Robot control method and system combining fuzzy control and neural network

Abstract

The embodiments of the present invention relate to the field of robot control technology, and more specifically, to a robot control method and system that combines fuzzy control and neural networks. The method includes: obtaining a multi-dimensional state data set of a robot in a target operation scenario, performing fuzzy control feature extraction processing on the multi-dimensional state data set, calling a pre-trained neural network control model to perform hierarchical decision processing on the fuzzy control feature set, generating a real-time control instruction set for the robot, adjusting the operating state of the robot in the target operation scenario based on the real-time control instruction set, and feeding back the adjusted operating state to the multi-dimensional state data set to perform cyclic control processing. In this way, more precise, intelligent, and dynamic control of the robot can be achieved, effectively solving the problems of inaccurate control and poor adaptability of the existing technology in complex scenarios.

Description

Technical Field

[0001] The present invention relates to the field of robot control technology, and in particular to a robot control method and system combining fuzzy control and neural network. Background Art

[0002] In the field of robotic control, as robotic application scenarios become increasingly complex and diverse, higher requirements are placed on the intelligent control and adaptability of robots. Traditional robotic control technologies often focus only on single-dimensional data information or have poor adaptability to complex environments and changing tasks.

[0003] Existing robot control solutions, when faced with complex and ever-changing operational scenarios, often rely solely on a single type of data to make decisions. These solutions fail to fully consider the various factors affecting the robot's operation, resulting in inaccurate decisions and difficulty coping with the challenges posed by complex environments. Furthermore, existing technologies lack effective mechanisms for handling uncertain and fuzzy data, making it difficult to make accurate control decisions when encountering ambiguous or uncertain data. Furthermore, traditional control methods struggle to dynamically adjust control strategies based on real-time changes, making it prone to operational instability and low task execution efficiency during robot execution. Summary of the Invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, one of the objectives of the present invention is to provide a robot control method and system combining fuzzy control and neural network.

[0005] An embodiment of the present invention provides a robot control method that combines fuzzy control and neural networks, including: obtaining a multidimensional state data set of the robot in a target operation scenario, the multidimensional state data set including device posture data, environmental interference data and task constraints; performing fuzzy control feature extraction processing on the multidimensional state data set to obtain a fuzzy control feature set of the robot in the current operation cycle, the fuzzy control feature set including membership distribution features and fuzzy rule matching features; calling a pre-trained neural network control model to perform hierarchical decision processing on the fuzzy control feature set to generate a real-time control instruction set for the robot, the real-time control instruction set including motion trajectory adjustment instructions and device state compensation instructions; adjusting the operating state of the robot in the target operation scenario based on the real-time control instruction set, and feeding back the adjusted operating state to the multidimensional state data set to execute cyclic control processing.

[0006] An embodiment of the present invention also provides a robot control system, comprising a processor and a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-mentioned robot control method combining fuzzy control and neural network.

[0007] An embodiment of the present invention further provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the above-mentioned robot control method combining fuzzy control and neural network.

[0008] By applying the embodiments of the present invention, a multi-dimensional state data set of the robot in the target operation scenario (covering various information such as device posture, environmental interference, and task constraints) is first obtained, laying the foundation for a comprehensive understanding of the robot's operating status. Then, fuzzy control feature extraction processing is performed on the multi-dimensional state data set. The generated fuzzy control feature set includes membership distribution features and fuzzy rule matching features, which can accurately characterize data characteristics with fuzzy logic and effectively handle the uncertainty and ambiguity in the data. Then, a pre-trained neural network control model is called to perform hierarchical decision processing. Combined with the nonlinear processing capabilities of the neural network, a real-time control instruction set including motion trajectory adjustment instructions and device state compensation instructions can be generated to achieve precise decision-making. Finally, based on the real-time control instruction set, the robot's operating state is adjusted and feedback is provided to form a dynamic optimization mechanism, which enables the robot to continuously adapt to environmental changes and task requirements, significantly improving the robot's operating stability, task execution accuracy, and environmental adaptability in complex scenarios, and ensuring the efficient and reliable completion of various tasks.

[0009] In summary, the embodiments of the present invention can achieve more precise, intelligent and dynamic control of the robot by acquiring a multi-dimensional state data set, performing fuzzy control feature extraction, and combining the hierarchical decision processing and feedback adjustment mechanism of the neural network control model, effectively solving the problems of inaccurate control and poor adaptability of existing technologies in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 This is a flow chart of a robot control method combining fuzzy control and neural network provided by an embodiment of the present invention.

[0012] Figure 2 A block diagram of a robot control system provided by an embodiment of the present invention.

[0013] icon:

[0014] 100-Robot control system;

[0015] 101 - processor; 102 - memory; 103 - bus. DETAILED DESCRIPTION

[0016] The exemplary embodiments disclosed herein will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0017] In order to better understand the above technical solution, the technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0018] Figure 1 This is a flow chart of a robot control method combining fuzzy control and neural network according to an embodiment of the present invention, which is applied to a robot control system and includes steps 110 to 140.

[0019] Step 110: Acquire a multi-dimensional state data set of the robot in the target operation scenario, wherein the multi-dimensional state data set includes device posture data, environmental interference data, and task constraint conditions.

[0020] In an embodiment of the present invention, a robot is used in an industrial warehouse to perform cargo handling tasks. In this scenario, acquiring a multi-dimensional state data set is crucial for the robot's stable operation and task completion. Device posture data reflects the robot's own state, environmental interference data allows the robot to perceive its surroundings, and task constraints provide a clear task guide for the robot's actions. By comprehensively acquiring this data, the robot can better adapt to environmental changes, make accurate decisions, and ensure the efficient and safe completion of cargo handling tasks.

[0021] In one embodiment, obtaining a multi-dimensional state data set of the robot in the target operation scenario includes:

[0022] Step 111: Collecting the device attitude data through the inertial measurement unit of the robot, wherein the device attitude data includes three-dimensional acceleration, angular velocity and inclination.

[0023] In industrial warehousing scenarios, robots move and move goods in a variety of ways, and their postures constantly change. Inertial measurement units collect this critical data in real time. For example, when a robot travels in a straight line along a warehouse aisle, three-dimensional acceleration data can reflect its acceleration and deceleration. During a turn, angular velocity data can indicate the speed of the turn. And inclination data can be used to determine whether the robot is in a balanced state. For example, changes in inclination when traveling up or down a slope allow the robot to adjust its posture in a timely manner to prevent cargo from falling. This data is crucial for accurately assessing the robot's status and implementing appropriate control measures.

[0024] Step 112: Collecting the environmental interference data through the environmental sensor of the robot, wherein the environmental interference data includes the ground friction coefficient, the obstacle distribution density and the air resistance coefficient.

[0025] In industrial warehouse environments, factors such as ground conditions, obstacle distribution, and air conditions can interfere with a robot's operation. Environmental sensors can keenly capture this information. For example, different areas have different ground materials, resulting in different ground friction coefficients. A robot experiences different driving resistance on smooth and rough surfaces, which affects its power output and trajectory. The obstacle density reflects the distribution of obstacles such as goods and shelves in the warehouse. The robot needs to plan a safe driving path based on this data to avoid collisions. Although the air resistance coefficient is relatively small, it cannot be ignored when the robot is traveling at high speeds, as it affects the robot's energy consumption and speed control.

[0026] Step 113: Obtain the task constraints through the task scheduling unit of the robot, where the task constraints include target path priority, task execution time limit, and safety obstacle avoidance threshold.

[0027] In cargo handling tasks in industrial warehousing, the task scheduling unit plays a key command role. The target path priority clarifies the execution order of different path segments in conflict scenarios. For example, if multiple robots are simultaneously performing tasks in a warehouse and paths intersect, the robot on the path segment with the higher priority will have priority to ensure efficient execution of the overall task. The task execution time limit constrains the maximum time it takes for a robot to complete a task, thereby enabling the rational allocation of warehouse resources and improving work efficiency. The safety obstacle avoidance threshold determines the minimum allowable distance between the robot and obstacles, ensuring the robot's safety during operation and preventing collisions that could damage equipment and cause cargo loss.

[0028] In another embodiment, obtaining the task constraint condition by the task scheduling unit of the robot includes:

[0029] Step 1131: Receive the initial task instruction sent by the external control terminal, and parse the initial task instruction to obtain task path planning data and task execution condition data.

[0030] In industrial warehousing scenarios, the external control terminal is responsible for issuing initial task instructions. When a new cargo handling task is requested, the control terminal sends detailed instructions to the robot's task scheduling unit. The task scheduling unit then interprets the instructions. For example, the instructions may include a task to move cargo from warehouse area A to area B. The task path planning data will specify the robot's specific route from area A to area B, including information such as the passageways and turning points it passes through. The task execution condition data will specify the time required to complete the task, the cargo weight limit, and other conditions, providing a basis for subsequently determining the task constraints.

[0031] Step 1132: Determine the target path priority based on the task path planning data, where the target path priority is used to indicate the execution order of different path segments in a conflict scenario.

[0032] In complex industrial warehouse environments, multiple robots operating simultaneously can cause path conflicts. Based on the task path planning data, the task scheduling unit considers various factors to determine the target path priority. For example, if a path leads to an urgent delivery area, the path segment on this path will have a higher priority. If multiple robots cross paths during operation, the robot on the path segment with the higher priority will pass first, while other robots will need to wait or re-plan their paths to ensure the efficient execution of the overall task.

[0033] Step 1133: Determine the task execution time limit and the safety obstacle avoidance threshold based on the task execution condition data. The task execution time limit is used to constrain the maximum operating time of the robot, and the safety obstacle avoidance threshold is used to determine the minimum allowable distance between the robot and the obstacle.

[0034] It is understandable that the task execution condition data contains a lot of important information, and the task scheduling unit will extract key content from it. Regarding the time limit for task execution, for example, the task of transporting goods from a specific area to a designated location is required to be completed within one hour. This is to ensure the efficient operation of the entire warehousing process and avoid task delays that affect subsequent work. The determination of the safe obstacle avoidance threshold is based on factors such as the type and distribution of obstacles in the warehouse. For example, for large shelves, the safe obstacle avoidance threshold may be set larger to prevent the robot from colliding with the shelves during operation, causing goods to fall or equipment to be damaged; for small obstacles, the safe obstacle avoidance threshold can be relatively small, but a sufficient safety distance must also be ensured.

[0035] Step 1134: Associating the target path priority, the task execution time limit, and the safety obstacle avoidance threshold as the task constraint conditions.

[0036] Optionally, the task scheduling unit integrates target path priority, task execution time limit, and safety obstacle avoidance threshold to form a complete set of task constraints. These constraints are interrelated and mutually influential, providing clear guidance for robot operation. In industrial warehousing scenarios, robots need to meet these constraints simultaneously when performing tasks. They must complete tasks within the specified time and plan their routes reasonably according to path priorities while ensuring safety.

[0037] Step 120: Perform fuzzy control feature extraction processing on the multi-dimensional state data set to obtain a fuzzy control feature set of the robot in the current operation cycle, wherein the fuzzy control feature set includes a membership distribution feature and a fuzzy rule matching feature.

[0038] In industrial warehousing scenarios, multidimensional state data sets contain a vast amount of complex and uncertain information. Fuzzy control feature extraction and processing of this data can transform it into a more valuable, easier-to-understand, and process fuzzy control feature set. Membership distribution features can help robots better understand their own state and the degree of environmental factors, while fuzzy rule matching features enable robots to make informed decisions based on historical experience and current task requirements.

[0039] Optionally, performing fuzzy control feature extraction processing on the multi-dimensional state data set to obtain a fuzzy control feature set of the robot in the current operation cycle includes:

[0040] Step 121: performing membership function mapping processing on the device posture data to generate a first membership distribution feature, where the first membership distribution feature includes a deviation membership of the device posture data relative to a preset standard posture interval.

[0041] In industrial warehousing scenarios, robots' postures constantly change while handling goods. Preset standard posture ranges are set based on the robot's design and normal operating requirements. For example, when a robot is traveling in a straight line, its standard posture is to remain horizontal and its direction of travel stable. Membership function mapping processing is performed on the device posture data, comparing the actual collected device posture data, such as three-dimensional acceleration, angular velocity, and inclination angle, with the preset standard posture range. Taking inclination angle as an example, if the robot's inclination angle is within the standard range, its deviation membership relative to the standard posture range may be low; if the inclination angle exceeds a certain range, the deviation membership becomes medium or high. The first membership distribution feature generated in this way allows the robot to clearly understand the degree of deviation of its own posture from the standard posture.

[0042] Step 122: Perform fuzzy interval division processing on the environmental interference data to generate a second membership distribution feature, where the second membership distribution feature includes distribution weights of the environmental interference data in multiple interference intensity intervals.

[0043] In industrial storage environments, environmental interference factors are complex and diverse. The fuzzy interval division processing of environmental interference data is to divide it into multiple interference intensity intervals according to the characteristics and influence of different interference factors. For example, for the ground friction coefficient, it can be divided into low friction, medium friction and high friction intervals; for the obstacle distribution density, it can be divided into low density, medium density and high density intervals. Then, based on the actual collected environmental interference data, its distribution weight in each interference intensity interval is determined. For example, when the ground friction coefficient in the area where the robot is located is likely to be in the medium friction interval, the distribution weight of this interval will be higher. The second membership distribution feature generated by this can give the robot a more intuitive understanding of the environmental interference situation.

[0044] Step 123: Perform fuzzy rule matching processing on the task constraint conditions to generate the fuzzy rule matching feature, which includes the matching degree between the current task constraint conditions and each rule template in the historical task rule library.

[0045] In industrial warehousing scenarios, the historical task rule library has accumulated a wealth of experience and rules from past task execution. Fuzzy rule matching of task constraints involves comparing the current task constraints, such as target path priority, task execution time limit, and safety obstacle avoidance threshold, with the various rule templates in the historical task rule library. For example, if the current task requires cargo transportation to be completed in a short period of time and the path passes through an area with many obstacles, fuzzy rule matching can be used to find similar task rule templates in historical tasks and calculate the degree of match between the task rule templates. A high degree of match indicates that the current task is similar to the historical tasks, allowing for decision-making based on historical experience.

[0046] Step 124: Fusing the first membership distribution feature, the second membership distribution feature, and the fuzzy rule matching feature to obtain the fuzzy control feature set.

[0047] In industrial warehousing scenarios, the first membership distribution feature reflects the robot's posture, the second membership distribution feature reflects environmental disturbances, and the fuzzy rule matching feature provides a basis for decision-making based on historical experience. Fusion of these three features aims to integrate information from different perspectives. For example, a weighted fusion approach can be used, assigning different weights to each feature based on the actual situation. If the current environmental disturbance has a significant impact on the robot's operation, the weight of the second membership distribution feature can be set higher. This fusion approach results in a more comprehensive and integrated set of fuzzy control features, providing a better foundation for subsequent decision-making.

[0048] Step 130: Calling a pre-trained neural network control model to perform hierarchical decision processing on the fuzzy control feature set to generate a real-time control instruction set for the robot, wherein the real-time control instruction set includes motion trajectory adjustment instructions and device state compensation instructions.

[0049] In industrial warehousing scenarios, pre-trained neural network control models can be used for intelligent decision-making. The fuzzy control feature set contains a large amount of complex information. By invoking this model for hierarchical decision-making, this information can be converted into a specific, executable set of real-time control instructions. Motion trajectory adjustment instructions allow the robot to adjust its route based on the environment and its own state, while device state compensation instructions ensure that the robot's device posture and power output are in optimal conditions to better complete cargo handling tasks.

[0050] Preferably, the calling of a pre-trained neural network control model to perform hierarchical decision processing on the fuzzy control feature set to generate a real-time control instruction set for the robot includes:

[0051] Step 131: Input the fuzzy control feature set into the feature coding layer of the neural network control model to perform feature space conversion processing to generate a mapping feature vector.

[0052] In industrial warehousing scenarios, the feature encoding layer of a neural network control model can serve as an information converter. The fuzzy control feature set contains complex, multi-dimensional information, which the feature encoding layer processes. For example, the feature encoding layer integrates and transforms different types of features, such as first membership distribution features, second membership distribution features, and fuzzy rule matching features. These features are mapped from their original feature space to another feature space more suitable for neural network processing, generating a mapped feature vector. This mapped feature vector contains the recoded and integrated information, providing a better representation for subsequent decision-making.

[0053] Step 132: Calling the first decision layer of the neural network control model to perform motion trajectory prediction processing on the mapping feature vector to generate an initial trajectory adjustment instruction, wherein the first decision layer predicts trajectory deviation based on historical trajectory data and current environment data.

[0054] In industrial warehousing scenarios, the first decision layer performs trajectory prediction based on the mapped feature vectors, historical trajectory data, and current environmental data. For example, historical trajectory data can provide past experience with the robot's travel routes in similar environments and tasks, while current environmental data includes information such as obstacle distribution and ground conditions. The first decision layer analyzes this data to predict potential trajectory deviations. If a new obstacle is detected ahead, or if a change in the ground friction coefficient could cause the robot to shift its trajectory, it generates initial trajectory adjustment instructions to guide the robot to adjust its route in advance.

[0055] Step 133: Call the second decision layer of the neural network control model to perform device state compensation prediction processing on the mapping feature vector to generate an initial state compensation instruction. The second decision layer predicts state stability based on the device historical state data and current interference data.

[0056] In industrial warehousing scenarios, the second decision layer predicts and processes device state compensation based on the mapped feature vectors, historical device state data, and current interference data. Historical device state data records the robot's past posture, power output, and other conditions. Current interference data, such as ground friction and air resistance, can affect device state. The second decision layer analyzes this data to predict device state stability. If it detects that current environmental interference may cause the robot's center of gravity to shift or power output to become unstable, it generates initial state compensation instructions to adjust the robot's posture and power output.

[0057] Step 134: normalize the initial trajectory adjustment instruction and the initial state compensation instruction to generate a real-time control instruction set including the motion trajectory adjustment instruction and the device state compensation instruction.

[0058] In industrial warehousing scenarios, initial trajectory adjustment instructions and initial state compensation instructions may have different dimensions and value ranges. Normalization unifies these instructions so that they have the same scale and range. For example, a normalization algorithm maps the initial trajectory adjustment instructions and initial state compensation instructions to a standard interval. The resulting real-time control instruction set includes calibrated and integrated trajectory adjustment instructions and device state compensation instructions, which can more accurately and effectively guide robot operation.

[0059] In an alternative embodiment, the training process of the pre-trained neural network control model includes:

[0060] Step 210: Acquire a historical operation data set, where the historical operation data set includes equipment operation data, environmental data, and corresponding control instruction annotation data for multiple historical operation cycles.

[0061] During the long-term operation of industrial warehousing scenarios, a large amount of historical operation data has been accumulated. This data covers multiple historical operation cycles. The equipment operation data records the robot's posture, speed, and other information at different times. Environmental data includes ground conditions, obstacle distribution, and other conditions. The corresponding control instruction annotation data clearly defines the correct control instructions to be taken under these data conditions. For example, in a past operation cycle, the robot's equipment posture, speed, and corresponding control instructions were recorded when moving goods from point P1 to point P2 under a specific ground friction coefficient and obstacle distribution environment. This data provides rich material for training neural network control models.

[0062] Step 220: Perform fuzzy feature extraction processing on the historical operation data set to obtain a historical fuzzy control feature set.

[0063] Just as with the current multi-dimensional state data set, fuzzy feature extraction is also required for the historical operation data set. For device operation data, membership function mapping is used to generate membership distribution features related to device posture. Fuzzy interval partitioning is performed on environmental data to obtain membership distribution features for environmental interference data. Fuzzy rule matching is performed between control instruction annotation data and the historical task rule base to generate fuzzy rule matching features. These features are then fused to obtain a historical fuzzy control feature set, which encompasses the fuzzy control information in the historical data and provides suitable input for model training.

[0064] Step 230: Obtain an initial neural network model, where the initial neural network model includes an input layer, at least two hidden layers, and an output layer.

[0065] As you can understand, the initial neural network model is a basic architecture. The input layer receives external data, which in this scenario is the historical fuzzy control feature set. At least two hidden layers perform complex nonlinear transformations and feature extraction on the input data, enabling them to learn underlying patterns within the data. The output layer outputs predicted control instructions based on the hidden layer's processing results. For example, after receiving the historical fuzzy control feature set, the input layer passes it to the hidden layer. The hidden layer analyzes and processes the data through connections between neurons and weight adjustments. Finally, the output layer outputs the predicted control instruction set.

[0066] Step 240: Input the historical fuzzy control feature set into the initial neural network model for forward propagation processing to generate a predictive control instruction set.

[0067] In an industrial warehousing scenario, after a set of historical fuzzy control features is input into the initial neural network model, the data is forward propagated according to the model's structure. The input layer passes the data to the first hidden layer. The neurons in this hidden layer calculate and transform the data based on preset weights and activation functions, and then pass the results to the next hidden layer. After processing through multiple hidden layers, the output layer finally generates a set of predictive control instructions.

[0068] Step 250: Generate a loss function based on the difference between the predicted control instruction set and the control instruction labeled data, and update the weight parameters of the initial neural network model through the back propagation algorithm until the loss function converges to obtain the pre-trained neural network control model.

[0069] In industrial warehousing scenarios, the discrepancy between the predicted control instruction set and the labeled control instruction data reflects the accuracy of the model's predictions. An algorithm is set to generate a loss function, whose value represents the difference between the predicted and actual results. For example, algorithms such as mean squared error can be used to calculate the loss function. Then, using a backpropagation algorithm, starting from the output layer, the loss value is propagated back through the network structure, adjusting the weight parameters accordingly. This process is repeated until the loss function converges. At this point, the initial neural network model becomes a pretrained neural network control model, capable of more accurately processing input data and making decisions.

[0070] As an implementation manner, generating a loss function based on the difference between the predicted control instruction set and the control instruction labeled data includes:

[0071] Step 251: performing mean square error calculation on the motion trajectory adjustment instruction and the marked motion trajectory adjustment instruction in the prediction control instruction set to obtain a first loss component.

[0072] In industrial warehousing scenarios, mean squared error (MSE) calculation is a common method for measuring the difference between two trajectory adjustment instructions. The MSE algorithm calculates the sum of the squares of the differences between the predicted and labeled trajectory adjustment instructions in each dimension and then takes the average. For example, in a two-dimensional plane, a trajectory adjustment instruction may involve adjustments in the x- and y-directions. The squares of the differences between the predicted and labeled values ​​in these two directions are calculated, added, and then divided by the number of dimensions. The result is the first loss component, which reflects the prediction error in the trajectory adjustment.

[0073] Step 252: performing cosine similarity calculation on the device state compensation instruction in the predictive control instruction set and the labeled device state compensation instruction to obtain a second loss component.

[0074] In industrial warehousing scenarios, equipment state compensation instructions involve information from multiple dimensions, including the robot's posture compensation and power output compensation. Cosine similarity is used to measure the similarity between the predicted and labeled equipment state compensation instructions in vector space. The equipment state compensation instructions are considered vectors, and the cosine of the angle between the two vectors is calculated. For example, the equipment state compensation instruction vector may contain an angle vector for equipment posture adjustment and a power vector for power output adjustment. The cosine similarity value is calculated by taking the dot product of the two vectors and dividing it by the corresponding module length product. Subtracting this cosine similarity value from 1 yields the second loss component. This component reflects the degree of difference between the predicted equipment state compensation results and the true annotations, providing a basis for model adjustment.

[0075] Step 253: performing absolute value sum calculation on the overall deviation between the predicted control instruction set and the control instruction labeling data to obtain a third loss component.

[0076] In the industrial warehousing scenario, considering the overall deviation between the predicted control instruction set and the control instruction annotation data, the absolute value summation calculation is used to quantify this deviation. The predicted control instruction set contains multiple parts such as motion trajectory adjustment instructions and equipment state compensation instructions, and the control instruction annotation data also covers the corresponding accurate information. Take the absolute value of the difference between each instruction element in the predicted control instruction set and the corresponding element in the control instruction annotation data, and then add all these absolute values. The sum obtained is the third loss component. For example, the adjustment values ​​of the motion trajectory adjustment instructions in the x, y, and z directions and the compensation values ​​of the posture and power output in the equipment state compensation instructions, etc., are all calculated and accumulated. The absolute value of the difference comprehensively reflects the degree of deviation between the overall prediction result and the annotation data.

[0077] Step 254: Perform weighted summation on the first loss component, the second loss component, and the third loss component to obtain the loss function.

[0078] In industrial warehousing scenarios, different loss components may have different importance for model training, so a weighted summation is required to obtain the loss function. Different weights are assigned to the first, second, and third loss components. For example, based on practical experience and analysis of the importance of the task, if the accuracy of motion trajectory adjustment is critical for the robot to complete the task, the weight of the first loss component can be set higher; if the accuracy of equipment state compensation is relatively less important, the weight of the second loss component can be appropriately reduced; the third loss component reflects the overall deviation, and its weight is also set based on the overall situation. Then, each loss component is multiplied by its corresponding weight and added together, that is, the first loss component multiplied by weight 1, the second loss component multiplied by weight 2, and the third loss component multiplied by weight 3. The final loss function is obtained. This loss function integrates prediction error information from different aspects and can more comprehensively reflect the difference between the model prediction and the actual situation, providing an accurate basis for subsequent updating of the model weight parameters through the backpropagation algorithm.

[0079] Step 140: Adjusting the operating state of the robot in the target operating scenario based on the real-time control instruction set, and feeding back the adjusted operating state to the multi-dimensional state data set to perform cyclic control processing.

[0080] In industrial warehousing scenarios, real-time control instructions are the key basis for robots to adjust their operating states. Based on these instructions, robots can accurately adjust their motion and equipment status to adapt to changing environments and task requirements. Simultaneously, these adjusted operating states are fed back into multi-dimensional state data sets, forming a closed-loop control process that enables robots to continuously optimize their operations and better complete cargo handling tasks.

[0081] In one embodiment, adjusting the operating state of the robot in the target operating scenario based on the real-time control instruction set includes:

[0082] Step 141: parsing the motion trajectory adjustment instructions in the real-time control instruction set, determining the trajectory correction amount of the robot within the target time window, and adjusting the driving wheel speed and steering angle of the robot according to the trajectory correction amount.

[0083] In industrial warehousing scenarios, trajectory adjustment commands are used to ensure the robot can accurately reach the target location according to mission requirements. Upon receiving a trajectory adjustment command from the real-time control command set, the robot interprets it. For example, a command might indicate that within the next 10 seconds (the target time window), the robot needs to shift a certain distance to the left to avoid a new obstacle ahead. By interpreting the command, the robot determines the trajectory correction, which may include information such as lateral displacement and angular change. The robot then adjusts the drive wheel speed and steering angle based on this trajectory correction. If a left turn is required, the left drive wheel speed is appropriately reduced, while the right drive wheel speed remains unchanged or increases, thereby achieving the desired turn. Simultaneously, the overall drive wheel speed is adjusted based on the overall speed requirement and trajectory correction requirements to ensure the robot accurately reaches the target location within the specified time.

[0084] Step 142: parse the device state compensation instructions in the real-time control instruction set, determine the device posture compensation amount and power output compensation amount of the robot, adjust the center of gravity distribution of the robot according to the device posture compensation amount, and adjust the motor output power of the robot according to the power output compensation amount.

[0085] In industrial warehousing scenarios, device state compensation commands are used to ensure that the robot's equipment operates in optimal condition. After parsing the device state compensation commands, the robot's device posture compensation and power output compensation are determined. For example, when a robot is carrying heavy cargo, the shift in center of gravity may cause instability. Device posture compensation may require adjustments to the robot's mechanical structure to redistribute the center of gravity to a more stable position. Power output compensation is determined based on factors such as cargo weight and ground friction. If ground friction increases, motor output power needs to be increased to maintain the robot's normal speed; otherwise, it needs to be reduced appropriately. By adjusting the center of gravity distribution and motor output power, the robot can operate stably under varying operating conditions.

[0086] Step 143: Collect the adjusted driving wheel speed, steering angle, center of gravity distribution and motor output power to generate the adjusted operating state.

[0087] In industrial warehousing scenarios, after a robot adjusts its drive wheel speed, steering angle, center of gravity distribution, and motor output power, it needs to collect these parameters. Sensors installed in key areas of the robot capture this information in real time. For example, a speed sensor is installed on the drive wheel to measure speed, an angle sensor to obtain steering angle, a gravity sensor to sense center of gravity distribution, and a power sensor to collect motor output power. This collected data is integrated to form the adjusted operating status. This operating status data reflects the robot's actual operating status after executing real-time control commands, providing accurate data support for subsequent feedback and loop control.

[0088] In the next step, the adjusted operating state is fed back to the multi-dimensional state data set to perform a loop control process, including:

[0089] Step 144: writing the adjusted driving wheel speed and steering angle in the operating state into the multi-dimensional state data set as new device posture data.

[0090] In industrial warehousing scenarios, drive wheel speed and steering angle are crucial components of the device's posture. Writing these two adjusted parameters as new device posture data into the multi-dimensional state data set allows for timely updates of the robot's posture information. For example, after adjusting the robot's operating state, the drive wheel speed may be adjusted from 100 rpm to 120 rpm, and the steering angle may be changed from 0 degrees to 15 degrees to the left. Writing these new data into the device posture data portion of the multi-dimensional state data set allows the robot to perform subsequent processing and decision-making based on the latest device posture information the next time it retrieves the multi-dimensional state data set.

[0091] Step 145: writing the center of gravity distribution and motor output power in the adjusted operating state into the multi-dimensional state data set as new device state compensation data.

[0092] In industrial warehousing scenarios, the center of gravity distribution and motor output power reflect the robot's device state compensation. Writing the adjusted center of gravity distribution and motor output power as new device state compensation data into the multi-dimensional state data set allows the robot to promptly understand changes in its device state. For example, through adjustments, the robot's center of gravity position may be moved from an original coordinate point to a new position, and the motor output power may be adjusted from an original value to a new value. By updating this new data into the multi-dimensional state data set, the robot can then use this latest device state compensation data to better assess its state and make control decisions during subsequent operation.

[0093] Step 146: re-execute the fuzzy control feature extraction process and the hierarchical decision process based on the updated multi-dimensional state data set to generate a real-time control instruction set for the next operation cycle to implement cyclic control.

[0094] In industrial warehousing scenarios, the updated multi-dimensional state data set contains the robot's latest operating status information. Based on this updated data, the fuzzy control feature extraction process is re-executed to process the device posture data, environmental interference data, and task constraints to generate a new fuzzy control feature set. For example, new device posture data may cause changes in the first membership distribution feature, and updates to environmental interference data will also affect the second membership distribution feature. The new fuzzy control feature set is then input into the pre-trained neural network control model for hierarchical decision processing to generate a real-time control instruction set for the next operating cycle. This process is repeated continuously, allowing the robot to continuously adjust its operating state according to changes in the environment and its own state to better complete the cargo handling task and achieve efficient and stable cyclic control.

[0095] As a non-limiting embodiment, after the adjusted operating state is fed back to the multi-dimensional state data set to perform cyclic control processing, it also includes: generating fuzzy rule optimization parameters based on the correspondence between the adjusted operating state and the real-time control instruction set, the fuzzy rule optimization parameters include rule weight correction coefficients and membership function offsets; performing real-time update processing on the rule templates in the historical task rule library according to the fuzzy rule optimization parameters to generate an updated rule template set, the real-time update processing includes adjusting the rule trigger threshold and reallocating the center point coordinates of the membership function; applying the updated rule template set to the fuzzy rule matching processing of the next operating cycle to optimize the generation accuracy of the fuzzy control feature set.

[0096] In industrial warehousing scenarios, there's an inherent correspondence between adjusted operating states and sets of real-time control instructions. By analyzing this correspondence, fuzzy rule optimization parameters can be generated. For example, if a particular fuzzy rule performs well in one operating state but poorly in another, the rule weight correction coefficient can be adjusted accordingly, adjusting the weight of that rule to make it more influential in the appropriate scenario. Regarding the membership function offset, if the actual situation doesn't quite match the preset membership function, the offset can be adjusted to ensure the membership function better reflects the actual situation.

[0097] Based on the generated fuzzy rules, the parameters are optimized and the rule templates in the historical task rule library are updated in real time. For example, the rule trigger threshold is adjusted. If a rule triggers too frequently or too infrequently during the current operation, the trigger threshold is appropriately raised or lowered to better meet actual needs. The center point coordinates of the membership function are reallocated to enable the membership function to more accurately describe the relationship between data and concepts. After these processes, an updated set of rule templates is generated.

[0098] In the following operation cycle, the updated rule template set is applied to the fuzzy rule matching process. When the robot encounters new task constraints, it uses the updated rule templates for matching, enabling more accurate calculation of the matching degree and thus optimizing the accuracy of the generated fuzzy control feature set. For example, in a new task, the updated rule templates can more accurately determine the similarity between the current task constraints and historical experience, providing a more reliable basis for subsequent decision-making.

[0099] As another non-limiting embodiment, after the adjusted operating state is fed back to the multi-dimensional state data set to perform cyclic control processing, it also includes: extracting the device posture compensation and trajectory correction in the adjusted operating state to generate a derived training data set; performing spatiotemporal alignment processing on the derived training data set and the corresponding fuzzy control feature set to generate time-series associated training samples; performing online gradient update on the pre-trained neural network control model based on the time-series associated training samples, and adjusting the weight distribution of the feature coding layer and the activation function parameters of the first decision layer.

[0100] In industrial warehousing scenarios, the adjusted operating state contains a wealth of information. The device posture compensation and trajectory correction reflect the actual adjustments made to the robot during operation. This data is extracted to generate a derived training data set. For example, the device posture compensation may include adjustments to the robot's center of gravity, angle, and other aspects at different times, while the trajectory correction records adjustments to the robot's route at various stages.

[0101] The derived training data set is spatiotemporally aligned with the corresponding fuzzy control feature set. This means matching the fuzzy control feature set acquired at the same time or in a related time series with the derived training data. For example, at a specific moment, the robot made a decision based on the fuzzy control feature set, resulting in a certain amount of device posture compensation and trajectory correction. This data is then aligned in chronological order to generate time-series correlation training samples.

[0102] Based on these time-correlated training samples, the pre-trained neural network control model undergoes an online gradient update. Online gradient updating is a method for continuously adjusting parameters during model execution. By calculating the gradient of the time-correlated training samples, the weight distribution of the feature encoding layer is adjusted according to the gradient direction. For example, if a weight parameter is found to cause a large error when processing the current sample, the weight is adjusted based on the gradient information to more accurately process the data. Simultaneously, the activation function parameters of the first decision layer are adjusted. The activation function determines the output of the neurons. By adjusting these parameters, neurons can better respond to inputs in different situations, improving the model's decision accuracy and adaptability.

[0103] As another non-limiting embodiment, after the adjusted operating state is fed back to the multi-dimensional state data set to perform cyclic control processing, it also includes: performing sliding window statistical analysis on the fuzzy control feature set of N consecutive operating cycles to extract the mean vector and variance matrix of each feature; dynamically normalizing the fuzzy control feature set according to the mean vector and variance matrix to generate a standardized fuzzy feature vector; inputting the standardized fuzzy feature vector into the neural network control model to perform feature attention allocation between decision layers, and adjusting the priority weights of the motion trajectory adjustment instruction and the device state compensation instruction.

[0104] In industrial warehousing scenarios, to better analyze the changing trends and stability of the fuzzy control feature set, a sliding window statistical analysis is performed on the fuzzy control feature set for N consecutive operating cycles. For example, the sliding window size is set to 5 operating cycles, and the fuzzy control feature set within the window is statistically analyzed by moving one cycle at a time. The mean vector of each feature is extracted, and the mean vector reflects the average level of each feature over these N cycles. For example, for a dimension in the first membership distribution feature, its average value over N cycles is calculated to form an element of the mean vector. At the same time, the variance matrix is ​​calculated. The variance matrix can measure the fluctuation of each feature over these N cycles and reflect the degree of discreteness of the data.

[0105] Based on the calculated mean vector and variance matrix, the fuzzy control feature set is dynamically normalized. Dynamic normalization ensures that different features have comparable scales across different operating cycles. For example, for a particular feature, the value in the mean vector is subtracted from the feature's value, and then divided by the standard deviation of the corresponding dimension in the variance matrix to obtain the normalized value, thus generating a standardized fuzzy feature vector.

[0106] The standardized fuzzy feature vector is input into the neural network control model to allocate feature attention between decision layers. Within the model, different decision layers may pay varying degrees of attention to different features. By analyzing the standardized fuzzy feature vector, the system determines which features are more important for trajectory adjustment instructions and device state compensation instructions. For example, if a feature is found to be highly correlated with the trajectory, the attention weight of that feature in the first decision layer (responsible for trajectory prediction) is appropriately increased, giving the model greater attention to it when generating trajectory adjustment instructions. Features related to device state compensation are weighted accordingly in the second decision layer, adjusting the priority of trajectory adjustment instructions and device state compensation instructions. This allows the robot to more rationally allocate decision resources based on actual conditions, improving control effectiveness.

[0107] In an embodiment of the present invention, the following key processing can also be performed based on the existing fuzzy control theory, neural network modeling technology and industrial robot closed-loop control method: for the fuzzy feature extraction link, conventional membership functions (such as triangle and trapezoidal functions) can be used in combination with domain knowledge to preset standard posture intervals and interference intensity intervals, and the deviation membership and distribution weights can be fully generated by calculating the membership relationship between the actual data and the preset intervals; for fuzzy rule matching, similarity calculations (such as Euclidean distance and fuzzy proximity) can be used based on the historical task database to achieve quantitative evaluation of the rule template matching degree.

[0108] Optionally, at the neural network decision layer, standard feature encoding techniques (such as fully connected layers and convolution operations) can be used to transform feature space. Time series prediction models (such as LSTM and GRU) can be used to integrate historical trajectory / state data to generate initial adjustment instructions. When constructing the loss function, conventional evaluation methods such as mean squared error (for motion trajectory), cosine similarity (for multidimensional device state vectors), and absolute deviation can be combined, with weighting coefficients determined empirically or through grid search. Dynamic normalization in closed-loop control can employ sliding window Z-score normalization, and feature attention allocation can incorporate attention mechanisms (such as additive attention) to automatically learn decision-layer weights.

[0109] Furthermore, for online rule base updates, state-command correlations can be leveraged based on reinforcement learning to dynamically modify rule weights and membership function centers. For online model fine-tuning, incremental data after time-series alignment is combined with stochastic gradient descent to optimize encoding layer weights and activation function parameters. This effectively achieves further optimization through feature fuzzification, decision stratification, dynamic adaptation, and dimensional unification.

[0110] By applying the embodiments of the present invention, a multi-dimensional state data set of the robot in the target operation scenario (covering various information such as device posture, environmental interference, and task constraints) is first obtained, laying the foundation for a comprehensive understanding of the robot's operating status. Then, fuzzy control feature extraction processing is performed on the multi-dimensional state data set. The generated fuzzy control feature set includes membership distribution features and fuzzy rule matching features, which can accurately characterize data characteristics with fuzzy logic and effectively handle the uncertainty and ambiguity in the data. Then, a pre-trained neural network control model is called to perform hierarchical decision processing. Combined with the nonlinear processing capabilities of the neural network, a real-time control instruction set including motion trajectory adjustment instructions and device state compensation instructions can be generated to achieve precise decision-making. Finally, based on the real-time control instruction set, the robot's operating state is adjusted and feedback is provided to form a dynamic optimization mechanism, which enables the robot to continuously adapt to environmental changes and task requirements, significantly improving the robot's operating stability, task execution accuracy, and environmental adaptability in complex scenarios, and ensuring the efficient and reliable completion of various tasks.

[0111] In summary, the embodiments of the present invention can achieve more precise, intelligent and dynamic control of the robot by acquiring a multi-dimensional state data set, performing fuzzy control feature extraction, and combining the hierarchical decision processing and feedback adjustment mechanism of the neural network control model, effectively solving the problems of inaccurate control and poor adaptability of existing technologies in complex scenarios.

[0112] An embodiment of the present invention provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the robot control method combining fuzzy control and neural network is implemented.

[0113] An embodiment of the present invention provides a processor, which is used to run a program, wherein the robot control method combining fuzzy control and neural network is executed when the program is run.

[0114] In the embodiment of the present invention, Figure 2 As shown, the robot control system 100 includes at least one processor 101, and at least one memory 102 and a bus 103 connected to the processor 101; wherein the processor 101 and the memory 102 communicate with each other through the bus 103; the processor 101 is used to call the program instructions in the memory 102 to execute the above-mentioned robot control method combining fuzzy control and neural network.

[0115] The present invention is described with reference to flowcharts and / or block diagrams of methods, robot control systems (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0116] In a typical configuration, a robot control system includes one or more processors (CPUs), memory, and a bus. The robot control system may also include input / output interfaces, network interfaces, and the like.

[0117] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip. Memory is an example of a computer-readable medium.

[0118] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology for information storage. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage, computer-readable storage media, or any other non-transmission media that can be used to store information that can be accessed by the robot control system. As defined herein, computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0119] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or computer-readable storage medium that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, commodity, or computer-readable storage medium. In the absence of further limitations, an element defined by the phrase "comprises a..." does not preclude the presence of additional identical elements in the process, method, commodity, or computer-readable storage medium that includes the element.

[0120] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0121] The above are merely embodiments of the present invention and are not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A robot control method combining fuzzy control and neural network, characterized in that: include: Acquire a multi-dimensional state data set of the robot in a target operation scenario, wherein the multi-dimensional state data set includes device posture data, environmental interference data, and task constraints; Performing fuzzy control feature extraction processing on the multi-dimensional state data set to obtain a fuzzy control feature set of the robot in the current operation cycle, the fuzzy control feature set including a membership distribution feature and a fuzzy rule matching feature; Calling a pre-trained neural network control model to perform hierarchical decision processing on the fuzzy control feature set to generate a real-time control instruction set for the robot, the real-time control instruction set including motion trajectory adjustment instructions and device state compensation instructions; Adjusting the operating state of the robot in the target operating scenario based on the real-time control instruction set, and feeding back the adjusted operating state to the multi-dimensional state data set to perform a cyclic control process; The performing fuzzy control feature extraction processing on the multi-dimensional state data set to obtain the fuzzy control feature set of the robot in the current operation cycle includes: Performing membership function mapping processing on the device posture data to generate a first membership distribution feature, where the first membership distribution feature includes a deviation membership of the device posture data relative to a preset standard posture interval; Performing fuzzy interval division processing on the environmental interference data to generate a second membership distribution feature, where the second membership distribution feature includes distribution weights of the environmental interference data in multiple interference intensity intervals; Performing fuzzy rule matching processing on the task constraint conditions to generate the fuzzy rule matching feature, wherein the fuzzy rule matching feature includes the matching degree between the current task constraint conditions and each rule template in the historical task rule library; Performing feature fusion on the first membership distribution feature, the second membership distribution feature, and the fuzzy rule matching feature to obtain the fuzzy control feature set; The calling of the pre-trained neural network control model to perform hierarchical decision processing on the fuzzy control feature set to generate a real-time control instruction set for the robot includes: Inputting the fuzzy control feature set into the feature coding layer of the neural network control model for feature space conversion processing to generate a mapping feature vector; Calling a first decision layer of the neural network control model to perform motion trajectory prediction processing on the mapped feature vector to generate an initial trajectory adjustment instruction, wherein the first decision layer predicts trajectory deviation based on historical trajectory data and current environment data; Calling the second decision layer of the neural network control model to perform device state compensation prediction processing on the mapping feature vector to generate an initial state compensation instruction, wherein the second decision layer performs state stability prediction based on historical device state data and current interference data; The initial trajectory adjustment instruction and the initial state compensation instruction are normalized to generate a real-time control instruction set including a motion trajectory adjustment instruction and a device state compensation instruction.

2. The robot control method combining fuzzy control and neural network according to claim 1, characterized in that: The training process of the pre-trained neural network control model includes: Acquire a historical operation data set, wherein the historical operation data set includes equipment operation data, environmental data, and corresponding control instruction annotation data for multiple historical operation periods; Performing fuzzy feature extraction processing on the historical operation data set to obtain a historical fuzzy control feature set; Obtaining an initial neural network model, wherein the initial neural network model includes an input layer, at least two hidden layers, and an output layer; Inputting the historical fuzzy control feature set into the initial neural network model for forward propagation processing to generate a predictive control instruction set; A loss function is generated based on the difference between the predicted control instruction set and the control instruction labeled data, and the weight parameters of the initial neural network model are updated through a back propagation algorithm until the loss function converges to obtain the pre-trained neural network control model.

3. The robot control method combining fuzzy control and neural network according to claim 2, characterized in that: Generating a loss function based on a difference between the predicted control instruction set and the control instruction labeled data includes: performing mean square error calculation on the motion trajectory adjustment instruction and the marked motion trajectory adjustment instruction in the prediction control instruction set to obtain a first loss component; performing cosine similarity calculation on the device state compensation instruction in the prediction control instruction set and the labeled device state compensation instruction to obtain a second loss component; performing a summation calculation of absolute values ​​of overall deviations between the predicted control instruction set and the control instruction labeling data to obtain a third loss component; The first loss component, the second loss component and the third loss component are weightedly summed to obtain the loss function.

4. The robot control method combining fuzzy control and neural network according to claim 1, characterized in that: The adjusting the operating state of the robot in the target operating scenario based on the real-time control instruction set includes: parsing the motion trajectory adjustment instructions in the real-time control instruction set, determining a trajectory correction amount for the robot within a target time window, and adjusting a driving wheel speed and a steering angle of the robot according to the trajectory correction amount; parsing a device state compensation instruction in the real-time control instruction set, determining a device posture compensation amount and a power output compensation amount of the robot, adjusting a center of gravity distribution of the robot according to the device posture compensation amount, and adjusting a motor output power of the robot according to the power output compensation amount; The adjusted driving wheel speed, steering angle, center of gravity distribution and motor output power are collected to generate the adjusted operating state.

5. The robot control method combining fuzzy control and neural network according to claim 4, characterized in that: Feeding back the adjusted operating state to the multi-dimensional state data set to perform a loop control process includes: Writing the adjusted driving wheel speed and steering angle in the operating state into the multi-dimensional state data set as new device posture data; Writing the adjusted center of gravity distribution and motor output power in the operating state into the multi-dimensional state data set as new device state compensation data; The fuzzy control feature extraction process and the hierarchical decision process are re-executed based on the updated multi-dimensional state data set to generate a real-time control instruction set for the next operation cycle to implement cyclic control.

6. The robot control method combining fuzzy control and neural network according to claim 1, characterized in that: The step of obtaining a multi-dimensional state data set of the robot in the target operation scenario includes: Collecting the device attitude data through the inertial measurement unit of the robot, the device attitude data including three-dimensional acceleration, angular velocity and inclination; Collecting the environmental interference data through the environmental sensor of the robot, wherein the environmental interference data includes ground friction coefficient, obstacle distribution density and air resistance coefficient; Obtaining the task constraints through the task scheduling unit of the robot, wherein the task constraints include target path priority, task execution time limit, and safety obstacle avoidance threshold; The obtaining of the task constraint condition by the task scheduling unit of the robot includes: Receiving an initial task instruction sent by an external control terminal, parsing the initial task instruction to obtain task path planning data and task execution condition data; determining the target path priority based on the task path planning data, wherein the target path priority is used to indicate the execution order of different path segments in a conflict scenario; determining the task execution time limit and the safe obstacle avoidance threshold based on the task execution condition data, wherein the task execution time limit is used to constrain the maximum operating time of the robot, and the safe obstacle avoidance threshold is used to determine the minimum allowable distance between the robot and the obstacle; The target path priority, the task execution time limit and the safety obstacle avoidance threshold are associated as the task constraint conditions.

7. A robot control system, characterized in that: It comprises a processor and a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other via the bus; the processor is used to call program instructions in the memory to execute the robot control method combining fuzzy control and neural network as described in any one of claims 1-6.

8. A computer-readable storage medium, characterized in that A program is stored thereon, and when the program is executed by a processor, the robot control method combining fuzzy control and neural network as described in any one of claims 1 to 6 is implemented.

Images