Intelligent robot control method based on multi-sensor cooperation

Abstract

The invention discloses an intelligent robot control method based on multi-sensor cooperation, and relates to the field of intelligent robots, and the method comprises the following operation steps: S1, carrying out the architecture deployment of sensors; s2, establishing communication; s3, edge intelligent calculation enhancement; s4, a dynamic adaptive fusion strategy; s5, carrying out group intelligent decision making; and S6, carrying out sensor role dynamic conversion. According to the intelligent robot control method based on multi-sensor cooperation, the sensor data credibility can be adjusted in real time according to the environment condition, the fusion weight can be dynamically allocated, the data accuracy is ensured through the cross check of the adjacent sensor data, and meanwhile, the intelligent robot control method based on multi-sensor cooperation can be realized through the ant colony cooperation algorithm in combination with the self-adaptive pheromone volatilization regulation and control. The path exploration strategy can be flexibly adjusted according to task complexity and environment dynamic change, the search efficiency can be effectively improved, various environments can be more quickly adapted, and the optimal action path can be quickly found.

Description

Technical Field

[0001] The present invention relates to the field of intelligent robots, and in particular to an intelligent robot control method based on multi-sensor collaboration. Background Art

[0002] Intelligent robots have a variety of internal and external information sensors, such as vision, hearing, touch, and smell. In addition to receptors, they also have effectors as a means of acting on the surrounding environment. These are muscles, or self-synchronizing motors, which make the hands, feet, long noses, tentacles, etc. move. It can be seen from this that intelligent robots must have at least three elements: sensory elements, reaction elements, and thinking elements.

[0003] When working, traditional intelligent robots mostly adopt a decentralized layout of single-function sensors, and each sensor works independently. When faced with auxiliary environmental changes, the response speed will be affected, which will also affect the control decisions of the intelligent robot. At the same time, when making control decisions, intelligent robots are mostly based on fixed rules and simple machine learning models, and lack the ability to dynamically adapt to complex environments.

[0004] Therefore, it is necessary to propose an intelligent robot control method based on multi-sensor collaboration to solve the above problems. Summary of the invention

[0005] The main purpose of the present invention is to provide an intelligent robot control method based on multi-sensor collaboration, which can effectively solve the problems in the background technology.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] An intelligent robot control method based on multi-sensor collaboration includes the following steps:

[0008] S1: Deployment of sensor architecture, building sensor modules, the sensor modules are used to receive data collected by sensors, the sensors are selected from visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors, the visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors are designed with bionic honeycomb mechanical structures, and there are at least two of each of the visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors;

[0009] S2: Communication construction, optimizing the data transmission of the intelligent robot, designing the priority of the control instructions of the intelligent robot, finely dividing the communication frequency band, and opening up exclusive channels for data of different priorities according to self-set instructions;

[0010] S3: Enhanced edge intelligent computing: The intelligent robot performs online learning based on data collected by visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors, and adjusts deep learning model parameters online based on real-time collected data.

[0011] S4: Dynamic adaptive fusion strategy, based on the set initial sensor credibility weight as the basis for subsequent dynamic adjustment of the intelligent robot, and real-time update of the credibility of the current environment, so that the intelligent robot can make full use of sensor data to maintain accurate and reliable perception capabilities;

[0012] S5: Group intelligent decision-making, using ant algorithm to assist the control decision-making of intelligent robots;

[0013] S6: Dynamic conversion of sensor roles. Based on the comprehensive capability values ​​of sensors, when the intelligent robot performs tasks, sensors with high capability values ​​are selected as key sensors, and the remaining sensors are used for auxiliary functions, cooperating with each other to assist the intelligent robot in control.

[0014] Preferably, in S1, the visual sensor integrates infrared and ultraviolet imaging to provide accurate visual information for the intelligent robot; the distance detection sensor includes a lidar and a multi-angle ultrasonic sensor, and the distance detection sensor is used to perform distance detection and environmental mapping of the action position of the intelligent robot; the environmental monitoring sensor includes a gas sensor array and a temperature, humidity and dust sensor, and the gas sensor array includes a broad-spectrum gas detection based on metal oxide semiconductors, an electrochemical sensor, and a photoionization detector, and the environmental monitoring sensor is used to judge the quality of the surrounding environment of the intelligent robot; the tactile sensor is used to sense and monitor the tactile sense of the intelligent machine, and the intelligent robot performs corresponding drive control based on the data monitored by the visual sensor, the distance detection sensor, the environmental monitoring sensor, and the tactile sensor.

[0015] Preferably, S3 includes the following steps:

[0016] S301: Real-time data learning. During operation, the visual sensor, distance detection sensor, environmental monitoring sensor, and tactile sensor continuously collect new perception data. Each time a new data is collected, it is compared with the historical data. The incremental learning algorithm based on contrastive learning is used to quickly identify new objects, scenes, and environmental change features in the data. When unknown features are found, the small sample learning process is immediately started. By performing multiple iterative training on limited local data, the target recognition model is updated, so that the intelligent robot can identify new obstacles, tools, operation objects, scenes, and environments in a short time, which is used to improve the ability to adapt to unfamiliar environments.

[0017] S302: Dynamic adjustment of model parameters: The distance detection sensor includes point cloud processing algorithm parameters. The distance detection sensor optimizes the point cloud processing algorithm parameters in real time according to different environmental conditions. When the intelligent robot enters a complex geometric environment, it automatically increases the density threshold in the point cloud clustering algorithm to finely distinguish overlapping obstacles at close range. In open areas, the threshold is appropriately lowered to speed up processing and reduce misjudgments. Through the policy gradient algorithm, the model parameters are continuously adjusted according to environmental feedback, including the number of collisions and navigation efficiency, to ensure that accurate distance information can be output in various scenarios to assist the intelligent robot in decision-making.

[0018] Preferably, the step S4 includes the following steps:

[0019] S401: Initialization of data credibility assessment. During the startup phase of the intelligent robot, each visual sensor, distance detection sensor, environmental monitoring sensor, and tactile sensor sets an initial credibility weight for each of its own sensors based on historical experience data and preset sensor performance parameters, and stores the initial credibility weight as the basis for subsequent dynamic adjustments.

[0020] S402: Real-time environmental monitoring and credibility update. The environmental sensor collects the surrounding environmental data of the intelligent robot, and updates the credibility under the influence of fog and dust, the credibility under the influence of light, and the credibility under the combined influence of temperature and humidity. The credibility update under the influence of fog and dust includes:

[0021] Credibility update under the influence of light includes: setting a moving average window with a length of 5 sets of data, i.e., data within the past 25 seconds, calculating the average dust and fog concentration in the window, and if the average value exceeds 1 mg / m 3 , it is determined that the air quality has a potential impact on vision; if the concentration continues to rise and exceeds 3mg / m 3 If it lasts for more than 10 seconds, it is judged as a high pollution environment, and the credibility of the visual data needs to be adjusted immediately. The credibility update model based on exponential decay is adopted, and the initial credibility of the visual data is set to C 0 , the concentration of fog and dust is d, and the calculation formula of the updated credibility C is:

[0022] C=C 0 ×e -k×d ;

[0023] Where k is the attenuation coefficient and e is a natural constant. When the calculated credibility C is less than the preset initial credibility C of the visual data, 0 When , it is judged that the credibility of the visual data sensor is low;

[0024] The credibility under the influence of light and the credibility under the combined influence of temperature and humidity include: data verification, the normal range of light intensity is set at 0lx to 100000lx, when the rate of change of light intensity exceeds 30%, that is, the difference between two adjacent acquisition values ​​divided by the current value is greater than 0.3, it is judged that the lighting conditions have changed significantly, for light intensity greater than 80000lx, record the duration of strong light, if it exceeds 5 seconds, mark it as a strong light interference environment; for light intensity less than 100lx, if it lasts for more than 10 seconds, mark it as a weak light environment, these marks are used for subsequent visual data credibility adjustment, for strong light interference environment, a linear decreasing model is used, when in a strong light interference environment, the credibility decreases by 0.05 per second, assuming that the duration of strong light is t, then the formula of credibility C is:

[0025] C=C 0 -0.05×t;

[0026] When the calculated credibility C is less than the preset initial credibility C of the visual data 0 When , it is judged that the credibility of the visual data sensor is low;

[0027] For low-light environments, the reliability decreases logarithmically. Assuming the light intensity in low-light environments is I, the calculation formula for the reliability C is:

[0028]

[0029] When the calculated credibility C is less than the preset initial credibility C of the visual data 0 When , it is judged that the credibility of the visual data sensor is low;

[0030] The credibility update under the influence of light and the credibility update under the combined influence of temperature and humidity include: each time a new set of temperature and humidity data is collected, a simple data check is first performed to determine whether the data is within a reasonable physical range, where the temperature is between -40℃ and 85℃, and the humidity is between 0% and 100%. If not, it is marked as abnormal data. For normal data, the difference between the current temperature and humidity and the average temperature and humidity in the previous 10 minutes is calculated. If the absolute value of the difference is greater than the set threshold, the temperature change threshold is set to 2℃, and the humidity change threshold is set to 5%. It is determined that the ambient temperature and humidity have changed significantly, triggering the subsequent credibility adjustment process for the visual sensor. Specifically, a two-dimensional lookup table is constructed, with the row index being the temperature range, one interval of every 5℃, from -20℃ to 40℃; the column index being the humidity range, one interval of every 10%RH, from 0% to 100%; each element in the table stores a credibility adjustment factor f, where the element is a cell with a row index of the temperature range and a column index of the humidity range. When the current temperature and humidity are monitored, the corresponding adjustment factor f is obtained according to the lookup table. The calculation formula for the updated credibility C is:

[0031] C=C 0 ×(1-f);

[0032] When the calculated credibility C is less than the preset initial credibility C of the visual data 0 When , it is judged that the credibility of the visual data sensor is low;

[0033] S403: Dynamic allocation of fusion weights: Based on the real-time updated credibility, when the visual sensor, the distance detection sensor, the environmental monitoring sensor, and the tactile sensor are performing data fusion, a weighted average method is used to allocate fusion weights to different sensor data;

[0034] S404: Regular data synchronization comparison: The data synchronization mechanism is started every 100 milliseconds between adjacent sensors. Vision sensors transmit key features of the currently acquired image, including object contours and color histograms, to each other through wireless communication; distance detection sensors share distance measurement data and preliminary point cloud clustering results; tactile sensors communicate contact force distribution and vibration frequency;

[0035] S405: Difference analysis and correction: When two adjacent visual sensors compare image features, if the difference in the contour of the same object is found to exceed the preset threshold, the parameters of each image acquisition are immediately traced back, including the shooting angle, lens focal length, and illumination compensation value. The image registration algorithm is used to perform geometric transformation and grayscale adjustment on the image according to the difference in these parameters to realign the features and correct the deviation. If the deviation is caused by uneven illumination, the illumination compensation parameters are adjusted and the image is re-acquired to ensure the consistency of the visual data.

[0036] S406: Collaborative error correction feedback: In the working operation scenario of the intelligent robot, the distance detection sensor and the tactile sensor work together. Once the tactile sensor detects a sudden change in the grasping force that exceeds the normal grasping force fluctuation range, it immediately sends a feedback signal to the distance detection sensor. After receiving the signal, the distance detection sensor quickly performs a second distance measurement, recalculates the grasping target position and posture, and compares it with the previous distance measurement results. If deviations are found, the intelligent robot's actions are adjusted in time to correct the grasping position.

[0037] Preferably, in S402, the credibility adjustment process includes: once the environmental monitoring sensor detects that the fog concentration exceeds the standard, the light changes suddenly, or the temperature and humidity fluctuate greatly, it immediately sends an interrupt signal to the visual sensor. After receiving the notification, the visual sensor suspends the current operation and calls the corresponding credibility update algorithm. If fog or dust is detected, the real-time collected fog and dust concentration values ​​are substituted into the exponential decay formula to quickly calculate the new credibility value. Then, the visual sensor propagates the updated credibility value to the distance detection sensor and the environmental monitoring sensor, so as to increase the weight of its own data in the subsequent data fusion link, and participate in the fusion of sensor data according to the new weight distribution rule. At the same time, the current environmental status is marked, and the cause and magnitude of this credibility adjustment are recorded and stored in the log.

[0038] Preferably, the step S5 includes the following steps:

[0039] S501: Initially, pheromones are evenly distributed in the entire task space of the intelligent robot, and the data of each sensor randomly selects the direction of travel for exploration. As the number of explorations increases, pheromones gradually accumulate on the paths leading to successful results, and subsequent sensors tend to select these paths with high concentrations of pheromones.

[0040] S502: Setting the pheromone to volatilize 10% every fixed time period is used to gradually reduce the pheromone of old and inefficient paths, prompting the sensor to re-explore possible new paths, maintaining adaptability to environmental changes, and preventing excessive accumulation of pheromones from causing path exploration to fall into a local optimum;

[0041] When the intelligent robot completes a small phased goal or the entire task, it will give additional pheromone rewards to the sensors on the successful path;

[0042] S503: The control direction of the intelligent robot is updated by guiding pheromones through sensor data. The visual sensor captures the environment image in real time. Once a new potential target area is found, the visual signal is broadcast to the surrounding sensors to guide the nearby sensors to explore in that direction. The concentration of pheromones is pre-set on the initial exploration path to speed up the exploration of new areas.

[0043] Among the distance detection sensors, the laser radar and ultrasonic sensors continuously monitor surrounding obstacles. If they detect a new obstacle or narrow passage ahead, they quickly reduce the pheromone concentration on the current path and send warning signals to other sensors, prompting them to avoid dangerous areas and reselect a safe path with a relatively high pheromone concentration, thus preventing the intelligent robot from being damaged by collision.

[0044] S504: Task priority control adjustment based on pheromone concentration. The intelligent robot periodically scans the pheromone concentration on each path and associates the most urgent task to be completed by the intelligent robot with the path with high pheromone concentration.

[0045] Paths where pheromone concentrations have been low for a long time and have not significantly promoted the mission are determined to be inefficient paths. The intelligent robot suspends sensor activity on the path, re-analyzes the collected data, and attempts to optimize sensor parameters and adjust action strategies. If multiple optimizations are unsuccessful, the path exploration is completely abandoned, and sensors are reallocated to other potential areas to rationally utilize resources and improve overall mission effectiveness.

[0046] Preferably, S6 also includes designing a real-time capability evaluation system: each cellular sensor calculates a comprehensive capability value based on its current hardware performance, data quality, and past task completion performance. During the task execution process, once new task requirements or emergency situations arise, sensors with high capability values ​​quickly take over key roles and cooperate with other sensors to perform tasks.

[0047] Compared with the prior art, the present invention provides an intelligent robot control method based on multi-sensor collaboration, which has the following beneficial effects:

[0048] 1. This intelligent robot control method based on multi-sensor collaboration can adjust the credibility of sensor data in real time according to environmental conditions, dynamically allocate fusion weights, and ensure data accuracy through cross-checking of adjacent sensor data. At the same time, through the ant colony collaboration algorithm combined with adaptive pheromone volatilization regulation, it can flexibly adjust the path exploration strategy according to the complexity of the task and the dynamic changes of the environment, which can effectively improve the search efficiency, adapt to various environments more quickly, and quickly find the optimal action path.

[0049] 2. The intelligent robot control method based on multi-sensor collaboration can learn data in real time and compare the data collected during operation with previous data, so as to quickly identify the features in the data. Based on this, the intelligent robot can quickly identify new objects encountered, and can effectively improve the intelligent robot's ability to adapt to unfamiliar and complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0051] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0052] like Figure 1As shown, a multi-sensor collaborative intelligent robot control method includes the following steps:

[0053] S1: Architecture deployment of sensors. Build sensor modules. The sensor modules are used to receive data collected by sensors. The sensors include visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors. The visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors are designed with bionic honeycomb mechanical structures. There are at least two visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors.

[0054] Visual sensors integrate infrared and ultraviolet imaging to provide accurate visual information for intelligent robots; distance detection sensors include lidar and multi-angle ultrasonic sensors, and are used to perform distance detection and environmental mapping of the intelligent robot's action position; environmental monitoring sensors include gas sensor arrays and temperature, humidity and dust sensors, and the gas sensor array includes metal oxide semiconductor-based broad-spectrum gas detection, electrochemical sensors, and photoionization detectors, and environmental monitoring sensors are used to judge the quality of the surrounding environment of the intelligent robot; tactile sensors are used to sense and monitor the tactile sense of intelligent machines, and the intelligent robot performs corresponding drive control based on the data monitored by visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors.

[0055] S2: Communication construction, optimize the data transmission of the intelligent robot, design the priority of the control instructions of the intelligent robot, divide the communication frequency band finely, and open up exclusive channels for data of different priorities according to self-set instructions.

[0056] S3: Edge intelligent computing is enhanced. The intelligent robot performs online learning through data collected by visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors. The deep learning model parameters are adjusted online based on the real-time collected data. The specific operation steps include the following:

[0057] S301: Real-time data learning. During operation, the visual sensor, distance detection sensor, environmental monitoring sensor, and tactile sensor continuously collect new perception data. Each time a new data is collected, it is compared with the historical data. The incremental learning algorithm based on contrastive learning is used to quickly identify new objects, scenes, and environmental change features in the data. When unknown features are found, the small sample learning process is immediately started. By performing multiple iterative training on limited local data, the target recognition model is updated, so that the intelligent robot can identify new obstacles, tools, operation objects, scenes, and environments in a short time, which is used to improve the ability to adapt to unfamiliar environments.

[0058] S302: Dynamic adjustment of model parameters: The distance detection sensor includes point cloud processing algorithm parameters. The distance detection sensor optimizes the point cloud processing algorithm parameters in real time according to different environmental conditions. When the intelligent robot enters a complex geometric environment, it automatically increases the density threshold in the point cloud clustering algorithm to finely distinguish overlapping obstacles at close range. In open areas, the threshold is appropriately lowered to speed up processing and reduce misjudgments. Through the policy gradient algorithm, the model parameters are continuously adjusted according to environmental feedback, including the number of collisions and navigation efficiency, to ensure that accurate distance information can be output in various scenarios to assist the intelligent robot in decision-making.

[0059] S4: Dynamic adaptive fusion strategy, based on the set initial sensor credibility weight as the basis for subsequent intelligent robot dynamic adjustment, and real-time update of the credibility of the current environment, so that the intelligent robot can make full use of sensor data to maintain accurate and reliable perception capabilities, including the following steps:

[0060] S401: Data credibility assessment initialization. During the startup phase of the intelligent robot, each visual sensor, distance detection sensor, environmental monitoring sensor, and tactile sensor sets an initial credibility weight for each sensor based on historical experience data and preset sensor performance parameters. The initial credibility weight is stored as the basis for subsequent dynamic adjustment. Under normal lighting and clear vision, the visual sensor gives a higher initial credibility, such as 0.8, to the image data of the visible light camera. In areas with frequent foggy weather, the initial credibility of the lidar data is set to 0.7, and the credibility of the visual data is relatively lowered to 0.3, etc.

[0061] S402: Real-time environmental monitoring and credibility update. The environmental sensor collects the surrounding environmental data of the intelligent robot, and updates the credibility under the influence of fog and dust, the credibility under the influence of light, and the credibility under the combined influence of temperature and humidity. The credibility update under the influence of fog and dust includes:

[0062] Credibility update under the influence of light includes: setting a moving average window with a length of 5 sets of data, i.e., data within the past 25 seconds, calculating the average dust and fog concentration in the window, and if the average value exceeds 1 mg / m 3 , it is determined that the air quality has a potential impact on vision; if the concentration continues to rise and exceeds 3mg / m 3 If it lasts for more than 10 seconds, it is judged as a high pollution environment, and the credibility of the visual data needs to be adjusted immediately. The credibility update model based on exponential decay is adopted, and the initial credibility of the visual data is set to C 0 , the concentration of fog and dust is d, and the calculation formula of the updated credibility C is:

[0063] C=C 0 ×e -k×d;

[0064] Where k is the attenuation coefficient and e is a natural constant. When the calculated credibility C is less than the preset initial credibility C of the visual data, 0 When , it is judged that the credibility of the visual data sensor is low;

[0065] The credibility under the influence of light and the credibility under the combined influence of temperature and humidity include: data verification, the normal range of light intensity is set at 0lx to 100000lx, when the rate of change of light intensity exceeds 30%, that is, the difference between two adjacent acquisition values ​​divided by the current value is greater than 0.3, it is judged that the lighting conditions have changed significantly, for light intensity greater than 80000lx, record the duration of strong light, if it exceeds 5 seconds, mark it as a strong light interference environment; for light intensity less than 100lx, if it lasts for more than 10 seconds, mark it as a weak light environment, these marks are used for subsequent visual data credibility adjustment, for strong light interference environment, a linear decreasing model is used, when in a strong light interference environment, the credibility decreases by 0.05 per second, assuming that the duration of strong light is t, then the formula of credibility C is:

[0066] C=C 0 -0.05×t;

[0067] When the calculated credibility C is less than the preset initial credibility C of the visual data 0 When , it is judged that the credibility of the visual data sensor is low;

[0068] For low-light environments, the reliability decreases logarithmically. Assuming the light intensity in low-light environments is I, the calculation formula for the reliability C is:

[0069]

[0070] When the calculated credibility C is less than the preset initial credibility C of the visual data 0 When , it is judged that the credibility of the visual data sensor is low;

[0071] The credibility update under the influence of light and the credibility update under the combined influence of temperature and humidity include: each time a new set of temperature and humidity data is collected, a simple data check is first performed to determine whether the data is within a reasonable physical range, where the temperature is between -40℃ and 85℃, and the humidity is between 0% and 100%. If not, it is marked as abnormal data. For normal data, the difference between the current temperature and humidity and the average temperature and humidity in the previous 10 minutes is calculated. If the absolute value of the difference is greater than the set threshold, the temperature change threshold is set to 2℃, and the humidity change threshold is set to 5%. It is determined that the ambient temperature and humidity have changed significantly, triggering the subsequent credibility adjustment process for the visual sensor. Specifically, a two-dimensional lookup table is constructed, with the row index being the temperature range, one interval of every 5℃, from -20℃ to 40℃; the column index being the humidity range, one interval of every 10%RH, from 0% to 100%; each element in the table stores a credibility adjustment factor f, where the element is a cell with a row index of the temperature range and a column index of the humidity range. When the current temperature and humidity are monitored, the corresponding adjustment factor f is obtained according to the lookup table. The calculation formula for the updated credibility C is:

[0072] C=C 0 ×(1-f);

[0073] When the calculated credibility C is less than the preset initial credibility C of the visual data 0 When , it is judged that the credibility of the visual data sensor is low;

[0074] S403: Dynamic allocation of fusion weights: Based on the real-time updated credibility, when performing data fusion, visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors use the weighted average method to allocate fusion weights for different sensor data. Taking the construction of an environmental map as an example, when fusing visual features with lidar point cloud data, if the visual credibility is 0.4 and the lidar credibility is 0.6, then when generating the coordinate information of the map, the coordinate weight corresponding to the visual feature is 0.4, and the coordinate weight of the lidar point cloud is 0.6, ensuring that the final fusion result is closer to the real environment and effectively making up for the shortcomings of a single sensor;

[0075] S404: Regular data synchronization comparison: The data synchronization mechanism is started every 100 milliseconds between adjacent sensors. Vision sensors transmit key features of the currently acquired image, including object contours and color histograms, to each other through wireless communication; distance detection sensors share distance measurement data and preliminary point cloud clustering results; tactile sensors communicate contact force distribution and vibration frequency;

[0076] S405: Difference analysis and correction: When two adjacent visual sensors compare image features, if the difference in the contour of the same object is found to exceed the preset threshold, the parameters of each image acquisition are immediately traced back, including the shooting angle, lens focal length, and illumination compensation value. The image registration algorithm is used to perform geometric transformation and grayscale adjustment on the image according to the difference in these parameters to realign the features and correct the deviation. If the deviation is caused by uneven illumination, the illumination compensation parameters are adjusted and the image is re-acquired to ensure the consistency of the visual data.

[0077] S406: Collaborative error correction feedback: In the working operation scenario of the intelligent robot, the distance detection sensor and the tactile sensor work together. Once the tactile sensor detects a sudden change in the grasping force that exceeds the normal grasping force fluctuation range, it immediately sends a feedback signal to the distance detection sensor. After receiving the signal, the distance detection sensor quickly performs a second distance measurement, recalculates the grasping target position and posture, and compares the previous distance measurement results. If a deviation is found, the intelligent robot action is adjusted in time to correct the grasping position;

[0078] The credibility adjustment process includes: once the environmental monitoring sensor detects that the fog concentration exceeds the standard, the light changes suddenly, and the temperature and humidity fluctuate greatly, it immediately sends an interrupt signal to the visual sensor. After receiving the notification, the visual sensor suspends the current operation and calls the corresponding credibility update algorithm. If it is detected that it is affected by fog or dust, the real-time collected fog and dust concentration values ​​are substituted into the exponential decay formula to quickly calculate the new credibility value. Then, the visual sensor propagates the updated credibility value to the distance detection sensor and the environmental monitoring sensor, which is used to increase its own data weight in the subsequent data fusion link, and participate in the fusion of sensor data according to the new weight distribution rule. At the same time, the current environmental status is marked, the reason and magnitude of this credibility adjustment are recorded, and stored in the log. This not only facilitates subsequent troubleshooting, but also provides a data basis for the intelligent robot to continuously learn the relationship between environmental characteristics and sensor performance, so that it can respond more intelligently and quickly when encountering similar environmental changes next time, ensuring the accuracy of perception and decision-making in complex environments.

[0079] S5: Group intelligent decision-making, using ant algorithm to assist the control decision of intelligent robots, including the following steps:

[0080] S501: Initially, pheromones are evenly distributed in the entire task space of the intelligent robot, and the data of each sensor randomly selects the direction of travel for exploration. As the number of explorations increases, pheromones gradually accumulate on the paths leading to successful results, and subsequent sensors tend to select these paths with high concentrations of pheromones.

[0081] S502: Setting the pheromone to volatilize 10% every fixed time period is used to gradually reduce the pheromone of old and inefficient paths, prompting the sensor to re-explore possible new paths, maintaining adaptability to environmental changes, and preventing excessive accumulation of pheromones from causing path exploration to fall into a local optimum;

[0082] When the intelligent robot completes a small phased goal or the entire task, it will give additional pheromone rewards to the sensors on the successful path. This includes increasing the pheromone concentration of the search path by 50% after successfully locating the trapped person in a rescue scenario, encouraging subsequent sensors to give priority to this path strategy in similar scenarios, and continuously strengthening the dominant decision-making path.

[0083] S503: Guide pheromones through sensor data to update the control direction of the intelligent robot. The visual sensor captures the environment image in real time. Once a new potential target area is found, it broadcasts visual signals to surrounding sensors to guide nearby sensors to explore in that direction. It also pre-sets the concentration of pheromones on the initial exploration path to speed up the exploration of new areas. For example, in a warehouse search task, the visual sensor finds a closed compartment suspected of containing important materials. After receiving the signal, the surrounding sensors quickly move closer to the compartment and leave pheromones on the route to the compartment to improve the overall search efficiency.

[0084] Among the distance detection sensors, the laser radar and ultrasonic sensors continuously monitor surrounding obstacles. If they detect a new obstacle or narrow passage ahead, they quickly reduce the pheromone concentration on the current path and send warning signals to other sensors, prompting them to avoid dangerous areas and reselect a safe path with a relatively high pheromone concentration, thus preventing the intelligent robot from being damaged by collision.

[0085] S504: Task priority control adjustment based on pheromone concentration. The intelligent robot regularly scans the pheromone concentration on each path and associates the most urgent task for the intelligent robot to be completed with the path with high pheromone concentration. If a path with high pheromone concentration points to an area where people may be trapped, the intelligent robot will prioritize allocating its own resources to the sensors on the path to accelerate the completion of key tasks and ensure timely rescue.

[0086] Paths where pheromone concentrations have been low for a long time and have not significantly promoted the mission are determined to be inefficient paths. The intelligent robot suspends sensor activity on the path, re-analyzes the collected data, and attempts to optimize sensor parameters and adjust action strategies. If multiple optimizations are unsuccessful, the path exploration is completely abandoned, and sensors are reallocated to other potential areas to rationally utilize resources and improve overall mission effectiveness.

[0087] S6: Dynamic conversion of sensor roles. According to the comprehensive capability value of the sensor, when the intelligent robot performs a task, the sensor with a high capability value is selected as the key sensor, and the remaining sensors are used for auxiliary purposes, cooperating with each other to assist the intelligent robot in control. It also includes the design of a real-time capability evaluation system: each honeycomb sensor calculates the comprehensive capability value based on its current hardware performance, data quality, and past task completion performance. During the task execution process, once new task requirements or emergencies arise, sensors with high capability values ​​quickly take up key roles and cooperate with other sensors to perform tasks. For example, at the scene of a sudden fire, the environmental monitoring sensor detects a sharp increase in toxic gases. The sensor with the strongest gas analysis capability is mainly responsible for guiding the intelligent robot to avoid dangerous areas. At the same time, other sensors cooperate to ensure safe evacuation.

[0088] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An intelligent robot control method based on multi-sensor collaboration, characterized in that: The steps include: S1: Deployment of sensor architecture, building sensor modules, the sensor modules are used to receive data collected by sensors, the sensors are selected from visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors, the visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors are designed with bionic honeycomb mechanical structures, and there are at least two of each of the visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors; S2: Communication construction, optimizing the data transmission of the intelligent robot, designing the priority of the control instructions of the intelligent robot, finely dividing the communication frequency band, and opening up exclusive channels for data of different priorities according to self-set instructions; S3: Enhanced edge intelligent computing: The intelligent robot performs online learning based on data collected by visual sensors, distance detection sensors, environmental monitoring sensors, and tactile sensors, and adjusts deep learning model parameters online based on real-time collected data. S4: Dynamic adaptive fusion strategy, based on the set initial sensor credibility weight as the basis for subsequent dynamic adjustment of the intelligent robot, and real-time update of the credibility of the current environment, so that the intelligent robot can make full use of sensor data to maintain accurate and reliable perception capabilities; S5: Group intelligent decision-making, using ant algorithm to assist the control decision-making of intelligent robots; S6: Dynamic conversion of sensor roles. Based on the comprehensive capability values ​​of sensors, when the intelligent robot performs tasks, sensors with high capability values ​​are selected as key sensors, and the remaining sensors are used for auxiliary functions, cooperating with each other to assist the intelligent robot in control.

2. The intelligent robot control method based on multi-sensor collaboration according to claim 1 is characterized in that: In S1, the visual sensor integrates infrared and ultraviolet imaging to provide accurate visual information for the intelligent robot; the distance detection sensor includes a laser radar and a multi-angle ultrasonic sensor, and the distance detection sensor is used to perform distance detection and environmental mapping of the action position of the intelligent robot; the environmental monitoring sensor includes a gas sensor array and a temperature, humidity and dust sensor, and the gas sensor array includes a broad-spectrum gas detection based on metal oxide semiconductors, an electrochemical sensor, and a photoionization detector, and the environmental monitoring sensor is used to judge the quality of the surrounding environment of the intelligent robot; the tactile sensor is used to sense and monitor the tactile sense of the intelligent machine, and the intelligent robot performs corresponding drive control based on the data monitored by the visual sensor, the distance detection sensor, the environmental monitoring sensor, and the tactile sensor.

3. The intelligent robot control method based on multi-sensor collaboration according to claim 1 is characterized in that: The S3 includes the following steps: S301: Real-time data learning. During operation, the visual sensor, distance detection sensor, environmental monitoring sensor, and tactile sensor continuously collect new perception data. Each time a new data is collected, it is compared with the historical data. The incremental learning algorithm based on contrastive learning is used to quickly identify new objects, scenes, and environmental change features in the data. When unknown features are found, the small sample learning process is immediately started. By performing multiple iterative training on limited local data, the target recognition model is updated, so that the intelligent robot can identify new obstacles, tools, operation objects, scenes, and environments in a short time, which is used to improve the ability to adapt to unfamiliar environments. S302: Dynamic adjustment of model parameters: The distance detection sensor includes point cloud processing algorithm parameters. The distance detection sensor optimizes the point cloud processing algorithm parameters in real time according to different environmental conditions. When the intelligent robot enters a complex geometric environment, it automatically increases the density threshold in the point cloud clustering algorithm to finely distinguish overlapping obstacles at close range. In open areas, the threshold is appropriately lowered to speed up processing and reduce misjudgments. Through the policy gradient algorithm, the model parameters are continuously adjusted according to environmental feedback, including the number of collisions and navigation efficiency, to ensure that accurate distance information can be output in various scenarios to assist the intelligent robot in decision-making.

4. The intelligent robot control method based on multi-sensor collaboration according to claim 1 is characterized in that: The S4 includes the following steps: S401: Initialization of data credibility assessment. During the startup phase of the intelligent robot, each visual sensor, distance detection sensor, environmental monitoring sensor, and tactile sensor sets an initial credibility weight for each of its own sensors based on historical experience data and preset sensor performance parameters, and stores the initial credibility weight as the basis for subsequent dynamic adjustments. S402: Real-time environmental monitoring and credibility update. The environmental sensor collects the surrounding environmental data of the intelligent robot, and updates the credibility under the influence of fog and dust, the credibility under the influence of light, and the credibility under the combined influence of temperature and humidity. The credibility update under the influence of fog and dust includes: Credibility update under the influence of light includes: setting a moving average window with a length of 5 sets of data, i.e., data within the past 25 seconds, calculating the average dust and fog concentration in the window, and if the average value exceeds 1 mg / m 3 , it is determined that the air quality has a potential impact on vision; if the concentration continues to rise and exceeds 3mg / m 3 If it lasts for more than 10 seconds, it is judged as a highly polluted environment, and the credibility of the visual data needs to be adjusted immediately. The credibility update model based on exponential decay is adopted. Assuming the initial credibility of the visual data is C0, and the concentration of fog and dust is d, the calculation formula of the updated credibility C is: C=C0×e -k×d ; Where k is the attenuation coefficient, e is a natural constant, and when the calculated credibility C is less than the preset initial credibility C0 of the visual data, it is judged that the credibility of the visual data sensor is low; The credibility under the influence of light and the credibility under the combined influence of temperature and humidity include: data verification, the normal range of light intensity is set at 0lx to 100000lx, when the rate of change of light intensity exceeds 30%, that is, the difference between two adjacent acquisition values ​​divided by the current value is greater than 0.3, it is judged that the lighting conditions have changed significantly, for light intensity greater than 80000lx, record the duration of strong light, if it exceeds 5 seconds, mark it as a strong light interference environment; for light intensity less than 100lx, if it lasts for more than 10 seconds, mark it as a weak light environment, these marks are used for subsequent visual data credibility adjustment, for strong light interference environment, a linear decreasing model is used, when in a strong light interference environment, the credibility decreases by 0.05 per second, assuming that the duration of strong light is t, then the formula of credibility C is: C = C0-0.05×t; When the credibility C calculated here is less than the preset initial credibility C0 of the visual data, it is judged that the credibility of the visual data sensor is low; For low-light environments, the reliability decreases logarithmically. Assuming the light intensity in low-light environments is I, the calculation formula for the reliability C is: When the credibility C calculated here is less than the preset initial credibility C0 of the visual data, it is judged that the credibility of the visual data sensor is low; The credibility update under the influence of light and the credibility update under the combined influence of temperature and humidity include: each time a new set of temperature and humidity data is collected, a simple data check is first performed to determine whether the data is within a reasonable physical range, where the temperature is between -40℃ and 85℃, and the humidity is between 0% and 100%. If not, it is marked as abnormal data. For normal data, the difference between the current temperature and humidity and the average temperature and humidity in the previous 10 minutes is calculated. If the absolute value of the difference is greater than the set threshold, the temperature change threshold is set to 2℃, and the humidity change threshold is set to 5%. It is determined that the ambient temperature and humidity have changed significantly, triggering the subsequent credibility adjustment process for the visual sensor. Specifically, a two-dimensional lookup table is constructed, with the row index being the temperature range, one interval of every 5℃, from -20℃ to 40℃; the column index being the humidity range, one interval of every 10%RH, from 0% to 100%; each element in the table stores a credibility adjustment factor f, where the element is a cell with a row index of the temperature range and a column index of the humidity range. When the current temperature and humidity are monitored, the corresponding adjustment factor f is obtained according to the lookup table. The calculation formula for the updated credibility C is: C = C0 × (1-f); When the credibility C calculated here is less than the preset initial credibility C0 of the visual data, it is judged that the credibility of the visual data sensor is low; S403: Dynamic allocation of fusion weights: Based on the real-time updated credibility, when the visual sensor, the distance detection sensor, the environmental monitoring sensor, and the tactile sensor are performing data fusion, a weighted average method is used to allocate fusion weights to different sensor data; S404: Regular data synchronization comparison: The data synchronization mechanism is started every 100 milliseconds between adjacent sensors. Vision sensors transmit key features of the currently acquired image, including object contours and color histograms, to each other through wireless communication; distance detection sensors share distance measurement data and preliminary point cloud clustering results; tactile sensors communicate contact force distribution and vibration frequency; S405: Difference analysis and correction: When two adjacent visual sensors compare image features, if the difference in the contour of the same object is found to exceed the preset threshold, the parameters of each image acquisition are immediately traced back, including the shooting angle, lens focal length, and illumination compensation value. The image registration algorithm is used to perform geometric transformation and grayscale adjustment on the image according to the difference in these parameters to realign the features and correct the deviation. If the deviation is caused by uneven illumination, the illumination compensation parameters are adjusted and the image is re-acquired to ensure the consistency of the visual data. S406: Collaborative error correction feedback: In the working operation scenario of the intelligent robot, the distance detection sensor and the tactile sensor work together. Once the tactile sensor detects a sudden change in the grasping force that exceeds the normal grasping force fluctuation range, it immediately sends a feedback signal to the distance detection sensor. After receiving the signal, the distance detection sensor quickly performs a second distance measurement, recalculates the grasping target position and posture, and compares it with the previous distance measurement results. If deviations are found, the intelligent robot's actions are adjusted in time to correct the grasping position.

5. The intelligent robot control method based on multi-sensor collaboration according to claim 4 is characterized in that: In S402, the credibility adjustment process includes: once the environmental monitoring sensor detects that the fog concentration exceeds the standard, the light changes suddenly, or the temperature and humidity fluctuate greatly, it immediately sends an interrupt signal to the visual sensor. After receiving the notification, the visual sensor suspends the current operation and calls the corresponding credibility update algorithm. If fog or dust is detected, the real-time collected fog and dust concentration values ​​are substituted into the exponential decay formula to quickly calculate the new credibility value. Then, the visual sensor propagates the updated credibility value to the distance detection sensor and the environmental monitoring sensor, which is used to increase its own data weight in the subsequent data fusion link, and participate in the fusion of sensor data according to the new weight allocation rule. At the same time, the current environmental status is marked, and the cause and magnitude of this credibility adjustment are recorded and stored in the log.

6. The intelligent robot control method based on multi-sensor collaboration according to claim 1, characterized in that: The S5 includes the following steps: S501: Initially, pheromones are evenly distributed in the entire task space of the intelligent robot, and the data of each sensor randomly selects the direction of travel for exploration. As the number of explorations increases, pheromones gradually accumulate on the paths leading to successful results, and subsequent sensors tend to select these paths with high concentrations of pheromones. S502: Setting the pheromone to volatilize 10% every fixed time period is used to gradually reduce the pheromone of old and inefficient paths, prompting the sensor to re-explore possible new paths, maintaining adaptability to environmental changes, and preventing excessive accumulation of pheromones from causing path exploration to fall into a local optimum; When the intelligent robot completes a small phased goal or the entire task, it will give additional pheromone rewards to the sensors on the successful path; S503: The control direction of the intelligent robot is updated by guiding pheromones through sensor data. The visual sensor captures the environment image in real time. Once a new potential target area is found, the visual signal is broadcast to the surrounding sensors to guide the nearby sensors to explore in that direction. The concentration of pheromones is pre-set on the initial exploration path to speed up the exploration of new areas. Among the distance detection sensors, the laser radar and ultrasonic sensors continuously monitor surrounding obstacles. If they detect a new obstacle or narrow passage ahead, they quickly reduce the pheromone concentration on the current path and send warning signals to other sensors, prompting them to avoid dangerous areas and reselect a safe path with a relatively high pheromone concentration, thus preventing the intelligent robot from being damaged by collision. S504: Task priority control adjustment based on pheromone concentration. The intelligent robot periodically scans the pheromone concentration on each path and associates the most urgent task to be completed by the intelligent robot with the path with high pheromone concentration. Paths where pheromone concentrations have been low for a long time and have not significantly promoted the mission are determined to be inefficient paths. The intelligent robot suspends sensor activity on the path, re-analyzes the collected data, and attempts to optimize sensor parameters and adjust action strategies. If multiple optimizations are unsuccessful, the path exploration is completely abandoned, and sensors are reallocated to other potential areas to rationally utilize resources and improve overall mission effectiveness.

7. The intelligent robot control method based on multi-sensor collaboration according to claim 1 is characterized in that: The S6 also includes designing a real-time capability evaluation system: each honeycomb sensor calculates a comprehensive capability value based on its current hardware performance, data quality, and past task completion performance. During the task execution process, once new task requirements or emergency situations arise, sensors with high capability values ​​quickly take over key roles and cooperate with other sensors to perform tasks.

Images