Dynamic scene self-adaptive data acquisition mechanical arm system of intelligent robot with body

Abstract

The invention belongs to the technical field of mechanical arms, and particularly relates to a self-adaptive data acquisition mechanical arm system for a dynamic scene of an intelligent robot with a body, which comprises mechanical arm sections, and a plurality of mechanical arm section arrays are spliced to form a flexible mechanical arm; the driving mechanism is arranged at the head end of the flexible mechanical arm and used for driving the flexible mechanical arm to operate; each mechanical arm section is composed of a joint framework, a variable-rigidity flexible joint, a frame connecting piece and a flexible touch skin, the variable-rigidity flexible joints achieve synchronous rotation by means of the flexible bending deformation characteristic of the variable-rigidity flexible joints when the single mechanical arm sections are bent, and the rigidity of bending deformation of the variable-rigidity flexible joints is variable. The flexible touch skin is used for monitoring the contact force borne by the flexible mechanical arm. The flexible mechanical arm matched with the magnetic current damper has the rigidity variable function, is suitable for various working environments, and is matched with multi-source sensing equipment to achieve dynamic obstacle avoidance and road stiffness re-planning in a dynamic scene.

Description

Technical Field

[0001] This invention belongs to the field of robotic arm technology, specifically relating to a robotic arm system for dynamic scene adaptive data acquisition of embodied intelligent robots. Background Technology

[0002] Embossed intelligent robots are artificial intelligence systems with physical entities. They interact with the environment through multimodal sensors and integrate technologies such as machine vision and natural language understanding to achieve closed-loop control of perception, decision-making, and execution. Robotic arms are complex systems with high precision, multiple inputs and outputs, high nonlinearity, and strong coupling. Due to their unique operational flexibility, they have been widely used in industrial assembly, safety and explosion protection, and other fields. Robotic arms are complex systems with uncertainties such as parameter perturbations, external interference, and unmodeled dynamics.

[0003] Problems with existing technology: Existing flexible robotic arms are designed for use in unstructured environments or in scenarios involving human contact with high safety requirements. However, their flexible structure is not suitable for other operational scenarios that require a certain degree of force and precision. Their application scenarios are limited and their functions are limited. The robotic arm itself does not have the function of variable stiffness. Existing robotic arms suffer from severe responsiveness issues in dynamic environments. They are prone to loss of control in dynamic scenarios such as moving obstacles, have poor environmental adaptability, and lack the ability to dynamically avoid obstacles and replan routes. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic scene adaptive data acquisition robotic arm system for embodied intelligent robots. The flexible robotic arm equipped with a magnetohydrodynamic damper has variable stiffness and is suitable for various working environments. When combined with multi-source sensing devices, it can achieve dynamic obstacle avoidance and path replanning in dynamic scenes.

[0005] The specific technical solution adopted by this invention is as follows: An embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system includes: A robotic arm segment, wherein several robotic arm segments are arrayed and spliced ​​together to form a flexible robotic arm; A drive mechanism is provided at the head end of the flexible robotic arm and is used to drive the flexible robotic arm to operate. The drive mechanism is internally provided with a winding power mechanism and a rotation power mechanism. An actuator is mounted at the end of a flexible robotic arm; One of the robotic arm segments consists of a joint skeleton, a variable stiffness flexible joint, a frame connector, and a flexible tactile skin. The frame connector works with a winding power mechanism to achieve bending of the flexible robotic arm. The variable stiffness flexible joint achieves synchronous rotation by relying on its own flexible bending deformation characteristics when a single robotic arm segment bends, and the stiffness of the bending deformation of the variable stiffness flexible joint is variable. The flexible tactile skin is used to monitor the contact force borne by the flexible robotic arm.

[0006] Each of the robotic arm segments has a joint skeleton distributed at both ends. The upper and lower sides of the center of the joint skeleton are integrally provided with bone tubes. A protective mesh is provided between the bone tubes on the opposite side of two adjacent joint skeletons to protect the inner variable stiffness flexible joint. Millimeter radar is embedded at the end of each support rod of the joint skeleton.

[0007] Each of the variable stiffness flexible joints has an inner rotating tube symmetrically arranged at both ends. One end of the inner rotating tube is rotatably assembled inside the corresponding bone tube. The inner rotating tubes distributed on both sides of a joint skeleton are connected in an overlapping transmission manner. The edges of two inner rotating tubes facing each other are connected in an array with end connecting pieces. The ends of the end connecting pieces in the same group are connected to a ring connecting piece. A middle connecting piece is distributed in a ring between two adjacent ring connecting pieces, and the middle connecting pieces are connected in an alternating manner.

[0008] The inner rotating tube has an inner cavity at one end opposite to the adjacent bone tube. The inner cavity contains a magnetic fluid tube and a coil, with the coil sleeved on the outer surface of the middle part of the magnetic fluid tube. The open end of the inner rotating tube is fitted with a ring cap for fixing the magnetic fluid tube and the coil. A through tube is integrally provided at the central axis of the magnetic fluid tube. Magnetic fluid channels are arranged in a ring array through the middle part of the magnetic fluid tube and at the edge of the through tube.

[0009] Liquid plug 1 and liquid plug 2 are slidably assembled inside both ends of the magnetofluid tube, and both liquid plug 1 and liquid plug 2 are slidably sleeved on the outer surface of the tube. Two liquid plugs 1 inside a variable stiffness flexible joint are distributed inside the two magnetofluid tubes facing each other. Liquid plug tube rods are integrally connected to the two liquid plugs 1 facing each other. Braided corrugated hoses are connected between the ends of two adjacent liquid plug tube rods. The coils are connected in series by wires.

[0010] A transmission component is connected to one end of the inner rotating tube located at the head end of the flexible robotic arm. The transmission component is connected to the rotation power mechanism. One end of the inner rotating tube located at the tail end of the flexible robotic arm is connected to the execution end component. Each support rod at the end of the joint skeleton located at the end of the flexible robotic arm is connected to a traction rope. The traction rope passes through all the other joint skeletons inside the flexible robotic arm and is connected to the winding power mechanism. The top of the actuator is connected to a steel wire rope for driving the actuator to perform clamping operations. The steel wire rope passes through the flexible robotic arm and is connected to the winding power mechanism. The steel wire rope movably passes through the tube, the liquid plug rod, the braided corrugated hose, the joint skeleton, and the transmission component. The winding power mechanism consists of several sets of servo motors and reducers, and is used for wire drive control of the wire rope and traction rope.

[0011] The frame connector is connected to each support end of two adjacent joint skeletons. The frame connector includes a rope locking plate. Both ends of the frame connector are connected to spring ropes, and the ends of the spring ropes are connected to the ends of the corresponding joint skeleton support rods. An extender is fixedly installed on the inner side of the rope locking plate, and the extender output end of the extender is connected to a toothed block 1 that moves through the rope locking plate. A toothed block 2 is slidably assembled on the inner side of the toothed block 1. Side plates are integrally connected to both sides of the toothed block 2. The side plates slide through both sides of the toothed block 1 and are fixedly connected to a stop rod at their ends. An oblique protrusion is integrally provided on the side wall of the rope locking plate away from the toothed block 1, and the stop rod and the oblique protrusion form a pressing contact. The traction rope passes through the gap between the toothed block 1 and the toothed block 2.

[0012] The flexible tactile skin includes a ring shell and wing plates. The ring shell is sleeved on the outer surface of the corresponding bone canal. The wing plates are installed in a ring array on the outer wall of the joint skeleton and are located within the intervals of each support rod of the joint skeleton. Pressure sensors are symmetrically fixedly installed on both sides of the inner wall of the ring shell. A sleeve is movably sleeved on the outer side of the ring shell. Press blocks are symmetrically fixedly installed on both sides of the inner wall of the sleeve. A spring is connected to the end of the press block near the adjacent pressure sensor. Each of the wing plates is hinged to both ends with an elastic telescopic tube, and the ends of each set of elastic telescopic tubes are movably connected to a protective plate. The inner wall of one end of each protective plate is connected to the outer wall of the casing by a connecting rod.

[0013] The specific steps of the method for adaptive operation of embodied intelligent robots in dynamic scenes are as follows: Multi-source perception and dynamic modeling stage: Compound eye camera clusters scan environmental geometry information, multispectral sensors simultaneously capture material properties, flexible tactile skin monitors contact force in real time, acoustic arrays predict object slippage trends through friction acoustic patterns, solid-state millimeter radar and IMU are fused to generate environmental point clouds, and temperature, humidity / air pressure data are combined to correct airflow disturbance errors and update obstacle movement trajectories; visual, tactile, and acoustic data are fused at the feature level to generate dynamic scene maps, and real-time data cleaning, noise removal, and labeling are performed; Intelligent decision-making and collaborative control stage: pre-simulate disturbance scenario strategies, generate task sequences, and use RBF neural networks to compensate for system uncertainties; super-torsional sliding mode controllers suppress vibrations, and variable stiffness flexible joints switch stiffness to absorb sudden impacts; depth cameras generate obstacle bounding boxes, replan trajectories based on artificial potential field methods, and provide collision warnings simultaneously. Flexible execution and structural self-adaptation stage: The magnetorheological fluid damping drive joint, composed of components of the magnetorheological fluid tube, realizes dynamic adjustment of stiffness; the execution end component performs electromagnetic adsorption and vacuum gripping operations; Data closed loop and system evolution stage.

[0014] A dynamic scene adaptive operation system, comprising: An environmental perception and dynamic modeling system is provided with a multimodal perception fusion module, which integrates a vision system, tactile feedback and environmental dynamic modeling. Vision system: A multi-camera cluster combined with multispectral sensors enables multi-scale scene scanning and material recognition; Tactile and acoustic feedback: Flexible tactile skin monitors contact force; Acoustic microphone array predicts object slippage trends through friction sound patterns; Environmental dynamic modeling: Solid-state millimeter radar and IMU construct a SLAM++ system to predict obstacle movement trajectories in real time, and integrate temperature, humidity / barrier sensors to perceive environmental disturbances. The intelligent decision-making and adaptive control system has a hierarchical control architecture and obstacle avoidance and fault tolerance mechanisms. Layered control architecture: Millisecond-level stiffness adjustment of variable stiffness flexible joints to cope with sudden impacts; online parameter identification algorithm updates inertia / friction coefficient model in real time, combined with RBF neural network to dynamically compensate for system uncertainties; spatiotemporal graph convolutional network analyzes the correlation of object motion, and reinforcement learning pre-plays strategies for thousand-level interference scenarios; Obstacle avoidance and fault tolerance mechanisms: The depth camera generates bounding boxes for human bodies / obstacles, and the trajectory is replanned in real time based on the artificial potential field method to ensure a safe distance; The flexible execution and structural innovation system features an internal biomimetic drive structure, consisting of magnetorheological fluid damping drive joints made up of magnetorheological fluid tube components, enabling dynamic stiffness adjustment; modular execution end components support electromagnetic adsorption and vacuum gripping operations. The data closed-loop and simulation optimization system is equipped with edge-cloud collaborative computing and simulation transfer training.

[0015] The technical effects achieved by this invention are as follows: (1) In this invention, a magnetohydrodynamic damper assembled from components such as magnetohydrodynamic tubes is combined with a mechanical arm segment composed of flexible joints, and a flexible mechanical arm assembled from several mechanical arm segments has a variable stiffness function. When it is in a flexible state, it can keep the arm compliant during operation, while the end effector can also perform rotation and clamping actions. It is suitable for use in non-structural environments or scenarios where there is contact with people and high safety requirements. When it is in a rigid state, it can control the end effector to perform operations that require a certain force and precision. In addition, when there is a momentary switch from a rigid state to a flexible state, it can be used for scenarios involving collision events to weaken the impact of collision events on the equipment.

[0016] (2) The present invention uses protective plates distributed around each joint skeleton to directly bear the contact force of other objects. After detecting the pressure, it is determined that the robotic arm segment is in contact with other objects at the node. Then, in conjunction with other sensing devices, the sensing and detection work is realized, laying the data foundation for subsequent obstacle avoidance path planning.

[0017] (3) This invention achieves millimeter-level contact force perception and material recognition through compound eye camera cluster, multispectral sensor and flexible electronic skin. The SLAM++ system constructed by solid-state lidar and IMU can predict the movement trajectory of obstacles. The depth camera generates human body / obstacle bounding box and replans trajectory in real time based on artificial potential field method to maintain a safe distance of 0.45m. The spatiotemporal Transformer model provides 0.5-second collision warning and magnetorheological joint locks the impact within 8ms. Attached Figure Description

[0018] Figure 1 This is a front view structural diagram of the robotic arm system provided in an embodiment of the present invention; Figure 2 This is a combined structural diagram of the various robotic arm segments provided in the embodiments of the present invention; Figure 3 This is a structural diagram of the robotic arm segment provided in an embodiment of the present invention; Figure 4 This is a structural diagram of the joint skeleton and frame connector provided in an embodiment of the present invention; Figure 5 This is a structural diagram of a variable stiffness flexible joint provided in an embodiment of the present invention; Figure 6 This is a cross-sectional view of the combination of the bone tube and the variable stiffness flexible joint provided in an embodiment of the present invention; Figure 7 This is a cross-sectional view of a variable stiffness flexible joint provided in an embodiment of the present invention; Figure 8 yes Figure 4 A magnified view of the structure at point A in the middle; Figure 9 This is a structural diagram of the combination of the joint skeleton and the flexible tactile skin provided in the embodiments of the present invention; Figure 10 This is a structural diagram of the flexible tactile skin provided in an embodiment of the present invention; Figure 11 This is a system diagram of the dynamic scene adaptive operation system provided in the embodiments of the present invention; Figure 12 This is a flowchart of the dynamic scene adaptive operation method provided in the embodiments of the present invention.

[0019] The attached diagram lists the components represented by each number as follows: 1. Robotic arm segment; 2. Joint skeleton; 201. Bone tube; 202. Protective mesh fence; 203. Millimeter radar; 3. Variable stiffness flexible joint; 301. Inner rotating tube; 302. End connecting piece; 303. Ring connecting piece; 304. Middle connecting piece; 305. Inner cavity; 306. Magnetorheological fluid tube; 307. Coil; 308. Ring cover; 309. Magnetorheological fluid channel; 310. Through tube; 311. Liquid plug one; 312. Liquid plug two; 313. Liquid plug tube rod; 314. Braided corrugated hose; 315. Wire; 4. Frame connector; 401. Spring rope; 402. Rope locking plate; 403. Telescopic device; 404. Tooth block one; 405. Side plate; 406. Tooth block two; 407. Oblique protrusion; 408. Support rod; 5. Soft tactile skin; 501. Ring shell; 502. Pressure sensor; 503. Press block; 504. Wing plate; 505. Elastic telescopic tube; 506. Protective plate; 507. Connecting rod; 508. Sleeve; 6. Actuating end piece; 601. Steel wire rope; 7. Transmission component; 8. Traction rope; 9. Drive mechanism. Detailed Implementation

[0020] To make the objectives and advantages of this invention clearer, the invention will be specifically described below with reference to embodiments. It should be understood that the following text is merely used to describe one or more specific embodiments of the invention and does not strictly limit the scope of protection specifically claimed by the invention.

[0021] like Figures 1-10 As shown, the embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system includes: robotic arm segment 1, drive mechanism 9 and execution end piece 6; See attached document Figures 1-2 A flexible robotic arm is formed by assembling several robotic arm segments 1 in an array; a drive mechanism 9 is located at the head end of the flexible robotic arm and is used to drive the operation of the flexible robotic arm. The drive mechanism 9 is the body part of the embodied intelligent robot. The drive mechanism 9 is equipped with a winding power mechanism and a rotation power mechanism; an execution end piece 6 is installed at the end of the flexible robotic arm. See attached document Figure 3A robotic arm segment 1 consists of a joint skeleton 2, a variable stiffness flexible joint 3, a frame connector 4, and a flexible tactile skin 5. The frame connector 4 works with a winding power mechanism to achieve bending of the flexible robotic arm. When a single robotic arm segment 1 bends, the variable stiffness flexible joint 3 achieves synchronous rotation by relying on its own flexible bending deformation characteristics. The stiffness of the bending deformation of the variable stiffness flexible joint 3 is variable. The flexible tactile skin 5 is used to monitor the contact force borne by the flexible robotic arm.

[0022] Example 1: See attached document Figure 4 Each robotic arm segment 1 has a joint skeleton 2 distributed at both ends. The joint skeleton 2 has a bone tube 201 integrally installed on both the upper and lower sides of the center. A protective mesh 202 for protecting the inner variable stiffness flexible joint 3 is installed between the bone tubes 201 on the opposite side of two adjacent joint skeletons 2. A millimeter radar 203 is embedded at the end of each support rod of the joint skeleton 2. See attached document Figures 5-7 Each variable stiffness flexible joint 3 has an inner rotating tube 301 symmetrically arranged at both ends. One end of the inner rotating tube 301 is rotatably assembled inside the corresponding bone tube 201. The inner rotating tubes 301 distributed on both sides of a joint skeleton 2 are connected in an overlapping transmission manner. The edges of the two inner rotating tubes 301 facing each other are connected in an array with end connecting pieces 302. The ends of the same set of end connecting pieces 302 are connected together with a ring connecting piece 303. The two adjacent ring connecting pieces 303 are connected in a ring-distributed manner with middle connecting pieces 304, and the middle connecting pieces 304 are connected in an alternating manner. See attached document Figures 5-7 The inner rotating tube 301 has an inner cavity 305 at one end away from the adjacent bone tube 201. The inner cavity 305 is equipped with a magnetic fluid tube 306 and a coil 307. The coil 307 is sleeved on the outer surface of the middle part of the magnetic fluid tube 306. The open end of the inner rotating tube 301 is equipped with a ring cover 308 for fixing the magnetic fluid tube 306 and the coil 307. A through tube 310 is integrally provided at the central axis of the magnetic fluid tube 306. The middle part of the magnetic fluid tube 306 and the edge of the through tube 310 are provided with a magnetic fluid channel 309 in a ring array. See attached document Figures 5-7 The inner ends of the magnetofluid tube 306 are respectively slidably assembled with liquid plug 311 and liquid plug 312, and both liquid plug 311 and liquid plug 312 are slidably sleeved on the outer surface of the tube 310. The two liquid plugs 311 inside a variable stiffness flexible joint 3 are distributed inside the two magnetofluid tubes 306 facing one end. The two liquid plugs 311 are integrally connected to the liquid plug tube rod 313 on the facing side. The ends of two adjacent liquid plug tube rods 313 are connected with a braided corrugated hose 314. The coils 307 are connected in series by wires 315. According to the above, when the robotic arm segment 1 is driven to bend and deform, its inner variable stiffness flexible joint 3 will bend at the braided corrugated hose 314 as the turning point. The angles of the two inner rotating tubes 301 will change accordingly, and the tennis ball-shaped structure formed by the end connecting piece 302, the ring connecting piece 303, and the middle connecting piece 304 will also deflect accordingly. The function of this tennis ball-shaped structure is to ensure that the two inner rotating tubes 301 can be in a synchronous rotation transmission relationship under the premise that the angles of the two inner rotating tubes 301 change. Inside the inner rotating tube 301, the coils 307 are all When not powered on, the two inner tubes 301 are flexibly connected through the tennis ball-shaped structure and the protective mesh 202. In this state, the connection stiffness between the two inner tubes 301 is relatively weak. When the variable stiffness flexible joint 3 bends, due to the bendable but non-stretchable nature of the braided corrugated hose 314, the angle between the two connected liquid plug rods 313 changes synchronously, and the corresponding liquid plug 311 will move in the corresponding magnetic fluid tube 306, so that the magnetic fluid at both ends inside the magnetic fluid tube 306 flows and is replaced through the magnetic fluid channel 309. When it is necessary to increase the connection stiffness to ensure the force of the robotic arm in performing operations, all coils 307 are energized. The magnetic field generated by the energized coils 307 causes the magnetic particles in the magnetic fluid to aggregate and block the magnetic fluid channel 309, thereby hindering the flow of liquid at both ends of the magnetic fluid tube 306. This puts the corresponding liquid plug rod 313 and the magnetic fluid tube 306 in a state similar to a fixed connection, thereby increasing the stiffness of the variable stiffness flexible joint 3. Furthermore, it increases the overall stiffness of the flexible robotic arm, enabling it to perform operations that require a certain structural stiffness. Conversely, if a collision event is detected during the operation, the robotic arm can be restored to a flexible state by momentarily de-energizing the coils 307, thereby absorbing and weakening the collision force. For the specific execution process, please refer to Example 3 below. The above describes a magnetofluid damper assembled from components such as the magnetofluid tube 306, combined with a robotic arm segment 1 composed of flexible joints, and a flexible robotic arm assembled from several robotic arm segments 1. This flexible robotic arm has variable stiffness. When in a flexible state, it can maintain arm compliance during operation while the end effector 6 can also perform rotation and clamping actions, making it suitable for use in non-structural environments or scenarios involving human contact and requiring high safety. When in a rigid state, it can control the end effector 6 to perform operations requiring a certain amount of force and precision. In addition, when switching from a rigid state to a flexible state instantaneously, it can be used in scenarios involving collision events to mitigate the impact of collision events on the equipment.

[0023] The working principle of this invention is as follows: When neither of the coils 307 inside the inner rotating tube 301 is energized, the two inner rotating tubes 301 are flexibly connected through the aforementioned tennis ball-shaped structure and the protective mesh 202. In this state, the connection stiffness between the two inner rotating tubes 301 is relatively weak. When the variable stiffness flexible joint 3 bends, due to the bendable but non-stretchable characteristic of the braided corrugated hose 314, the angle between the two connected liquid plug rods 313 changes synchronously, and the corresponding liquid plug 311 will move in the corresponding magnetic fluid tube 306, causing the magnetic fluid to move. The magnetic fluid at both ends of the fluid tube 306 flows and is replaced through the magnetic fluid channel 309. When it is necessary to increase the stiffness of the connection to ensure the force of the robotic arm to perform the operation, all coils 307 are energized. The magnetic field generated by the energized coils 307 causes the magnetic particles in the magnetic fluid to aggregate and block the magnetic fluid channel 309, thereby hindering the flow of liquid at both ends of the magnetic fluid tube 306. This puts the corresponding liquid plug rod 313 and the magnetic fluid tube 306 in a state similar to a fixed connection, thereby improving the stiffness of the variable stiffness flexible joint 3.

[0024] See attached document Figures 1-2 One end of the inner rotating tube 301 at the head of the flexible robotic arm is connected to a transmission component 7, which is connected to a rotation power mechanism. One end of the inner rotating tube 301 at the tail of the flexible robotic arm is connected to an execution end component 6. The top of the execution end component 6 is connected to a wire rope 601 for driving the execution end component 6 to perform a clamping operation. The wire rope 601 passes through the flexible robotic arm and is connected to a winding power mechanism. The wire rope 601 movably passes through the through tube 310, the liquid plug rod 313, the braided corrugated hose 314, the joint skeleton 2, and the transmission component 7. See attached document Figures 1-2 , Figure 7 Each support rod of the joint frame 2 located at the end of the flexible robotic arm is connected to a traction rope 8. The traction rope 8 runs through all the other joint frames 2 inside the flexible robotic arm and is connected to the winding power mechanism. The winding power mechanism is composed of several sets of servo motors and reducers, and is used to perform wire drive control on the wire rope 601 and the traction rope 8.

[0025] Based on the above, the inner rotating tubes 301 distributed on both sides of a joint skeleton 2 (i.e., the two inner rotating tubes 301 that connect two adjacent variable stiffness flexible joints 3) are connected in an overlapping transmission manner, so that all variable stiffness flexible joints 3 can rotate synchronously. The inner rotating tube 301 located at the head end of the flexible robotic arm is connected to the rotation power mechanism through the transmission component 7. The rotation power mechanism consists of a servo motor and a reducer. When the power mechanism is in operation, it can drive all variable stiffness flexible joints 3 to rotate together. The execution end piece 6 located at the end of the flexible robotic arm also rotates accordingly. One of the servo motors and reducers in the winding power mechanism can control the execution end piece 6 to perform the gripping operation by winding and unwinding the steel wire rope 601. The specific structure is all existing technology and will not be described in detail here.

[0026] See attached document Figure 8 The frame connector 4 is connected to each support end of two adjacent joint frames 2. The frame connector 4 includes a rope locking plate 402. Both ends of the frame connector 4 are connected to spring ropes 401. The end of the spring rope 401 is connected to the end of the corresponding support rod of the joint frame 2. See attached document Figure 8 An extension device 403 is fixedly installed on the inner side of the rope locking plate 402, and the extension output end of the extension device 403 is connected to a tooth block 404 that moves through the rope locking plate 402. A tooth block 406 is slidably assembled on the inner side of the tooth block 404. Side plates 405 are integrally connected to both sides of the tooth block 406. The side plates 405 slide through both sides of the tooth block 404 and are fixedly connected to the end of each side plate 405. An oblique protrusion 407 is integrally provided on the side wall of the rope locking plate 402 away from the tooth block 404, and the abutment 408 and the oblique protrusion 407 form a pressing contact. The traction rope 8 passes through the gap between the tooth block 404 and the tooth block 406.

[0027] According to the above, when it is necessary to control the bending of a certain section of the robotic arm 1, the telescopic device 403 in the corresponding direction inside the robotic arm 1 is activated according to the bending direction. The telescopic control tooth block 404 and tooth block 406 move down together. During the downward movement, due to the pressure contact between the abutment rod 408 and the inclined protrusion 407, the tooth block 406 moves closer to the tooth block 404 while moving down, until it cooperates with the tooth block 404 to clamp and bite the traction rope 8 on that side. Then, the corresponding set of servo motors and reducers in the winding power mechanism is activated to make it work and wind up the traction rope 8 on that side, so that bending can be achieved with the robotic arm 1 as the turning point.

[0028] Example 2: See attached document Figures 9-10The flexible tactile skin 5 includes a ring shell 501 and a wing plate 504. The ring shell 501 is sleeved on the outer surface of the corresponding bone tube 201. The wing plate 504 is installed in a ring array on the outer wall of the joint skeleton 2, and the wing plate 504 is located in the interval of each support rod of the joint skeleton 2. Pressure sensors 502 are symmetrically fixed on both sides of the inner wall of the ring shell 501. A sleeve 508 is movably sleeved on the outer side of the ring shell 501. Press blocks 503 are symmetrically fixed on both sides of the inner wall of the sleeve 508. A spring is connected to one end of the press block 503 near the adjacent pressure sensor 502. See attached document Figures 9-10 Each wing plate 504 has an elastic telescopic tube 505 hinged to both ends, and each set of elastic telescopic tubes 505 has a protective plate 506 movably connected to the end of the tube. The inner wall of one end of each protective plate 506 is connected to the outer wall of the housing 508 through a connecting rod 507.

[0029] Based on the above, the millimeter radar 203 distributed at the ends of each support rod of each joint frame 2 is used to detect the position of surrounding objects in real time. The guard plate 506 distributed around each joint frame 2 is used to directly bear the contact force of other objects. When any guard plate 506 comes into contact with other objects, it will shift, the corresponding elastic telescopic tube 505 on the inner side will contract, and the corresponding connecting rod 507 will synchronously drive the housing 508 to rotate. The housing 508 and the ring housing 501 will rotate relative to each other. The corresponding button 503 on a certain side will move closer to the adjacent pressure sensor 502 and squeeze the pressure sensor 502 with the help of the spring. After detecting the pressure, the pressure sensor 502 determines that the robotic arm segment 1 has come into contact with other objects at this node. In conjunction with other sensing devices, the sensing and detection work is realized, laying the data foundation for subsequent obstacle avoidance path planning.

[0030] The working principle of this invention is as follows: the guard plates 506 distributed around each joint skeleton 2 are used to directly bear the contact force of other objects. After any guard plate 506 on any side comes into contact with other objects, it will shift, the corresponding elastic telescopic tube 505 on the inner side will contract, and the corresponding connecting rod 507 will synchronously drive the sleeve 508 to rotate. The sleeve 508 and the ring shell 501 will rotate relative to each other. The corresponding push block 503 on a certain side will move closer to the adjacent pressure sensor 502 and squeeze the pressure sensor 502 with the help of the spring. After detecting the pressure, the pressure sensor 502 determines that the robotic arm segment 1 has come into contact with other objects at this node.

[0031] Example 3: like Figure 11-12 As shown, the dynamic scene adaptive operation system includes: I. An environmental perception and dynamic modeling system, which is equipped with a multimodal perception fusion module, which integrates a vision system, tactile feedback and environmental dynamic modeling; Vision system: A compound eye multi-camera cluster (wide-angle + macro lens) combined with a multispectral sensor (visible light / infrared / depth) enables multi-scale scene scanning and material recognition. This camera cluster and multispectral sensor are installed inside the embodied intelligent robot. Haptic and acoustic feedback: 5. Flexible tactile skin monitors contact force (±0.05N sensitivity); an acoustic microphone array predicts the sliding trend of objects through frictional sound patterns, and the acoustic microphone array is also installed inside the embodied intelligent robot; Environmental dynamic modeling: A solid-state millimeter radar 203 (100Hz scanning frequency) and an IMU are used to build a SLAM++ system to predict the trajectory of obstacles in real time (error <5mm), and temperature, humidity and air pressure sensors are integrated to sense environmental disturbances.

[0032] II. Intelligent decision-making and adaptive control system, which has a hierarchical control architecture and obstacle avoidance and fault tolerance mechanisms. Layered control architecture: Low-level execution: Super-twisted sliding mode control (ST-SMC) suppresses high-frequency vibrations (bandwidth > 500Hz), combined with millisecond-level stiffness adjustment of the variable stiffness flexible joint 3 to cope with sudden impacts; Mid-level optimization: Online parameter identification algorithm updates the inertia / friction coefficient model in real time, combined with RBF neural network to dynamically compensate for system uncertainties; High-level planning: Spatiotemporal graph convolutional network (ST-GCN) analyzes the correlation of object motion, and reinforcement learning rehearses strategies for thousand-level interference scenarios; Obstacle avoidance and fault tolerance mechanisms: Dynamic obstacle avoidance: Depth camera generates bounding boxes for human / obstacles, and replans the trajectory in real time based on artificial potential field method to ensure safe distance (average 0.45m); Fault recovery: Redundant degrees of freedom reconstruct the kinematic chain, and the task can still be completed under single joint failure; Vibration sensor provides 30-day advance warning of bearing wear.

[0033] III. Flexible execution and structural innovation system, which is equipped with a biomimetic drive structure, a magnetorheological fluid damping drive joint composed of components of magnetorheological fluid tube 306, to achieve dynamic adjustment of stiffness; modular execution end piece 6 supports operations such as electromagnetic adsorption and vacuum gripping.

[0034] IV. Data closed-loop and simulation optimization system, which is internally equipped with edge-cloud collaborative computing and simulation transfer training; Edge-cloud collaborative computing: The robotic arm's edge NPU (10TOPS computing power) handles real-time control, while the cloud optimizes long-term strategies through digital twins; a dynamic scene database is established (100,000+ sample annotations of environment-action-result triples). Simulation transfer training: High-fidelity physics engines (such as RoboMirage) simulate fluid / software interactions, and domain randomization techniques (random textures / lighting) improve the model's generalization ability.

[0035] The specific operation process of this dynamic scene adaptive running system is as follows: Step 1: Multi-source sensing and dynamic modeling stage Step 1.1: Environmental Perception and Data Synchronization Parallel acquisition by heterogeneous sensors: A cluster of compound eye cameras (wide-angle + macro) scans environmental geometry at a frequency of 100Hz, while multispectral sensors simultaneously capture material properties (such as metallic reflectivity and biological tissue humidity). Distributed haptic feedback: The flexible haptic skin 5 monitors the contact force in real time (±0.05N accuracy), the acoustic array predicts the object's slippage trend through friction sound patterns, and the data is aligned to a unified clock source through timestamps; Dynamic SLAM++ construction: Solid-state millimeter radar 203 and IMU are fused to generate environmental point clouds, and temperature, humidity and air pressure data are combined to correct airflow disturbance errors and update obstacle movement trajectories; Step 1.2: Multimodal data fusion A federated learning framework is used to aggregate data across domains: visual, tactile, and acoustic data are fused at the feature level to generate dynamic scene maps, and the ST-GCN network analyzes the motion correlation of objects (such as the chain displacement of stacked objects on a conveyor belt). The edge NPU (10 TOPS computing power) performs real-time data cleaning, removes noise, and labels the "environment-action-result" triple before uploading it to the cloud-based digital twin platform.

[0036] Step 2: Intelligent Decision-Making and Collaborative Control Stage Step 2.1: Hierarchical Decision Generation High-level planning: The cloud-based reinforcement learning model calls upon the historical database (100,000+ samples) to pre-simulate interference scenario strategies (such as strong winds deflecting target objects) and generate task sequences; Mid-level optimization: The online parameter identification algorithm updates the inertia / friction coefficient of the robotic arm joints in real time, and the RBF neural network compensates for system uncertainties; Low-level control: The super-torsional sliding mode controller (ST-SMC) suppresses vibration with a bandwidth of >500Hz, and the variable stiffness flexible joint 3 switches stiffness (0.1-100N·m / rad) within 8ms to absorb sudden impacts; Step 2.2, Real-time Obstacle Avoidance and Fault Tolerance Dynamic obstacle avoidance: The depth camera generates obstacle bounding boxes and replans the trajectory based on the artificial potential field method, maintaining a safe distance of 0.45m; the spatiotemporal Transformer model provides a collision warning of 0.5 seconds; Fault recovery: redundant degrees of freedom reconstruct the kinematic chain (such as enabling backup joints when a single joint fails), and vibration sensors provide a 30-day advance warning of bearing wear.

[0037] Step 3: Flexible Execution and Structural Adaptation Stage Modular actuator scheduling: Actuator 6 performs operations such as electromagnetic adsorption and vacuum gripping.

[0038] Step 4: Data Closure and System Evolution Stage Simulation-Reality Transfer: High-fidelity physics engines (such as RoboMirage) simulate fluid / software interactions, and domain randomization techniques (random textures / lighting) generate adversarial training data; spiking neural networks (SNNs) reduce power consumption by 30%, and the ADMM distributed algorithm shortens planning time by 18.65%. Continuous optimization and security protection: Level 3 protection: pressure-sensitive shell triggers joint magnetorheological locking (8ms response); federated learning framework shares multi-robot anomaly handling experience; self-diagnostic evolution: cloud-based digital twin compares simulation and actual operation data, iteratively updating control parameters; wear sensor data drives spare parts pre-maintenance.

[0039] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system, characterized in that, include: A robotic arm segment (1) is formed by assembling several robotic arm segments (1) into an array; The drive mechanism (9) is located at the head end of the flexible robotic arm and is used to drive the flexible robotic arm to operate. The drive mechanism (9) is provided with a winding power mechanism and a rotation power mechanism inside. An actuator (6) is mounted on the end of a flexible robotic arm; A robotic arm segment (1) consists of a joint skeleton (2), a variable stiffness flexible joint (3), a frame connector (4), and a flexible tactile skin (5). The frame connector (4) works with a winding power mechanism to achieve bending of the flexible robotic arm. The variable stiffness flexible joint (3) rotates synchronously when a single robotic arm segment (1) bends, relying on its own flexible bending deformation characteristics. The stiffness of the bending deformation of the variable stiffness flexible joint (3) is variable. The flexible tactile skin (5) is used to monitor the contact force borne by the flexible robotic arm.

2. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 1, characterized in that: Each of the robotic arm segments (1) has a joint skeleton (2) distributed at both ends. The joint skeleton (2) has a bone tube (201) integrally arranged on both the upper and lower sides of the center. A protective mesh (202) for protecting the inner variable stiffness flexible joint (3) is provided between the bone tubes (201) on the opposite side of two adjacent joint skeletons (2). A millimeter radar (203) is embedded at the end of each support rod of the joint skeleton (2).

3. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 2, characterized in that: Each of the variable stiffness flexible joints (3) is symmetrically provided with an inner rotating tube (301) at both ends. One end of the inner rotating tube (301) is rotatably assembled inside the corresponding bone tube (201). The inner rotating tubes (301) distributed on both sides of a joint skeleton (2) are connected in an overlapping transmission manner. The edges of the two inner rotating tubes (301) facing each other are connected in an array with end connecting pieces (302). The ends of the same set of end connecting pieces (302) are connected together with a ring connecting piece (303). The two adjacent ring connecting pieces (303) are connected in a ring-distributed manner with a middle connecting piece (304), and the middle connecting pieces (304) are connected in an alternating manner.

4. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 3, characterized in that: The inner rotating tube (301) has an inner cavity (305) at one end away from the adjacent bone tube (201). The inner cavity (305) is equipped with a magnetic fluid tube (306) and a coil (307). The coil (307) is sleeved on the outer surface of the middle part of the magnetic fluid tube (306). The open end of the inner rotating tube (301) is equipped with a ring cap (308) for fixing the magnetic fluid tube (306) and the coil (307). A through tube (310) is integrally provided at the central axis of the magnetic fluid tube (306). The middle part of the magnetic fluid tube (306) and the edge of the through tube (310) are provided with a magnetic fluid channel (309) in a ring array.

5. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 4, characterized in that: The magnetic fluid tube (306) has a liquid plug 1 (311) and a liquid plug 2 (312) slidably assembled inside both ends. The liquid plug 1 (311) and the liquid plug 2 (312) are slidably sleeved on the outer surface of the tube (310). The two liquid plugs 1 (311) inside a variable stiffness flexible joint (3) are distributed inside the two magnetic fluid tubes (306) facing each other. The liquid plug tube rod (313) is integrally connected to the side of the two liquid plugs 1 (311) facing each other. The ends of the two adjacent liquid plug tube rods (313) are connected by a braided corrugated hose (314). The coils (307) are connected in series by wires (315).

6. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 5, characterized in that: A transmission component (7) is connected to one end of the inner rotating tube (301) located at the head end of the flexible robotic arm. The transmission component (7) is connected to the rotation power mechanism. One end of the inner rotating tube (301) located at the end of the flexible robotic arm is connected to the execution end component (6). Each support rod of the joint frame (2) located at the end of the flexible robotic arm is connected to a traction rope (8). The traction rope (8) runs through all the other joint frames (2) inside the flexible robotic arm and is connected to the winding power mechanism. The top of the actuator (6) is connected to a wire rope (601) for driving the actuator (6) to perform clamping operations. The wire rope (601) passes through the flexible robotic arm and is connected to the winding power mechanism. The wire rope (601) movably passes through the tube (310), the liquid plug rod (313), the braided corrugated hose (314), the joint skeleton (2), and the transmission component (7). The winding power mechanism consists of several sets of servo motors and reducers, and is used to perform line drive control on the wire rope (601) and the traction rope (8).

7. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 6, characterized in that: The frame connector (4) is connected to each support end of two adjacent joint skeletons (2). The frame connector (4) includes a rope locking plate (402). Both ends of the frame connector (4) are connected to spring ropes (401). The end of the spring rope (401) is connected to the end of the support rod of the corresponding joint skeleton (2). A telescopic device (403) is fixedly installed on the inner side of the rope locking plate (402), and the telescopic output end of the telescopic device (403) is connected to a tooth block one (404) that moves through the rope locking plate (402). A tooth block two (406) is slidably assembled on the inner side of the tooth block one (404). A side plate (405) is integrally connected to both sides of the tooth block two (406). The side plate (405) slides through both sides of the tooth block one (404) and a stop rod (408) is fixedly connected to each end. An oblique protrusion (407) is integrally provided on the side wall of the rope locking plate (402) away from the tooth block one (404), and the stop rod (408) and the oblique protrusion (407) form a squeezing contact. The traction rope (8) passes through the gap between the tooth block one (404) and the tooth block two (406).

8. The embodied intelligent robot dynamic scene adaptive data acquisition robotic arm system according to claim 7, characterized in that: The flexible tactile skin (5) includes a ring shell (501) and a wing plate (504). The ring shell (501) is sleeved on the outer surface of the corresponding bone tube (201). The wing plate (504) is installed in a ring array on the outer wall of the joint skeleton (2), and the wing plate (504) is located in the interval of each support rod of the joint skeleton (2). Pressure sensors (502) are symmetrically fixed on both sides of the inner wall of the ring shell (501). A sleeve (508) is movably sleeved on the outer side of the ring shell (501). Press blocks (503) are symmetrically fixed on both sides of the inner wall of the sleeve (508). A spring is connected to one end of the press block (503) near the adjacent pressure sensor (502). Each of the wing plates (504) is hinged to both ends with an elastic telescopic tube (505), and the ends of each set of elastic telescopic tubes (505) are movably connected to a guard plate (506). The inner wall of one end of each guard plate (506) is connected to the outer wall of the casing (508) through a connecting rod (507).

9. A dynamic scene adaptive operation method for an embodied intelligent robot, used to control the operation of a robotic arm system as described in any one of claims 1-8, characterized in that, The specific steps are as follows: Multi-source perception and dynamic modeling stage: The compound eye camera cluster scans the geometric information of the environment, the multispectral sensor captures the material properties in real time, the flexible tactile skin (5) monitors the contact force in real time, the acoustic array predicts the sliding trend of the object through friction sound pattern, the solid millimeter radar (203) is fused with the IMU to generate environmental point cloud, and the airflow disturbance error is corrected by combining temperature, humidity and air pressure data to update the obstacle movement trajectory; Visual, tactile, and acoustic data are fused at the feature level to generate dynamic scene maps, with real-time data cleaning, noise removal, and annotation. Intelligent decision-making and collaborative control stage: pre-rehearse interference scenario strategies, generate task sequences, and use RBF neural networks to compensate for system uncertainties; super-torsional sliding mode controllers suppress vibrations, and variable stiffness flexible joints (3) switch stiffness to absorb sudden impacts; depth cameras generate obstacle bounding boxes, replan trajectories based on artificial potential field methods, and provide collision warnings simultaneously; Flexible execution and structural self-adaptation stage: The magnetorheological fluid damping drive joint, composed of components of the magnetorheological fluid tube (306), realizes dynamic adjustment of stiffness; the execution end piece (6) performs electromagnetic adsorption and vacuum gripping operations; Data closed loop and system evolution stage.

10. A dynamic scene adaptive operation system, used to execute the dynamic scene adaptive operation method as described in claim 9, characterized in that, The dynamic scene adaptive operation system includes: An environmental perception and dynamic modeling system is provided with a multimodal perception fusion module, which integrates a vision system, tactile feedback and environmental dynamic modeling. Vision system: A multi-camera cluster with compound eyes combined with a multispectral sensor enables multi-scale scanning of scenes and material recognition; Tactile and acoustic feedback: Flexible tactile skin (5) monitors contact force; Acoustic microphone array predicts the slippage trend of objects through friction sound patterns; Environmental dynamic modeling: Solid-state millimeter radar (203) and IMU construct a SLAM++ system to predict the movement trajectory of obstacles in real time, and integrate temperature, humidity and air pressure sensors to perceive environmental disturbances; The intelligent decision-making and adaptive control system has a hierarchical control architecture and obstacle avoidance and fault tolerance mechanisms. Layered control architecture: Millisecond-level stiffness adjustment of variable stiffness flexible joint (3) to cope with sudden impact; online parameter identification algorithm updates inertia / friction coefficient model in real time, combined with RBF neural network to dynamically compensate for system uncertainty; spatiotemporal graph convolutional network analyzes the correlation of object motion, and reinforcement learning pre-plays thousand-level interference scenario strategy; Obstacle avoidance and fault tolerance mechanisms: The depth camera generates bounding boxes for human bodies / obstacles, and the trajectory is replanned in real time based on the artificial potential field method to ensure a safe distance; The flexible execution and structural innovation system has a biomimetic drive structure inside, which is a magnetorheological fluid damping drive joint composed of components of the magnetorheological fluid tube (306) to achieve dynamic adjustment of stiffness; the modular execution end piece (6) supports electromagnetic adsorption and vacuum gripping operations. The data closed-loop and simulation optimization system is equipped with edge-cloud collaborative computing and simulation transfer training.

Images