. The fully automated moxibustion robot based on artificial intelligence according to claim 1, characterized in that, The end effector is also equipped with a laser rangefinder and a temperature sensor. The laser rangefinder is used to detect the distance between the moxibustion box and the human skin, and the temperature sensor is located near the ventilation port of the moxibustion box to detect the internal temperature of the moxibustion box. The image acquisition module is a wide-angle distortion-free camera used to acquire images of the user's tongue coating and the user's posture on the bed. The end effector is also equipped with a microphone, and a speaker is located on the side of the box. The microphone and the speaker are electrically connected to the main control board to acquire user voice commands and broadcast voice prompts.

5. The fully automated moxibustion robot based on artificial intelligence according to claim 1, characterized in that, The replaceable moxibustion box has a metal patch on the top for adsorption to the electromagnet. The moxibustion box has two electrical contact points on the outside, which are electrically connected to the ignition mechanism. The end effector has two corresponding electrical connection contact points. When the moxibustion box is adsorbed onto the end effector, the electrical contact points abut against the electrical connection contact points.

6. A control method for a fully automated moxibustion robot based on artificial intelligence, applied to the fully automated moxibustion robot based on artificial intelligence as described in any one of claims 1 to 5, characterized in that, Includes the following steps: Step 1: After the device is powered on and completes self-test initialization, it enters the work-ready state. Step 2: Initiate voice interaction by collecting the user's voice input about discomfort symptoms through the microphone, converting it into text information through the speech recognition model, and capturing an image of the user's tongue coating through the camera. Step 3: Input the text information and the tongue image into the local visual language model. The visual language model outputs the moxibustion acupoint plan and duration suggestion, and confirms it with the user through the display screen and voice broadcast. Step 4: After the user confirms the plan, control the robotic arm to move to a top-down position, capture the user's lying posture image through the camera, identify the user's skeletal points through the human posture estimation model, and calculate the spatial coordinates of the target acupoints according to the rules of acupoint positioning in traditional Chinese medicine. Step 5: Solve the inverse kinematics of the robotic arm based on the spatial coordinates of the target acupoint, calculate the rotation angle of each joint of the robotic arm, and control the robotic arm to move the moxibustion box attached to the end effector to the target acupoint for moxibustion. Step six: After the moxibustion is completed or when the user's termination command is received, control the robotic arm to return to the standby position.

7. The control method for the fully automated moxibustion robot based on artificial intelligence according to claim 6, characterized in that, In step five, when solving the inverse kinematics of the robotic arm, the end effector maintains a vertically downward posture, converts the three-dimensional spatial coordinates into a polar coordinate system, and uses trigonometric functions and the law of cosines to solve the angles of each joint in the plane where each joint of the robotic arm is located. Let the coordinates of the target point T in the robotic arm coordinate system be (Xt, Yt, Zt). The base motor J1, shoulder motor J2, elbow motor J3, and wrist motor J4 of the robotic arm are arranged sequentially. The end effector is located at the end of the wrist motor J4. Let T′ be the projection point of the target point T onto the XY plane. Then: θ1 = Arctangnet(Yt / Xt), where θ1 is the target rotation angle of the base motor J1; R= Where R is the distance from the projection point T′ to the origin of the coordinate system; In the plane defined by the Z-axis and OT′, let the distance between the shoulder motor J2 and the elbow motor J3 be the upper arm with a length of L_OA, the distance between the elbow motor J3 and the wrist motor J4 be the forearm with a length of L_AB, and the distance between the wrist motor J4 and the end effector be L_BT. Then: The coordinates B of the wrist motor J4 in the plane are (R, Zt+L_BT). Connect the origin O and point B. In right triangle OBT′, the length of OB can be obtained by the Pythagorean theorem, and ∠BOT′ and ∠OBT′ can be obtained by trigonometric functions. In triangle OAB, the lengths of the three sides OA, AB, and OB are known. Use the Law of Cosines to find ∠AOB, ∠OAB, and ∠ABO respectively. The target angle θ2 of the shoulder motor J2 is ∠BOT′+∠AOB; The target angle θ3 of the elbow motor J3 is ∠OAB; The target angle θ4 of the wrist motor J4 is ∠ABO + ∠OBT′.

8. The control method for the fully automated moxibustion robot based on artificial intelligence according to claim 6, characterized in that, In step four, after the human posture estimation model identifies the user's skeletal points, it determines the human midline by connecting the midpoint of the shoulder and the midpoint of the hip. Combining the relative positional relationship between acupoints and skeletal points in traditional Chinese medicine theory, it calculates the pixel coordinates of the target acupoints in the image, and then converts the pixel coordinates into spatial coordinates in the robotic arm coordinate system through camera calibration parameters. The camera calibration parameters are obtained from multiple calibration points set in the calibration plane at the factory. These calibration points cover the camera's field of view. The pixel coordinates of each calibration point in the image correspond one-to-one with its spatial coordinates in the robotic arm coordinate system. After calibration, any image point in the camera's field of view can be converted into spatial coordinates in the robotic arm coordinate system through interpolation or coordinate transformation.

9. The control method for a fully automated moxibustion robot based on artificial intelligence according to claim 6, characterized in that, Each AI model is deployed and runs locally on the main control board. The visual language model and the speech recognition model are accelerated by the NPU, while the text-to-speech model and the human pose estimation model are run by the CPU. Each AI model is managed by an independent process, and each process provides services to the outside world through a local socket. The main control software calls the corresponding model service according to the current working stage. The visual language model and the speech recognition model share the NPU computing power in a time-sharing manner.

10. The control method for a fully automated moxibustion robot based on artificial intelligence according to claim 6, characterized in that, Also includes: The moxibustion box replacement process is as follows: when the temperature sensor detects that the temperature of the moxibustion box is lower than the preset threshold, the robotic arm is controlled to move above the waste moxibustion box recycling bin, the power supply of the electromagnet is disconnected to release the current moxibustion box, and then the robotic arm is controlled to move to the new moxibustion box storage bin, the electromagnet is connected to attract the new moxibustion box, and the moxibustion process continues. In addition, there is a power failure protection step. When the external power supply is interrupted, the battery module sends a power failure signal to the main control board and switches to battery power supply mode. The main control board controls the robotic arm to retract to the storage posture and controls the lifting platform to lower the robotic arm into the box.

Images