Multi-mode body-equipped intelligent robot control device

Abstract

The invention relates to the technical field of body-equipped intelligent robots, and discloses a multi-mode body-equipped intelligent robot control device, which comprises a brain-like decision model, a brain-like control strategy, a robot body and a sensing system which are connected in sequence, in the brain-like decision model part, an artificial neuron is adopted to construct a network to make a decision on behaviors of the robot, and an upper-layer control command is generated; the brain-like control strategy part is used for converting an upper-layer control command into a bottom-layer control command according to a specific application scene; the robot body executes a specified task and generates an action according to the received bottom layer control command; and the sensing system obtains the environment state change condition caused by the action of the robot body and updates the state information as the input of the brain-like decision model part at the next moment, so that a new round of control loop is started. According to the brain-like control strategy, a man-machine coordination motion control method is adopted for cooperative control, and a better interaction force control effect and higher position following precision are achieved for control over the robot body.

Description

Technical Field

[0001] The present invention relates to the technical field of embodied intelligent robots, and in particular to a multi-modal embodied intelligent robot control device. Background Art

[0002] Intelligent robots play an increasingly important role in industrial production and daily life. For example, in the field of intelligent manufacturing, higher and higher requirements are placed on the robot's dexterous motion control and fine force control capabilities; in the medical field, robots are required to work closely with humans to provide more intelligent and refined medical services; in extreme environments such as disaster relief, the application demand for robots is also growing.

[0003] At present, robot technology is developing towards embodied intelligence. Embodied intelligent robots aim to build robot systems with autonomous environmental perception, reliable intelligent decision-making and natural motion operation planning. With the increasing requirements for robot autonomy and intelligence, robot control still faces many challenges such as weak motion compliance and poor robustness. In particular, there is still a lot of room for improvement in maintaining coordinated motion control, anthropomorphic compliant operation, and efficient human-machine collaboration in unstructured environments. On the one hand, this is because the performance of existing robot systems is restricted by factors such as sensor accuracy, system repeatability, auxiliary mechanism performance, and energy consumption. On the other hand, this is because unstructured environments are complex, changeable, noisy, and highly uncertain. In the face of these complex and unknown environments, traditional robot control technology is laborious, complex, and error-prone. Summary of the invention

[0004] The present invention provides a multimodal embodied intelligent robot control device, whose brain-like control strategy adopts a human-machine coordinated motion control method for collaborative control, and the control of the robot body has better interactive force control effect and higher position following accuracy.

[0005] The present invention provides a multi-modal embodied intelligent robot control device, including a brain-like decision model, a brain-like control strategy, a robot body and a perception system connected in sequence;

[0006] The brain-like decision-making model part uses artificial neurons to construct a network, and makes decisions on the robot's behavior by analyzing the action goals and the external environment information at the current moment, and generates upper-level control commands; the brain-like control strategy part converts the upper-level control commands into low-level control commands according to the specific application scenarios; the robot body performs the designated tasks and generates actions according to the received low-level control commands; the perception system obtains the changes in the environmental state caused by the robot body's actions, and updates the state information as the input of the brain-like decision-making model part at the next moment, thereby starting a new round of control loop.

[0007] Furthermore, the brain-like decision model includes a perception module, a processing module, a decision module and a feedback module;

[0008] The perception module includes a camera, a microphone, a pressure sensor and a temperature sensor for collecting visual, auditory and tactile information, and is used to receive information about the external environment and convert the received external physical signals into digital signals that can be processed by the computer;

[0009] The processing module analyzes, processes and integrates the information obtained by the perception module. The processing module is composed of multiple neurons or computing units, which cooperate with each other through complex network connections to extract features and recognize patterns of information in order to understand and interpret the received information;

[0010] The decision module is used to make corresponding decisions according to the analysis results of the information by the processing module, and the decision layer selects a plan from multiple possible action plans based on the set decision rules or strategies;

[0011] The feedback module is used to provide feedback on the results of the decision module so as to evaluate and adjust the decision.

[0012] Furthermore, the brain-like control strategy adopts a human-machine coordinated motion control method for collaborative control, and the human-machine coordinated motion control method specifically includes:

[0013] A human-machine coordinated motion system model is established, a spring with a known stiffness coefficient k is added to the human-machine interface, and the expected impedance force f between the operator and the robot system is obtained according to the human-machine coordinated motion system model. e ,Right now The human-computer interaction force f is expressed as f = k(x d -x), where k e 、b e 、m e Respectively represent the equivalent stiffness coefficient, equivalent damping coefficient and equivalent mass of the operator's limbs, x e is the expected motion information of the operator's limbs, that is, the operator's expected motion trajectory, k is the stiffness of the added elastic element, x d is the motion information of the contact position between the robot system and the operator, which represents the actual motion information of the operator’s limbs in this system, and x is the motion information of the robot;

[0014] The motion execution system uses a servo system. The servo motor shaft is equipped with an encoder. By reading the encoder information, the motion state of the servo motor shaft can be obtained, and the motion information of the robot can be obtained: position x, speed Acceleration By spring deformation get: Obtaining motion information of the contact position between the robot system and the operator;

[0015] During the interaction between the robot system and the operator, the operator's motion parameters need to adjust the control system parameters in real time. According to the human-machine coordinated motion system model with elastic elements and on the basis of the adaptive impedance control algorithm, an adaptive impedance control system based on position-speed proportional compensation is adopted, including an adaptive controller, an impedance controller, a fuzzy controller, a PID controller, a servo motor and a motion actuator connected in sequence.

[0016] Furthermore, the implementation method of the adaptive controller includes:

[0017] Assume the expected human-computer interaction force is f d , the actual human-computer interaction force is f, and for the operator:

[0018]

[0019] When the robot is at low speed or stationary, it changes to ff d = kx(x e -x d ), human-computer interaction force f = k(x d -x), in order to achieve the desired contact force tracking, the target trajectory modified by impedance control is obtained, which is called the reference trajectory x r :

[0020]

[0021] Therefore, the contact force can be tracked by changing the desired trajectory. The contact force f is usually obtained by a force sensor. The sensor data has noise, so we get: Where e = x e -x;

[0022] Use x' e Represents the estimated operator's expected movement position, defined as: δx e =x' e -x, e'=e+δx e , substituting into the above formula and adding the adaptive adjustment term Ω, we get:

[0023] Assume that the learning rate is η, the controller sampling period is λ, and the regulation is To ensure stable error convergence;

[0024] Consider Ω as the sum of a sequence of p elements, set the initial value of the first term to 0, and set ε(t) = f(t)-f d (t), f'=kδx, we get After the transfer function is obtained through Laplace transform, consider when the system delay is λ, 0<λ<1 and p is large enough, and through Taylor expansion, the characteristic equation of the system is obtained:

[0025] λm e s 3 +λb e s 2 +λ(k e +k-kη)s+kη=0

[0026] According to the Routh criterion, for a system to be stable, it must satisfy:

[0027]

[0028] That is, η satisfies Keep the system stable.

[0029] Furthermore, the position speed is proportionally compensated, and the specific implementation method includes:

[0030] In human-machine coordinated motion, the expected force is defined as zero. Based on the force sensor data, the operator’s maximum interaction force f during motion can be obtained. max , then define the force reduction ratio k m , and get the interaction force reduction value f m is: f m =k m f max ;

[0031] Define the position speed compensation ratio as p x 、p v , the learning rate is η, the controller sampling period is λ, and it is stipulated that:

[0032]

[0033]

[0034] The formula with adaptive adjustment term Ω is rewritten as: Thus, the proportional compensation of position and speed is realized; the force reduction ratio k m Set according to operator comfort level.

[0035] Further, the robot body includes a control unit, an actuator, a controller object and a detection element connected in sequence to implement instructions for collaborative control using a human-machine coordinated motion control method;

[0036] The perception system includes an encoder, a force sensor, a distance sensor and a camera to perceive changes in environmental conditions and update the state information of the robot body to the brain-like decision model.

[0037] The beneficial effects of the present invention are:

[0038] The present invention includes a brain-like decision model, a brain-like control strategy, a robot body and a perception system connected in sequence; the brain-like decision model makes decisions on the robot's behavior and generates an upper-level control command, the brain-like control strategy converts the upper-level control command into a lower-level control command, and the robot body performs a designated task and generates an action according to the received lower-level control command; the perception system obtains the change in the environmental state caused by the robot body's action, and updates the state information as the input of the brain-like decision model part at the next moment, thereby starting a new round of control loop. In the brain-like control strategy, the method of adding elastic elements to the human-machine interface effectively reduces the human-machine interaction force, and further determines an adaptive impedance control method based on position-speed ratio compensation, which improves the control effect of the interaction force and the position following accuracy, makes it easier to realize the control of the human-machine interaction force, and has a better interaction force control effect and higher position following accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of the structure of the multi-modal embodied intelligent robot control device of the present invention.

[0040] Figure 2 It is a structural schematic diagram of the human-machine coordinated motion system model in the present invention.

[0041] Figure 3 It is a schematic diagram of the structure of the adaptive impedance control system based on position-speed proportional compensation in the present invention.

[0042] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0043] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0044] like Figure 1 As shown, the present invention provides a multi-modal embodied intelligent robot control device, including a brain-like decision model, a brain-like control strategy, a robot body and a perception system connected in sequence;

[0045] The brain-like decision-making model part uses artificial neurons to construct a network, and makes decisions on the robot's behavior by analyzing the action goals and the external environment information at the current moment, and generates upper-level control commands; the brain-like control strategy part converts the upper-level control commands into low-level control commands according to the specific application scenarios; the robot body performs the designated tasks and generates actions according to the received low-level control commands; the perception system obtains the changes in the environmental state caused by the robot body's actions, and updates the state information as the input of the brain-like decision-making model part at the next moment, thereby starting a new round of control loop.

[0046] Brain-like decision-making model:

[0047] The brain-like decision-making model includes a perception module, a processing module, a decision-making module and a feedback module;

[0048] (1) The perception module includes a camera, a microphone, a pressure sensor and a temperature sensor for collecting visual, auditory and tactile information, receiving information about the external environment, and converting the received external physical signals into digital signals that can be processed by the computer.

[0049] For vision, cameras are usually used as image acquisition devices, such as CMOS or CCD cameras, which can convert optical images into digital signals. Computer vision libraries such as OpenCV and Scikit-Image are used to read and preprocess images. Convolutional neural networks (CNNs) are used for image feature extraction and object recognition, such as classic AlexNet and VGGNet.

[0050] For hearing, microphone arrays are often used to collect sound signals. Some high-end ones are also equipped with audio preprocessing chips for preliminary amplification, filtering and other processing of the sound. Use Python's pyaudio library for sound recording and basic processing, or use professional audio processing software such as Adobe Audition to analyze audio data. Use Mel Frequency Cepstral Coefficients (MFCC) for speech feature extraction and combine with Hidden Markov Model (HMM) for speech recognition.

[0051] Touch can be realized through pressure sensors, temperature sensors, etc. In Python, relevant sensor driver libraries can be used, such as the Adafruit_DHT library for temperature sensors, to read and process sensor data. For tactile signal processing, signal filtering algorithms can be used to remove noise; for olfactory and taste sensor data, pattern recognition algorithms can be used to classify odors and tastes.

[0052] (2) The processing module analyzes, processes and integrates the information obtained by the perception module. The processing module is composed of multiple neurons or computing units, which cooperate with each other through complex network connections to extract features and recognize patterns of information in order to understand and interpret the received information.

[0053] The processing module includes general processing, graphics processor, digital signal processor and field programmable gate array. The general processor CPU is responsible for executing various instructions and data processing tasks, providing basic computing power for the processing layer. The graphics processor GPU is used to accelerate data processing and model training in deep learning, especially when processing images, videos and other data. The digital signal processor DSP is specially used for fast processing of digital signals, and can perform operations such as filtering, encoding and decoding. The field programmable gate array FPGA has reconfigurability and high-speed parallel processing capabilities. The hardware circuit can be customized according to specific needs to implement specific signal processing algorithms. At the same time, the algorithms used by the processing module include: feature extraction algorithms. In the field of images, there are algorithms such as scale-invariant feature transform (SIFT) and speeded up robust features (SURF) for extracting key features of images; in text processing, methods such as bag-of-words model and TF-IDF can extract text features. Data preprocessing algorithms, including normalization, standardization, data enhancement and other algorithms, are used to clean and transform raw data, improve data quality, and enhance the generalization ability of the model. Deep learning algorithms, convolutional neural networks (CNN) are used to process image and video data, recurrent neural networks (RNN) and its variants LSTM and GRU are used to process sequence data, and autoencoders (AE) can be used for data dimensionality reduction and feature learning.

[0054] (3) The decision-making module is used to make corresponding decisions based on the analysis results of the information by the processing module, and the decision-making layer selects a plan from multiple possible action plans based on the set decision rules or strategies.

[0055] The decision-making module includes high-performance servers and dedicated decision-making chips. High-performance servers have strong computing and storage capabilities, which are used to run complex decision-making models and process large amounts of data. They support multiple processors, large-capacity memory, and high-speed hard disks to meet the computing and data storage requirements in the decision-making process. Dedicated decision-making chips are artificial intelligence chips based on specific algorithms, which are optimized for decision-making tasks and can quickly process and analyze data to make decisions. The algorithms used are machine learning decision algorithms, including decision trees, random forests, support vector machines (SVMs), and other algorithms. By learning the patterns and rules in the data, decision models are established to classify and predict new data, thereby making decisions.

[0056] (4) The feedback module is used to provide feedback on the results of the decision module so as to evaluate and adjust the decision. The feedback mechanism can help the model continuously optimize the decision strategy based on the actual results and improve the accuracy and efficiency of the decision.

[0057] Brain-like control strategy:

[0058] The brain-like control strategy adopts a human-machine coordinated motion control method for collaborative control, and the human-machine coordinated motion control method specifically includes:

[0059] Establish a human-machine coordinated motion system model. The human limbs have certain damping and stiffness, and their size is related to the size of muscle force. When establishing a human-machine interaction model, the operator's limbs can be simplified into an impedance model. Add a spring with a known stiffness coefficient of k at the human-machine interface, so that the interaction between the entire robot system and the operator can be simplified into a single-degree-of-freedom model, such as Figure 2 As shown in the figure, M is the mass of the robot motion actuator and load.

[0060] According to the human-machine coordinated motion system model, the expected impedance force f of the operator and the robot system is obtained. e ,Right now:

[0061]

[0062] The human-computer interaction force f is expressed as:

[0063] f = k(x d -x) (2)

[0064] Among them, k e 、b e 、m e Respectively represent the equivalent stiffness coefficient, equivalent damping coefficient and equivalent mass of the operator's limbs, x e is the expected motion information of the operator's limbs, that is, the operator's expected motion trajectory, k is the stiffness of the added elastic element, x d is the motion information of the contact position between the robot system and the operator, which represents the actual motion information of the operator's limbs in this system, and x is the motion information of the robot; the stiffness k of the spring is known, and the human-machine interaction force f can be measured by the force sensor. In order to realize impedance control, the motion information of the environment, that is, the actual motion information of the operator's limbs, must be obtained.

[0065] The motion execution system uses a servo system. The servo motor shaft is equipped with an encoder. By reading the encoder information, the motion state of the servo motor shaft can be obtained, and the motion information of the robot can be obtained: position x, speed Acceleration From the spring deformation:

[0066]

[0067] get:

[0068]

[0069] With the above formulas (4) to (6), the motion information of the contact position between the robot system and the operator is obtained.

[0070] In the process of interaction between the robot system and the operator, the operator's motion parameters need to adjust the control system parameters in real time. According to the human-machine coordinated motion system model with elastic elements, on the basis of the adaptive impedance control algorithm, an adaptive impedance control system based on position-speed proportional compensation is adopted, such as Figure 3 As shown, it includes an adaptive controller, an impedance controller, a fuzzy controller, a PID controller, a servo motor and a motion actuator connected in sequence.

[0071] The implementation method of the adaptive controller includes:

[0072] Assume the expected human-computer interaction force is f d , the actual human-computer interaction force is f, and for the operator:

[0073]

[0074] When the robot is at low speed or stationary, formula (7) becomes:

[0075] ff d =k e (x e -x d ) (8)

[0076] Formula 2 obtains the human-computer interaction force f = k(x d -x), in order to achieve the desired contact force tracking, the target trajectory modified by impedance control is obtained according to Formula 2 and Formula (8), which is called the reference trajectory x r :

[0077]

[0078] Therefore, the contact force can be tracked by changing the desired trajectory. The contact force f is usually obtained by a force sensor. The sensor data has noise. According to formula (7) and formula (9), it is obtained:

[0079]

[0080] Where e = x e -x.

[0081] Use x' eIt indicates the estimated position of the operator's expected movement, and is defined as:

[0082] δx e =x' e -x (11)

[0083] e'=e+δx e (12)

[0084] Substituting formulas (11) and (12) into formula (10), and adding the adaptive adjustment term Ω, we obtain:

[0085]

[0086] Assume that the learning rate is η, the controller sampling period is λ, and the regulation is To ensure stable error convergence.

[0087] Substitute formulas (4) to (6) into formula (13), and consider Ω as the sum of a sequence of p elements, set the initial value of the first term to 0, and set ε(t) = f(t) - f d (t), f'=kδx, and substitute into formula (13), we get

[0088]

[0089] After the transfer function is obtained through Laplace transform, consider when the system delay is λ, 0<λ<1 and p is large enough, and through Taylor expansion, the characteristic equation of the system is obtained:

[0090] λm e s 3 +λb e s 2 +λ(k e +k-kη)s+kη=0

[0091] According to the Routh criterion, for a system to be stable, it must satisfy:

[0092]

[0093] That is, η satisfies Keep the system stable.

[0094] Based on the above implementation method of the adaptive controller, the position speed is proportionally compensated. The specific implementation methods include:

[0095] In human-machine coordinated motion, the expected force is defined as zero. Based on the force sensor data, the operator’s maximum interaction force f during motion can be obtained. max , then define the force reduction ratio k m , and get the interaction force reduction value f m for:

[0096] f m =k m f max (16)

[0097] Define the position speed compensation ratio as p x 、p v , the learning rate is η, the controller sampling period is λ, and it is stipulated that:

[0098]

[0099]

[0100] Formula (13) with the adaptive adjustment term Ω is rewritten as:

[0101]

[0102] Thus, the proportional compensation of position and speed is realized; the force reduction ratio k n Set according to operator comfort level.

[0103] In the inner loop of force control, fuzzy control is used to adjust PID parameters in real time. In terms of control effect, fuzzy PID has smaller overshoot and better stability than traditional PID. When designing the fuzzy controller, the domain of the input (e, ec) is set to [-3,3], and the output Δk P The domain of discourse is set to [-0.3, 0.3], Δk I The domain of discourse is set to [-0.06, 0.06], Δk D The domain is set to [-3,3]. In actual application, the domain needs to be scaled.

[0104] The robot body includes a control unit, an actuator, a controller object and a detection element connected in sequence, and implements instructions for collaborative control using a human-machine coordinated motion control method.

[0105] The perception system includes an encoder, a force sensor, a distance sensor and a camera to perceive changes in environmental conditions and update the state information of the robot body to the brain-like decision model.

[0106] The present invention includes a brain-like decision model, a brain-like control strategy, a robot body and a perception system connected in sequence; the brain-like decision model makes decisions on the robot's behavior and generates an upper-level control command, the brain-like control strategy converts the upper-level control command into a lower-level control command, and the robot body performs a designated task and generates an action according to the received lower-level control command; the perception system obtains the change in the environmental state caused by the robot body's action, and updates the state information as the input of the brain-like decision model part at the next moment, thereby starting a new round of control loop. In the brain-like control strategy, the method of adding elastic elements to the human-machine interface effectively reduces the human-machine interaction force, and further determines an adaptive impedance control method based on position-speed ratio compensation, which improves the control effect of the interaction force and the position following accuracy, makes it easier to realize the control of the human-machine interaction force, and has a better interaction force control effect and higher position following accuracy.

[0107] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0108] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A multi-modal embodied intelligent robot control device, characterized in that: It includes a brain-like decision-making model, a brain-like control strategy, a robot body and a perception system connected in sequence; The brain-like decision-making model part uses artificial neurons to construct a network, and makes decisions on the robot's behavior by analyzing the action goals and the external environment information at the current moment, and generates upper-level control commands; the brain-like control strategy part converts the upper-level control commands into low-level control commands according to the specific application scenarios; the robot body performs the designated tasks and generates actions according to the received low-level control commands; the perception system obtains the changes in the environmental state caused by the robot body's actions, and updates the state information as the input of the brain-like decision-making model part at the next moment, thereby starting a new round of control loop.

2. The multimodal embodied intelligent robot control device according to claim 1, characterized in that: The brain-like decision-making model includes a perception module, a processing module, a decision-making module and a feedback module; The perception module includes a camera, a microphone, a pressure sensor and a temperature sensor for collecting visual, auditory and tactile information, and is used to receive information about the external environment and convert the received external physical signals into digital signals that can be processed by the computer; The processing module analyzes, processes and integrates the information obtained by the perception module. The processing module is composed of multiple neurons or computing units, which cooperate with each other through complex network connections to extract features and recognize patterns of information in order to understand and interpret the received information; The decision module is used to make corresponding decisions according to the analysis results of the information by the processing module, and the decision layer selects a plan from multiple possible action plans based on the set decision rules or strategies; The feedback module is used to provide feedback on the results of the decision module so as to evaluate and adjust the decision.

3. The multimodal embodied intelligent robot control device according to claim 1, characterized in that: The brain-like control strategy adopts a human-machine coordinated motion control method for collaborative control, and the human-machine coordinated motion control method specifically includes: A human-machine coordinated motion system model is established, a spring with a known stiffness coefficient k is added to the human-machine interface, and the expected impedance force f between the operator and the robot system is obtained according to the human-machine coordinated motion system model. e ,Right now The human-computer interaction force f is expressed as f = k(x d -x), where k e 、b e 、m e Respectively represent the equivalent stiffness coefficient, equivalent damping coefficient and equivalent mass of the operator's limbs, x e is the expected motion information of the operator's limbs, that is, the operator's expected motion trajectory, k is the stiffness of the added elastic element, x d is the motion information of the contact position between the robot system and the operator, which represents the actual motion information of the operator’s limbs in this system, and x is the motion information of the robot; The motion execution system uses a servo system. The servo motor shaft is equipped with an encoder. By reading the encoder information, the motion state of the servo motor shaft can be obtained, and the motion information of the robot can be obtained: position x, speed Acceleration By spring deformation get: Obtaining motion information of the contact position between the robot system and the operator; During the interaction between the robot system and the operator, the operator's motion parameters need to adjust the control system parameters in real time. According to the human-machine coordinated motion system model with elastic elements and on the basis of the adaptive impedance control algorithm, an adaptive impedance control system based on position-speed proportional compensation is adopted, including an adaptive controller, an impedance controller, a fuzzy controller, a PID controller, a servo motor and a motion actuator connected in sequence.

4. The multi-modal embodied intelligent robot control device according to claim 3, characterized in that: The implementation method of the adaptive controller includes: Assume the expected human-computer interaction force is f d , the actual human-computer interaction force is f, and for the operator: When the robot is at low speed or stationary, it changes to ff d =k e (x e -x d ), human-computer interaction force f = k(x d -x), in order to achieve the desired contact force tracking, the target trajectory modified by impedance control is obtained, which is called the reference trajectory x r : Therefore, the contact force can be tracked by changing the desired trajectory. The contact force f is usually obtained by a force sensor. The sensor data has noise, so we get: Where e = x e -x; Use x' e Represents the estimated operator's expected movement position, defined as: δx e =x' e -x, e'=e+δx e , substituting into the above formula and adding the adaptive adjustment term Ω, we get: Assume that the learning rate is η, the controller sampling period is λ, and the regulation is To ensure stable error convergence; Consider Ω as the sum of a sequence of p elements, set the initial value of the first term to 0, and set ε(t) = f(t)-f d (t), f'=kδx, we get After the transfer function is obtained through Laplace transform, consider when the system delay is λ, 0<λ<1 and p is large enough, and through Taylor expansion, the characteristic equation of the system is obtained: λm e s 3 +λb e s 2 +λ(k e +k-kη)s+kη=0 According to the Routh criterion, for a system to be stable, it must satisfy: That is, η satisfies Keep the system stable.

5. The multimodal embodied intelligent robot control device according to claim 4, characterized in that: Proportional compensation of position speed is performed. The specific implementation methods include: In human-machine coordinated motion, the expected force is defined as zero. Based on the force sensor data, the operator’s maximum interaction force f during motion can be obtained. max , then define the force reduction ratio k m , and get the interaction force reduction value f m is: f m =k m f max ; Define the position speed compensation ratio as p x 、p v , the learning rate is η, the controller sampling period is λ, and it is stipulated that: The formula with adaptive adjustment term Ω is rewritten as: Thus, the proportional compensation of position and speed is realized; the force reduction ratio k m Set according to operator comfort level.

6. The multi-modal embodied intelligent robot control device according to claim 5, characterized in that: The robot body includes a control unit, an actuator, a controller object and a detection element connected in sequence, and implements instructions for collaborative control using a human-machine coordinated motion control method; The perception system includes an encoder, a force sensor, a distance sensor and a camera to perceive changes in environmental conditions and update the state information of the robot body to the brain-like decision model.

Images