High-risk operation robot control system and method simulating human nervous system

Abstract

The invention belongs to the field of robot control, and provides a high-risk operation robot control system and method simulating a human nervous system. The high-risk operation robot control system simulating the human nervous system comprises a sensing element, a lower computer, a middle computer and an upper computer. The sensing element is used for sensing robot state information and transmitting the robot state information to the corresponding lower computer; the lower computer is used for transmitting the received robot state information to the upper computer through the middle computer; the upper computer is used for forming instruction information according to a high-risk operation preset target in combination with the received robot state information and issuing the instruction information to the corresponding lower computer through the middle computer; the lower computer is further used for controlling the corresponding moving parts to execute the corresponding instruction information and feeding the received robot state information back to the upper computer through the middle computer, and sensing-decision-execution closed-loop control is formed.

Description

Technical Field

[0001] The present invention belongs to the field of robot control, and particularly relates to a high-risk operation robot control system and method that mimics the human nervous system. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention, and do not necessarily constitute prior art.

[0003] In a high-risk and complex environment, it is a challenging task for a robot to perform operations such as rescue and detection. Taking the post-earthquake environment as an example, the post-earthquake environment is complex, the surface features of cities and mountains are different, and aftershocks occur. When performing tasks such as rescue, the robot is required to have a high reaction speed - the ability of real-time perception and instantaneous reaction ability. This determines that the control system cannot adopt a distributed architecture and a hybrid architecture. These two architectures are not suitable for application in this kind of environment in terms of data real-time performance, cost, and complexity. The high-risk and complex environment has relatively high requirements for the motion complexity and motion accuracy of the robot. A single controller, such as directly using a Raspberry Pi 4B for control, although it has good real-time performance and low cost compression, has limited computing power, poor scalability, and is not suitable for high-risk and complex environments. Summary of the Invention

[0004] In order to solve the technical problems existing in the above background technique, the present invention provides a high-risk operation robot control system and method that mimics the human nervous system. It fully considers various factors such as the chip computing power level, expansion ability, ease of use, and high-risk and complex environment, refers to the human nervous system, and can ensure the rapid reaction and motion accuracy of the robot in a high-risk and complex environment.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions: The first aspect of the present invention provides a high-risk operation robot control system that mimics the human nervous system.

[0006] A high-risk operation robot control system that mimics the human nervous system includes: a sensing element, a lower computer, a middle computer, and an upper computer; The sensing element is used to sense the robot state information and transmit it to the corresponding lower computer; the lower computer is used to transmit the robot state information it receives to the upper computer through the middle computer; the upper computer is used to form command information according to the preset target of high-risk operation, combine the received robot state information, and issue it to the corresponding lower computer through the middle computer; the lower computer is also used to control the corresponding motion components to execute the corresponding command information, and at the same time transmit the robot state information it receives to the upper computer through the middle computer to form a perception-decision-execution closed-loop control; Among them, the communication between the lower computer and the middle computer, as well as between the middle computer and the upper computer, all adopt the general instruction format; the general instruction format is: [Frame Header][Control Flag][Instruction Type][Data Area][Check Code][Frame Tail].

[0007] As an implementation method, the data area consists of multiple data blocks, and the format of each data block is: [Action][Target][Value]; the value is variably encoded according to its range and type.

[0008] As an implementation method, in the data block, if the value range ≤ 255, 1 byte is used; if the value range ≤ 65535, 2 bytes are used; if it is a floating-point number, 4 bytes are used; As an implementation method, in the data block, the action is 1 byte, the high 2 bits represent the action type, and the low 6 bits are the action ID; As an implementation method, in the data block, the target is 1 byte, which is used to represent the target device or joint number, and its range is 0 to 255.

[0009] As an implementation method, the model of the upper computer is: ; Among them, is the upper computer instruction; S high is the perception data directly interacting with the upper computer; API LLM is the interaction result of the large language model API; S MLC is the feedback data integrated by the middle computer.

[0010] As an implementation method, the model of the middle computer includes: an instruction decomposition model, a data integration model, and a lightweight computing model; The instruction decomposition model is: ; The data integration model is: ; The lightweight computing model is: ; Among them, is the upper computer instruction; C high,i is the decomposed upper computer sub-instruction; F is the instruction decomposition function; D mid is the middle layer data; S MLC is the feedback data integrated by the middle computer; D low,i is the lower computer data; is the lightweight processed lower computer feedback data; G MCC is the integration function; is the lightweight computing function. As an implementation manner, the lower computer model includes: a control signal generation model and a data preprocessing model; The control signal generation model is: ; The data preprocessing model: ; Wherein, C low,i is the lower computer instruction; S low,i is the sensing data; U i is the control signal; D low,i is the preprocessed data; is the control signal generation function; is the data preprocessing function.

[0011] As an implementation manner, the instruction priority of the upper computer is higher than that of the middle computer; the instruction actually executed by the lower computer is determined by the following logic: ; Wherein, is the upper computer sub-instruction; is the middle computer sub-instruction.

[0012] As an implementation manner, the delay in the information transmission process between the lower computer and the middle computer and between the middle computer and the upper computer is optimized by using an information flow delay model: The information flow delay model from the upper computer to the middle computer is: ; The information flow delay model from the middle computer to the lower computer is: ; The information flow delay model from the lower computer to the middle computer is: ; The information flow delay model from the middle computer to the upper computer is: ; Wherein, T high→mid is the delay for the upper computer to send an instruction to the middle computer; T mid→low is the delay for the middle computer to send an instruction to the lower computer; T low→mid is the delay for the lower computer to send back data to the middle computer; T mid→high is the delay for the middle computer to send back data to the upper computer; C low,i is the lower computer instruction; is the upper instruction; D low,i is the lower computer data; is the upper computer data.

[0013] The second aspect of the present invention provides a control method for a high-risk operation robot that mimics the human nervous system.

[0014] A control method for a high-risk operation robot that mimics the human nervous system, based on the high-risk operation robot control system that mimics the human nervous system as described above; includes: The sensing element senses the robot state information and transmits it to the corresponding lower computer; the lower computer transmits the received robot state information to the upper computer through the middle computer; the upper computer forms command information according to the preset high-risk operation target, combines the received robot state information, and issues it to the corresponding lower computer through the middle computer; the lower computer controls the corresponding moving parts to execute the corresponding command information, and at the same time transmits the received robot state information to the upper computer through the middle computer to form a perception-decision-execution closed-loop control; Among them, the lower computer and the middle computer, as well as the middle computer and the upper computer, communicate using a general instruction format; the general instruction format is: [frame header][control flag][instruction type][data area][check code][frame tail].

[0015] The beneficial effects of the present invention are: (1) The high-risk operation robot control system of the present invention that mimics the human nervous system has a hierarchical architecture to optimize system performance and scalability; among them, modular design: the three-level hierarchical architecture of the upper computer (HLC), middle computer (MLC), and lower computer (LLC) realizes function decoupling. The upper computer focuses on global task planning and intelligent decision-making (such as calling the large language model API), the middle computer is responsible for instruction decomposition and lightweight calculation, and the lower computer executes specific motion control and low-precision data acquisition. This design significantly improves the parallel processing ability and resource allocation efficiency of the system; flexible expansion: through the data integration ability of the middle computer, the system can dynamically expand the number of sensors and actuators to adapt to the changing hardware requirements in high-risk environments (such as adding new sensors or robotic arm joints).

[0016] (2) The present invention has high reliability: the lower computer directly controls the key moving parts to ensure the instantaneous response of emergency instructions (such as obstacle avoidance), and the upper computer priority logic (Clow,i) ensures the control safety in high-risk scenarios.

[0017] (3) The general instruction format of the present invention is efficient and reliable: High transmission efficiency: The dynamic data area and variable-length coding design (such as the numerical field is allocated 1 / 2 / 4 bytes according to requirements) reduce redundant data. Compared with Modbus RTU, the transmission efficiency of small data volume is increased to 50% (N = 1), and the efficiency of large data volume reaches 89.5% (N = 10).

[0018] Strong fault tolerance ability: Adopting double frame headers / frame tails (0xAA55 and 0xFFEE) and CRC-16 checksum, the packet loss rate is reduced by 50% in a noisy environment, and the error detection rate reaches 99.998%, significantly enhancing the communication reliability in complex electromagnetic environments.

[0019] Dynamic expansion support: The instruction type field (1 byte) and the data block structure (action - target - value) support diverse instruction definitions, are compatible with multiple types of sensors and actuators, and adapt to the dynamic requirements in high - risk tasks.

[0020] (4) The bionic design of the present invention improves environmental adaptability: Nervous - system - like response: Referring to the information - transmission mechanism of the human nervous system (such as priority control and closed - loop feedback), an efficient closed - loop of "perception - decision - execution" is achieved, ensuring the rapid response and motion accuracy of the robot in complex environments (such as areas with frequent aftershocks).

[0021] Multi - sensor fusion: The upper computer integrates high - precision sensors such as lidar to achieve SLAM mapping, combines with the middle computer to process low - frequency sensor data, forms a multi - level environmental perception ability, and improves the operation stability under complex terrains.

[0022] (5) The comprehensive performance of the present invention has significant advantages: Low latency and high real - time performance: Through the optimization of the information - flow latency model (such as the compression of the latency from the middle computer to the lower computer Tmid→low), the overall response speed of the system meets the millisecond - level requirements of high - risk scenarios.

[0023] Advantages of additional aspects of the present invention will be partly given in the following description, partly will become obvious from the following description, or will be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The schematic diagram of the control system of the high - risk operation robot imitating the human nervous system in the embodiment of the present invention is used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention.

[0025] Figure 1 It is the schematic diagram of the control system of the high - risk operation robot imitating the human nervous system in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0026] The present invention will be further described below in conjunction with the drawings and embodiments.

[0027] It should be noted that the following detailed description is illustrative and is intended to provide a further description of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0028] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0029] In the present invention, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. They are only relational terms determined for the convenience of describing the structural relationship of each component or element of the present invention and do not specifically refer to any component or element in the present invention and should not be construed as a limitation to the present invention.

[0030] In the present invention, terms such as "fixed connection", "connected", "connected" should be understood in a broad sense, which may mean a fixed connection, an integral connection or a detachable connection; it may be directly connected or indirectly connected through an intermediate medium. For relevant scientific research or technical personnel in this field, the specific meanings of the above terms in the present invention can be determined according to specific circumstances and should not be construed as a limitation to the present invention.

[0031] Term Explanation: Host computer: The top core unit of the control system, responsible for global task planning, intelligent decision-making (such as calling the large language model API) and interaction with high-precision sensors (such as lidar).

[0032] Middle computer: The intermediate layer processing unit, responsible for instruction decomposition, data integration and lightweight computing tasks.

[0033] Lower computer: The bottom layer execution unit, directly controlling moving parts (such as motors, robotic arms) and low-precision, low-frequency sensors.

[0034] General instruction format: A standardized communication protocol for instruction transmission within the system.

[0035] Variable-length coding: A coding method that dynamically allocates storage space according to the data value range.

[0036] CRC-16 check: A type of Cyclic Redundancy Check, used to detect error codes in data transmission.

[0037] STM32F103C8T6: A microcontroller based on the ARM Cortex-M3 core, used for high-real-time motion control.

[0038] STM32F407: A high-performance ARM Cortex-M4 microcontroller that supports lightweight computing and multitasking.

[0039] Raspberry Pi 4B: A single-board computer based on the ARM architecture, serving as the core of the system control.

[0040] SLAM: Simultaneous Localization and Mapping technology that realizes dynamic environment modeling through sensors such as lidar.

[0041] Large Language Model API: An intelligent interaction interface based on Natural Language Processing (NLP) that supports semantic understanding and decision-making generation.

[0042] Priority control logic: A mechanism that ensures the execution of host computer instructions prior to the autonomous instructions of the middle-level computer.

[0043] Information delay model: Quantifies the delay parameters of data transmission between different levels of the system, used to optimize real-time performance.

[0044] Dynamic expandability: The system's ability to support the flexible addition of sensors or actuators.

[0045] As Figure 1 shown, the embodiment of the present invention provides a high-risk operation robot control system that mimics the human nervous system, which includes: sensing elements, lower-level computers, middle-level computers, and host computers. The sensing elements are used to sense the robot's status information and transmit it to the corresponding lower-level computer; the lower-level computer is used to transmit the robot's status information it receives to the host computer via the middle-level computer; the host computer is used to form command information based on the preset high-risk operation targets and the received robot's status information and issue it to the corresponding lower-level computer via the middle-level computer; the lower-level computer is also used to control the corresponding moving parts to execute the corresponding command information, and at the same time, transmit the robot's status information it receives to the host computer via the middle-level computer, forming a perception-decision-execution closed-loop control.

[0046] In the embodiment of the present invention, the sensing elements are sensors. Among them, the host computer is connected to most of the sensors and a small number of moving parts that play an auxiliary role, and is the control core of the entire system. The host computer uses the API of the large language model to make the entire system more intelligent. The control level of the host computer over the lower-level computer is higher than that of the middle-level computer. In the designed robot model, the lower-level computer uses stm32f103c8t6 (hereinafter referred to as f103), the middle-level computer uses stm32f407 (hereinafter referred to as f407), and the host computer uses Raspberry Pi 4B. Two f103s control a six-degree-of-freedom robotic arm, which is connected to f407. f407 controls four motors to control the robot's movement and is connected to Raspberry Pi 4B. And Raspberry Pi is connected to the lidar to achieve SLAM mapping.

[0047] In this embodiment, the communication between the lower computer and the middle computer, as well as between the middle computer and the upper computer, all adopt the general instruction format; the general instruction format is: [frame header][control flag][instruction type][data area][check code][frame tail].

[0048] Table 1 shows the field details

[0049] Among them, the data area consists of multiple data blocks, and the format of each data block is: [action][target][value]; the value is variably encoded according to its range and type.

[0050] In the data block, if the value range ≤ 255, 1 byte is used; if the value range ≤ 65535, 2 bytes are used; if it is a floating point number, 4 bytes are used (e.g., IEEE 754 standard); In the data block, the action is 1 byte, the high 2 bits represent the action type (such as move = 00, rotate = 01), and the low 6 bits are the action ID; The target is 1 byte, which is used to represent the target device or joint number, and its range is 0 to 255.

[0051] Performance index calculation: ① Total length formula:

[0052] Valid data:

[0053] Frame header: 2 bytes, which identifies the start position of the data frame.

[0054] Control flag: 1 byte, which is used to transmit control information (such as priority, retransmission flag, etc.).

[0055] Instruction type: 1 byte, which represents the type of the data frame or operation instruction (such as read, write, response, etc.).

[0056] Data block overhead: 2 bytes appended to each data block, which are usually used to store length or identification information.

[0057] L data: The length of the valid data in a single data block (unit: byte).

[0058] N: The number of data blocks.

[0059] Check code: 2 bytes, which are used for error detection (such as CRC check).

[0060] Frame tail: 2 bytes, which identifies the end position of the data frame.

[0061] Fixed overhead: Total 8 bytes (2 for frame header + 1 for control flag + 1 for instruction type + 2 for checksum + 2 for frame tail) ② Transmission efficiency:

[0062] Numerator: The sum of the effective data and additional overhead of all data blocks (i.e., the "effective part" actually transmitted).

[0063] Denominator: Total frame length (effective part + fixed overhead).

[0064] Physical meaning: Measures the proportion of effective data in the total transmitted data in the protocol. The closer the value is to 1, the higher the efficiency.

[0065] ③ Information density:

[0066] The physical meaning of the parameter is the same as that of the transmission efficiency. Here, it may represent the proportion of effective data per unit length. ④ Reliability Using CRC-16 checksum: Error detection rate 99.998%; Dual frame headers / frame tails: Reduces the packet loss rate (the packet loss rate is reduced by 50% in a noisy environment); In the embodiments of the present invention, the model of the host computer is: ; Wherein, is a high-level instruction; S high is the perception data directly interacting with the host computer; API LLM is the interaction result of the large language model API; S MLC is the feedback data integrated by the middle computer.

[0067] The model of the middle computer includes: an instruction decomposition model, a data integration model, and a lightweight computing model; The instruction decomposition model is: ; C high : Host computer instruction, representing a composite instruction or complex control command received by the system (such as "start full-process monitoring").

[0068] F(): Decomposition function, responsible for decomposing the high-level instruction into multiple independently executable sub-instructions.

[0069] C high,i : Decomposed host computer sub-instructions (such as "sensor initialization", "data acquisition start", etc.).

[0070] n: The number of sub-instructions, which depends on the complexity of the high-level instructions and the division of labor of the system execution units.

[0071] The data integration model is: ; D low,i : Lower computer data, representing raw data (such as temperature values, voltage signals) from different sources (such as sensors, subsystems).

[0072] G MCC : Integration function, responsible for fusing, filtering or structuring multi-source data (such as noise removal and time alignment).

[0073] D mid : Middle-level data, integrated structured data sets (such as time series data tables, feature vectors).

[0074] S MCC : Valid data subset, which represents key data that meets specific conditions after integration (such as anomaly detection results, key indicators).

[0075] The lightweight calculation model is: ; G ` MCC : Lightweight computing functions, using simplified algorithms or optimization strategies (such as approximate computing, dimensionality reduction processing) D ` mid : Lightweight middle-level data may sacrifice some accuracy to improve efficiency (such as compressed feature representation) in, is the set of instructions for the lower computer; F is the instruction decomposition function; For example: def F(C_high): # Predefined instruction template library templates = { "Monitoring type": "Sensor initialization (ID: {sensor_ids})", "Set data acquisition frequency to {frequency}", "Enable real-time data stream transmission to {target}" ], "Control type": "Execute device calibration (mode: {mode})", "Set threshold range: {threshold}" ] } # Step 1: Parse instruction type and parameters if "Full process monitoring" in C_high: instruction_type = "Monitoring type" params = { "sensor_ids": "S001 - S005", "frequency": "10Hz", "target": "Cloud" } else: raise ValueError("Unknown instruction type") # Step 2: Generate sub-instructions from templates sub_commands = [] for template in templates[instruction_type]: sub_command = template.format(**params) sub_commands.append(sub_command) return sub_commands D mid is the middle-level data; S MLC is the feedback data integrated by the middle-level computer; is the feedback data of the lower-level computer; is the feedback data of the lower-level computer after lightweight processing; G MCC is the integration function; is the lightweight calculation function. For example: import numpy as np def G_prime_MCC(*D_low): # Define lightweight processing function phi_i def sliding_window_avg(data, window_size = 100): # Sliding window average: Compress L data points into L / / window_size points compressed = [] for i in range(0, len(data), window_size): window = data[i:i + window_size] compressed.append(np.mean(window)) return compressed # Define feature extraction function (example) def extract_features(compressed_data): features = { "Mean": np.mean(compressed_data), "Variance": np.var(compressed_data), "Peak": np.max(compressed_data) } return features # Process each low-level data in turn D_mid_prime = {} for idx, data in enumerate(D_low): # Step 1: Sliding window compression compressed = sliding_window_avg(data) # Step 2: Extract key features features = extract_features(compressed) # Mark data sources (such as temperature, humidity) D_mid_prime[f"Sensor_{idx + 1}"] = features return D_mid_prime The lower computer model includes: a control signal generation model and a data preprocessing model; The control signal generation model is: ; The data preprocessing model: ; Among them, C low,i is the lower computer instruction; S low,i is the sensing data; U i is the control signal; D low,i is the lower computer preprocessed data; is the control signal generation function; is the data preprocessing function.

[0076] Taking the PID control signal as an example: class PIDController: def __init__(self, Kp, Ki, Kd): self.Kp = Kp self.Ki = Ki self.Kd = Kd self.integral = 0.0 self.prev_error = 0.0 def compute(self, error, dt): self.integral += error * dt derivative = (error - self.prev_error) / dt output = self.Kp * error + self.Ki * self.integral + self.Kd * derivative self.prev_error = error return output # Example: Control the i-th joint (e.g., i = 1 corresponds to the knee joint) f_LLC_i = PIDController(Kp = 2.0, Ki = 0.1, Kd = 0.5) current_angle = 30.0# Current angle (from sensor) target_angle = 45.0# Target angle (from high-level instruction) error = target_angle - current_angle control_signal = f_LLC_i.compute(error, dt = 0.1)# dt is the sampling period Taking the preprocessing of high-frequency noise removal as an example: import numpy as np from scipy.signal import butter, lfilter def g_LLC(D_raw, fc = 10.0, fs = 100.0): # Step 1: Low-pass filtering (Butterworth filter) b, a = butter(4, fc / (0.5 * fs), 'low') D_filtered = lfilter(b, a, D_raw) # Step 2: Normalize to [-1, 1] D_max = np.max(np.abs(D_filtered)) D_normalized = D_filtered / (D_max + 1e-6) # To avoid division by zero return D_normalized # Example: Processing raw joint angular velocity data D_raw = np.random.normal(0, 5, 1000) # Raw data with noise (mean = 0, standard deviation = 5） D_clean = g_LLC(D_raw, fc=5.0, fs=100.0) As an implementation, the instruction priority of the host computer is higher than that of the middle computer; The instruction actually executed by the slave computer is determined by the following logic: ; wherein, is the sub-instruction of the host computer; is the sub-instruction of the middle computer.

[0077] Specifically, the delays in the information transmission processes between the slave computer and the middle computer, and between the middle computer and the host computer, are optimized using an information flow delay model: Host computer → Middle computer: Delay T high→mid : ; Middle computer → Slave computer: Delay T mid→low : ; Slave computer → Middle computer: Delay T low→mid : ; Middle computer → Host computer: Delay T mid→high : .

[0078] T high→mid: Delay in the upper computer sending instructions to the middle computer; T mid→low : Delay in the middle computer sending instructions to the lower computer; T low→mid : Delay in the lower computer sending back data to the middle computer; T mid→high : Delay in the middle computer sending back data to the upper computer; C low,i is the instruction of the lower computer; is the upper instruction; D low,i is the data of the lower computer; is the data of the upper computer.

[0079] The overall information flow of the system can be expressed as the following recurrence relation: Information flow output by the upper computer:

[0080] C (k) high : Global goals or system configuration parameters (such as the overall task goal, operation mode), for example, "complete environmental monitoring task".

[0081] G: The state of the upper computer at time k or high-level sensor feedback (such as global positioning status, system health status).

[0082] S (k) high : Middle-level data (such as integrated environmental characteristics, historical records) fed back after processing by the middle computer in the previous cycle k-1.

[0083] D (k-1) mid : Interface for interaction with the large language model, used to generate or optimize high-level instructions (such as natural language instruction parsing).

[0084] API LLM : High-level control function, generating high-level instructions by integrating goals, states, historical data, and external models. Information flow processed by the middle computer:

[0085] C (k) high,i : High-level instructions output by the upper computer (such as "start full-process monitoring"); C (k) high : The i-th sub-instruction (such as "initialize sensors", "set acquisition frequency"); Information flow executed by the lower computer:

[0086] U (k) i : Execution actions of the lower-level machine k (such as motor drive signals, valve openings).

[0087] C (k) low,i : Low-level instructions assigned by the middle-level machine to the lower-level machine i (such as "read data of sensor S001").

[0088] S (k) low,i : Real-time status or sensor data of the lower-level machine i (such as current motor speed, temperature value).

[0089] Data feedback and update:

[0090] D (k) mid : Integrated middle-level data (such as structured feature tables, anomaly detection results).

[0091] D (k) low,n : Original data of the lower-level machine n at time k (such as original sensor sampling points).

[0092] In one or more embodiments, a control method for a high-risk operation robot imitating the human nervous system is based on the high-risk operation robot control system imitating the human nervous system as described above; it includes: The sensing element senses the robot state information and transmits it to the corresponding lower-level machine; the lower-level machine transmits the received robot state information to the upper-level machine through the middle-level machine; the upper-level machine forms command information according to the preset goals of high-risk operations, combines the received robot state information, and issues it to the corresponding lower-level machine through the middle-level machine; the lower-level machine controls the corresponding moving parts to execute the corresponding command information, and at the same time transmits the received robot state information to the upper-level machine through the middle-level machine, forming a perception-decision-execution closed-loop control; Among them, between the lower-level machine and the middle-level machine, and between the middle-level machine and the upper-level machine, communication is carried out using a general instruction format; the general instruction format is: [frame header][control flag][instruction type][data area][check code][frame tail].

[0093] The hardware selection of the present invention takes into account both computing power and real-time performance: Computing power balance: The upper-level machine uses Raspberry Pi 4B to provide high general computing power to support interaction with large language models; the middle-level machine uses STM32F407 for lightweight computing to avoid wasting upper-level machine resources; the lower-level machine uses STM32F103 to achieve high real-time motion control. The combination of the three achieves the optimal balance among cost, computing power, and real-time performance.

[0094] Cost controllable: The hardware selection is based on mature embedded platforms (such as STM32 series, Raspberry Pi), which can significantly reduce the development and deployment costs while ensuring performance.

[0095] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A high-risk operation robot control system that imitates the human nervous system, characterized in that: include: Sensor element, lower computer, middle computer and upper computer; The sensor element is used to sense the robot status information and transmit it to the corresponding lower computer; The lower computer is used to transmit the robot status information it receives to the upper computer via the intermediate computer; the upper computer is used to form command information based on the preset target of the high-risk operation and the received robot status information, and send it to the corresponding lower computer via the intermediate computer; the lower computer is also used to control the corresponding moving parts to execute the corresponding command information, and at the same time, feed back the robot status information it receives to the upper computer via the intermediate computer, forming a perception-decision-execution closed-loop control; Among them, the lower computer and the middle computer, as well as the middle computer and the upper computer, all communicate in a general instruction format; the general instruction format is: [frame header][control flag][instruction type][data area][check code][frame tail].

2. The high-risk operation robot control system imitating the human nervous system as claimed in claim 1, characterized in that: The data area is composed of multiple data blocks, and the format of each data block is: [action][target][value]; the value is variable-length encoded according to its range and type.

3. The high-risk operation robot control system that imitates the human nervous system as claimed in claim 2, characterized in that: In the data block, if the value range is ≤255, 1 byte is used; if the value range is ≤65535, 2 bytes are used; if it is a floating point number, 4 bytes are used.

4. The high-risk operation robot control system imitating the human nervous system as claimed in claim 1, characterized in that: In the data block, the action is 1 byte, the upper 2 bits indicate the action type, and the lower 6 bits are the action ID; Or / and In the data block, the target is 1 byte, which is used to represent the target device or joint number, and its range is 0~255.

5. The high-risk operation robot control system imitating the human nervous system as claimed in claim 1, characterized in that: The model of the host computer is: ； in, It is the host computer instruction; S high Perception data for direct interaction with the host computer; API LLM It is the result of interaction with the large language model API; S MLC The returned data integrated by the intermediate machine.

6. The high-risk operation robot control system that imitates the human nervous system as claimed in claim 1, characterized in that: The models of the middle machine include: instruction decomposition model, data integration model and lightweight computing model; The instruction decomposition model is: ; The data integration model is: ; The lightweight calculation model is: ; in, C is the host computer instruction; high,i is the decomposed host computer sub-instruction; F is the instruction decomposition function; D mid For middle-level data; S MLC The returned data integrated by the intermediate machine; D low,i It is the data of the lower computer; G is the feedback data of the lower computer after lightweight processing; MCC is the integration function; It is a lightweight calculation function.

7. The high-risk operation robot control system imitating the human nervous system as claimed in claim 1, characterized in that: The lower computer model includes: control signal generation model and data preprocessing model; The control signal generation model is: ; Data preprocessing model: ; Among them, C low,i S is the lower computer instruction; low,i is the sensor data; U i is the control signal; D low,i Preprocess data for the lower computer; Generate a function for the control signal; It is a data preprocessing function.

8. The high-risk operation robot control system that imitates the human nervous system as claimed in claim 1, characterized in that: The command of the upper computer has a higher priority than the command of the middle computer; the command actually executed by the lower computer Determined by the following logic: ； in, It is the command of the host computer; It is the middle machine instruction.

9. The high-risk operation robot control system that imitates the human nervous system as claimed in claim 1, characterized in that: The delay of the information transmission process between the lower computer and the middle computer and between the middle computer and the upper computer is optimized by using the information flow delay model: The information flow delay model from the host computer to the middle computer is: ; The information flow delay model from the middle computer to the lower computer is: ; The information flow delay model from lower computer to middle computer is: ; The information flow delay model from the middle computer to the upper computer is: ; Among them, T high→mid T is the delay of the upper computer sending instructions to the middle computer; mid→low T is the delay of the middle computer sending instructions to the lower computer; low→mid T is the delay of sending data from the lower computer to the middle computer; mid→high C is the delay of the intermediate computer transmitting data to the upper computer; low,i It is the instruction of the lower computer; D is the upper instruction; low,i It is the data of the lower computer; It is the host computer data.

10. A control method for a high-risk operation robot that imitates the human nervous system, characterized in that: A high-risk operation robot control system based on the imitation of the human nervous system according to any one of claims 1 to 9; comprising: The sensor element senses the robot status information and transmits it to the corresponding lower computer; the lower computer transmits the received robot status information to the upper computer via the intermediate computer; the upper computer forms command information based on the preset goals of high-risk operations and the received robot status information, and sends it to the corresponding lower computer via the intermediate computer; the lower computer controls the corresponding moving parts to execute the corresponding command information, and at the same time feeds back the received robot status information to the upper computer via the intermediate computer, forming a closed-loop control of perception-decision-execution; Among them, the communication between the lower computer and the middle computer, as well as between the middle computer and the upper computer, adopts a general instruction format; the general instruction format is: [frame header][control flag][instruction type][data area][check code][frame tail].

Images