AI modular perception robot service system based on precision sensing

Abstract

The application relates to the technical field of robot systems, in particular to an AI modular perception robot service system based on precision sensing, which comprises an environment perception module, a sensing sensitivity adjustment module, a sensor performance optimization module, an environment data repair module and a task rearrangement module. Through real-time data acquisition and environment fluctuation monitoring, the application effectively manages illumination and obstacle information, ensures efficient operation of the robot under various working conditions, dynamically adjusts the sensitivity of the sensor to allow the robot to maintain an optimal perception state when the environment changes, reduces energy consumption while improving reaction speed, and the real-time response to the intensity of the light and the frequency of obstacle detection not only improves the stability of the operation, but also reduces the error decision caused by environmental misreading. The introduction of the data repair mechanism improves the reliability of data identification and enhances the application ability of the robot in complex scenes, so that the robot can more accurately execute tasks and reduce potential risks and costs.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot systems, in particular to an AI modular perception robot service system based on precise sensing. BACKGROUND

[0002] Robot systems involve the design, development and application of robots capable of autonomous task execution, including the development of hardware, software and control systems for various types of robot systems, in which robots perceive the environment through various sensors and make corresponding decisions to complete predetermined tasks. Robot systems can be divided into industrial robots, service robots, medical robots, household robots and other different categories, and are widely used in automated production, logistics, medical care, home and intelligent manufacturing industries.

[0003] Among them, the AI modular perception robot service system based on precise sensing combines precise sensing, artificial intelligence (AI) and modular design, acquires environmental data through precise sensors, processes and analyzes the data using AI algorithms, and realizes the perception and intelligent decision-making of the robot to the environment. The robot system adopts modular design, so that it can be flexibly configured according to different application requirements, and is suitable for various service tasks, mainly applied to improve the service capability of robots in different scenarios, such as automated logistics, medical support, rescue, etc.

[0004] The prior art limits the response to dynamic environmental factors due to the static configuration of sensors and decision-making, for example, robots perform well in fixed lighting and obstacle environments, but cannot adjust in real time in scenarios where lighting conditions or obstacle layouts change frequently, resulting in decreased efficiency or operational errors. The prior art lacks an efficient data anomaly processing mechanism, when faced with incorrect or biased data, the robot continues to make decisions based on these inaccurate data, increasing the risk of operation, and in the field of medical robots, inaccurate processing of environmental perception data can affect the safety and accuracy of the diagnosis process, limiting the practicality of the robot system. SUMMARY

[0005] The purpose of the present application is to solve the shortcomings in the prior art, and an AI modular perception robot service system based on precise sensing is proposed.

[0006] In order to achieve the above purpose, the present application adopts the following technical scheme: an AI modular perception robot service system based on precise sensing, the system comprises:

[0007] The environmental perception module acquires the light intensity and the number of obstacles in the robot work area, uses a photoelectric sensor array to collect light data in real time, counts the number of times the obstacles are identified in the omnidirectional wheel navigation path grid, determines the fluctuation range of the two types of data, and generates a perception data fluctuation index;

[0008] The induction sensitivity adjustment module analyzes the fluctuation of the illumination intensity according to the perception data fluctuation index, and if the fluctuation amplitude exceeds the standard configuration range, adjusts the infrared sensor parameter to a high sensitivity state, and optimizes the image sensor power consumption, to obtain a sensing configuration state;

[0009] The sensor performance optimization module analyzes the fluctuation data of the illumination intensity and the obstacle detection frequency based on the sensing configuration state, adjusts the exposure time and the signal sampling frequency according to the analysis result, and obtains a sensing adjustment range.

[0010] The environmental data repair module analyzes the sensing adjustment range, identifies the data sequence deviating from the normal fluctuation range according to the fluctuation amplitude of the illumination intensity and the obstacle recognition data, processes the abnormal data segment by using data sorting and median extraction, replaces the unstable data segment, and generates an abnormal situation correction log.

[0011] The present application improves that the perception data fluctuation index includes illumination level and obstacle recognition frequency, the sensing configuration state includes sensitivity level and power consumption standard, the sensing adjustment range includes exposure parameter adjustment information and frequency setting, and the abnormal situation correction log includes data stability information and correction details.

[0012] The present application improves that the environmental perception module includes:

[0013] The illumination collection submodule obtains the illumination intensity and the number of obstacles in the robot operation area, compares the illumination data collected by the photoelectric sensor array with the array trigger boundary, judges whether the illumination intensity exceeds the boundary range, and screens the number of illumination points deviating from the boundary, to obtain the number of illumination deviations.

[0014] The obstacle recognition submodule collects the trigger frequency of obstacles at the path nodes according to the omnidirectional wheel navigation path grid, judges whether the trigger times reach the obstacle recognition standard according to the boundary trigger position data of the photoelectric sensor, and extracts the number of path nodes occupied by the recognized obstacles, to obtain the number of obstacle triggers.

[0015] The fluctuation statistical submodule calls the statistical data of the two types of data in multiple sensing cycles based on the number of illumination deviations and the number of obstacle triggers, analyzes the amplitude, direction difference and frequency fluctuation in the time sequence, and generates the perception data fluctuation index.

[0016] The present application improves that the induction sensitivity adjustment module includes:

[0017] The fluctuation recognition submodule obtains the illumination intensity detection data according to the perception data fluctuation index, calculates the illumination intensity change rate in a continuous time period, compares it with the illumination fluctuation reference interval, judges whether it appears fluctuation deviation, and obtains the illumination fluctuation recognition result.

[0018] The parameter adjusting submodule evaluates whether the current sensitivity of the infrared sensor needs to be adjusted based on the illumination fluctuation recognition result, and adjusts the infrared sensor parameters to a high sensitivity state if the illumination fluctuation exceeds the standard configuration range, to obtain an infrared sensitivity adjustment result.

[0019] The power consumption control submodule calls the infrared sensitivity adjustment result, adjusts the processing frame rate, gray scale number and compression encoding of the image sensor according to the power consumption configuration state of the current image sensor, performs power consumption optimization, and obtains a sensor configuration state.

[0020] The sensor performance optimization module comprises:

[0021] The illumination frequency analysis submodule calls the current infrared and image sensor configuration based on the sensor configuration state, compares the illumination intensity and obstacle detection frequency measurement data in the continuous time period, analyzes the joint change degree of illumination and frequency, determines the data fluctuation characteristics, and obtains illumination frequency fluctuation data.

[0022] The sensor adjusting submodule analyzes the compliance of the current exposure time and sampling frequency according to the illumination frequency fluctuation data, adjusts the adjustable range of the sensor according to the analysis result, updates the sensor range limit, matches the real-time environmental changes, and obtains a sensor adjustment range.

[0023] The environment data repair module comprises:

[0024] The range determination submodule analyzes the sensor adjustment range, extracts the numerical variation amplitude in the adjacent time period according to the illumination intensity and obstacle recognition data, judges whether it exceeds the upper and lower boundary intervals of the normal fluctuation range, marks the time points deviating from the fluctuation range, and generates an abnormal fluctuation time sequence.

[0025] The abnormality extraction submodule calls the abnormal fluctuation time sequence, collects the illumination intensity data and obstacle recognition data in the corresponding time period, sorts the two data sequences in ascending order according to the values, removes the first and last values, takes the median of the remaining data, and generates a reference median set.

[0026] The data replacement submodule calculates the difference between the original illumination intensity data and obstacle recognition data in the abnormal fluctuation time period according to the reference median set, calculates the deviation degree from the median, and obtains an abnormal situation correction log.

[0027] The system further comprises:

[0028] The task rearrangement module corrects the obstacle position data in the abnormal situation correction log, recalculates the associated coordinate points in the task path graph, rearranges the path priority according to the task number, execution time and path length, and obtains a navigation priority configuration;

[0029] The navigation priority configuration includes a priority sequence, a coordinate point adjustment result and a time allocation result.

[0030] The task rearrangement module includes:

[0031] The obstacle position correction submodule compares the old coordinates of the obstacles in the original path graph with the newly updated coordinates based on the abnormal situation correction log, identifies the coordinate differences between the two, detects whether the offset amplitude exceeds the target range, and updates the obstacle position in the path graph if it does, and generates an obstacle position offset amplitude.

[0032] The path priority recalculation submodule calls the obstacle position offset amplitude, recalculates the optimal path of the key task point associated with the obstacle, analyzes the influence of path length changes on task execution, rearranges the path priority according to the task number and task execution time, and obtains a navigation priority configuration.

[0033] Compared with the prior art, the application has the following advantages and positive effects:

[0034] In the application, real-time data acquisition and environmental fluctuation monitoring are used to effectively manage light and obstacle information, ensuring efficient operation of the robot under various working conditions. Dynamic sensor sensitivity adjustment allows the robot to maintain optimal perception state when the environment changes, reducing energy consumption while improving reaction speed. This real-time response to light intensity and obstacle detection frequency not only improves the stability of operation, but also reduces the risk of incorrect decisions due to environmental misreading. The introduction of the data repair mechanism improves the reliability of data recognition and enhances the application ability of the robot in complex scenarios, enabling more accurate task execution and reducing potential risks and costs. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 The system flowchart of the application;

[0036] Figure 2 The flowchart of the environmental perception module in the application;

[0037] Figure 3 The flowchart of the sensing sensitivity adjustment module in the application;

[0038] Figure 4 The flowchart of the sensor performance optimization module in the application;

[0039] Figure 5A flow chart of the environment data repairing module in the present application;

[0040] Figure 6 A flow chart of the task rearrangement module in the present application. DETAILED DESCRIPTION

[0041] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0042] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the present application and simplify the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly and specifically limited.

[0043] EMBODIMENT

[0044] Please refer to Figure 1 The present application provides a technical solution: an AI modular perception robot service system based on precision sensing includes:

[0045] The environment perception module acquires the light intensity and the number of obstacles in the robot working area, uses the photoelectric sensor array to collect the light data in real time, compares with the array trigger boundary, and counts the number of times the obstacles are identified in the omnidirectional wheel navigation path grid, determines the fluctuation range of the two data types, and generates the perception data fluctuation index;

[0046] The sensing sensitivity adjustment module analyzes the fluctuation of the light intensity according to the perception data fluctuation index, compares with the standard sensitivity configuration of the infrared sensor and the image sensor, if the fluctuation amplitude exceeds the standard configuration range, adjusts the infrared sensor parameter to the high sensitivity state, and optimizes the image sensor power consumption, to obtain the sensing configuration state;

[0047] The sensor performance optimization module calls the current infrared and image sensor configuration based on the sensing configuration state, analyzes the fluctuation data of the light intensity and the obstacle detection frequency, adjusts the exposure time and the signal sampling frequency according to the analysis result, and obtains the sensing adjustment range;

[0048] The environmental data repair module analyzes the sensing adjustment range, identifies data sequences deviating from the normal fluctuation range based on the light intensity and obstacle recognition data fluctuation amplitude, processes abnormal data segments using data sorting and median extraction, replaces unstable data segments, and generates an abnormal situation correction log;

[0049] The task rearrangement module recalculates the associated coordinate points in the task path graph based on the updated obstacle position data in the abnormal situation correction log, rearranges the path priority based on the task number, execution time, and path length, and obtains the navigation priority configuration.

[0050] The perception data fluctuation indicators include light levels and obstacle recognition frequencies, the sensing configuration states include sensitivity levels and power consumption standards, the sensing adjustment ranges include exposure parameter adjustment information and frequency settings, the abnormal situation correction logs include data stability information and correction details, and the navigation priority configurations include priority sequences, coordinate point adjustment results, and time allocation results.

[0051] The photoelectric sensor array is composed of multiple photoelectric sensors for real-time detection and conversion of light signals into electrical signals. This array allows the system to accurately measure light intensity in a wide area. The omni-directional wheel navigation path grid is a virtual or actual path division used by the robot for navigation, where the omni-directional wheels allow the robot to move in any direction, improving its maneuverability in complex environments. The perception data fluctuation indicators are statistical measures based on real-time changes in light intensity and obstacle detection frequency to assess environmental stability. They are used to evaluate whether changes in the environment exceed the robot's normal operating range. The sensing configuration state reflects the adjusted working mode of the current sensor, including the sensitivity of the infrared sensor and the power consumption level of the image sensor. It is the result of dynamic adjustment based on environmental conditions to optimize performance and energy efficiency. The sensing adjustment range refers to the range of adjustment parameters based on environmental data and sensor performance feedback, including the adjustment interval of exposure duration and signal sampling frequency to adapt to the actual needs of environmental light and obstacle detection. The abnormal situation correction log contains all data anomaly records identified and corrected by the system, detailing the nature, processing process, and results of each data anomaly. The navigation priority configuration covers the path priority ordering of task execution, time scheduling of each task node, and updated coordinate point positions. The configuration ensures the efficiency and accuracy of the robot during task execution.

[0052] Please refer to Figure 2 , the environmental perception module includes:

[0053] The illumination collection sub-module obtains the illumination intensity and the number of obstacles in the robot working area, compares the illumination data collected by the photoelectric sensor array with the array trigger boundary, judges whether the illumination intensity exceeds the boundary range, and screens the number of illumination points deviating from the boundary to obtain the number of illumination deviations;

[0054] First, photoelectric sensor arrays need to be laid out in the area covered by the robot working path, for example, 25 uniform measuring points are laid out in a 5m x 5m working area, each measuring point continuously collects the illumination intensity at the current time, in Lux, the sampling interval is set to 1 second, the sensor reads the values such as: measuring point P1 is 380 Lux, P2 is 410 Lux, P3 is 290 Lux, P4 is 470 Lux, P5 is 520 Lux, the system performs boundary comparison operation on all data, the illumination boundary range is set to lower limit 350 Lux and upper limit 500 Lux, which is determined with reference to the stable range of lighting in the industrial working scene, the boundary value setting process is measured by the average standard deviation method during the normal operation of the historical working area, the average value of the illumination under normal operation is 425 Lux, and the fluctuation range is ± 75 Lux, after setting, the illumination intensity of each measuring point is compared, for example, P3 is 290 Lux, which is lower than the lower limit 350, and is judged as deviating; P5 is 520 Lux, which exceeds the upper limit 500, and is also judged as deviating; P1, P2 and P4 are within the normal range, and are judged as not deviating, the system records the numbers of all deviating points and counts the number of them, in the above data, there are 2 deviating points, namely P3 and P5, so the number of illumination deviations in the current period is 2, the judgment process does not need complex algorithm, only needs to compare each measuring point reading once with the upper and lower limits of the boundary, and use Boolean variable to record True or False, and finally add the Boolean value to obtain the number of deviations.

[0055] The obstacle recognition sub-module collects the trigger frequency of obstacles at the path nodes according to the omni-directional wheel navigation path grid, judges whether the trigger frequency reaches the obstacle recognition standard according to the boundary trigger position data of the photoelectric sensor, and extracts the number of path nodes occupied by the recognized obstacles to obtain the number of obstacle triggers;

[0056] The robot uses an omnidirectional wheel navigation system to traverse a path grid within an area. Each path node is equipped with an obstacle-triggered perception mechanism. The system records the number of triggers for each path node within a sensing cycle, using node numbers. The path grid is set to 4 rows × 4 columns, with a total of 16 nodes numbered from N1 to N16. For example, the trigger frequencies recorded in a certain cycle are as follows: N3 triggers 4 times, N5 triggers 2 times, N7 triggers 3 times, and N8 triggers 1 time. The system sets the obstacle judgment standard as follows: if a single node triggers ≥ 3 times within a single cycle, it is considered to have an obstacle. This judgment criterion is derived from the statistical analysis of navigation deviations and path interruptions during operational testing. The statistical results show that continuous... Over 90% of the scenarios corresponding to 3 triggers have obstacle blockage, so this is set as a reasonable recognition lower limit. This setting is determined based on empirical data and can also be dynamically adjusted. According to the above records, N3 and N7 meet the conditions, so the number of obstacle triggers in this cycle is 2. The data recording method is to automatically accumulate the trigger counter of each node. At the end of the cycle, the trigger values ​​of all nodes are read, and the total number of nodes that meet the obstacle recognition conditions is selected according to the judgment criteria. When the judgment operation is executed, the current trigger count of each node is called, and it is compared with the value 3 to be greater than or equal to. The result is a list of Boolean values. The total number of True values ​​is the recognition count, which is used as the obstacle trigger count analysis.

[0057] The fluctuation statistics submodule, based on the number of illumination offsets and the number of obstacle triggers, calls statistical data from these two types of data over multiple sensing cycles to analyze the amplitude, directional differences, and frequency fluctuations in the time series, using the following formula:

[0058] ;

[0059] Generate perceptual data fluctuation indicators ,in, Representing the The number of illumination offsets within a period represents the number of data points recorded by the photoelectric sensor that deviate from the average illumination data within that period. This is the periodic average of the number of illumination offsets. This is a weighting factor for illumination shift, used to adjust the degree of influence of illumination data in the total fluctuation index. For the first The number of obstacle triggers within a cycle indicates the number of times the omnidirectional wheel navigation system detects and triggers an obstacle within that cycle. This represents the periodic average number of obstacle triggers. This is the weighting coefficient for obstacle triggering, used to adjust the degree of influence of obstacle data in the total volatility index. The number of illumination data cycles. Number of obstacle data cycles;

[0060] Data on illumination offset and obstacle trigger counts over multiple consecutive periods are aggregated and analyzed to obtain the dynamic fluctuations of the overall operating environment. The illumination offset sequence is as follows: , , , The average value over the calculation period is: Let the illumination weighting coefficient be... =0.6, the sequence of obstacle trigger counts is: , , , The average value is: Let the obstacle weight coefficient be... =0.8, the weighting coefficient is set based on the relative comparison of the influence of the two types of data on the fluctuation trend. The impact of light fluctuation on system response delay is less than that of obstacle recognition, so the weight is lower. Substitute it into the formula for calculation:

[0061] The first part, the sum of the absolute differences in the illumination offset, is:

[0062] ;

[0063] After multiplying by the weights, it becomes:

[0064] ;

[0065] Part Two, the sum of squared differences of barrier-triggered fluctuations is:

[0066] ;

[0067] Multiplied by the weight:

[0068] ;

[0069] Calculation after merging:

[0070] ;

[0071] The obtained perception data fluctuation index DS=0.475. This value represents the overall amplitude level of the fluctuation of light and obstacle data within the current time window. If compared with the set judgment range [0.2, 0.7], this value is in the medium to high range, indicating that there is a certain degree of fluctuation in light and obstacles in the current working environment.

[0072] Please see Figure 3 The sensor sensitivity adjustment module includes:

[0073] The fluctuation recognition submodule obtains light intensity detection data based on the fluctuation index of the sensing data, calculates the rate of change of light intensity within a continuous time period, compares it with the light fluctuation benchmark interval, determines whether there is a fluctuation shift, and obtains the light fluctuation recognition result.

[0074] In a certain lighting area, the light intensity is automatically recorded once per minute. Five consecutive records yield values ​​of 300, 320, 280, 310, and 330 Lux. After acquiring the data, the rate of change can be calculated by analyzing the intensity changes between adjacent time points. The calculation method is to subtract the intensity of the previous time point from the current light intensity, resulting in a rate of change for each time period: +20, -40, +30, and +20 Lux. This sequence shows alternating positive and negative values ​​with large absolute values. To further determine if there is any fluctuation or deviation, a fluctuation baseline range needs to be set. Assuming this range is currently set to ±10 Lux based on historical data and the scene, each rate of change needs to be evaluated individually. The evaluation method involves taking the absolute value of each rate of change and comparing it with a threshold. For example, for the change... If the rate is -40, the absolute value is 40, which is greater than 10. Therefore, this time point is judged as a fluctuation offset. Similarly, if the rate of change is +20, its absolute value is 20, which is also judged as an offset item. The judgment is carried out in sequence. If the rate of change is +5, it is lower than the upper and lower limits of the benchmark and is not considered as an abnormal fluctuation. Four offset items are judged, namely the first, second, third and fourth time periods. For example, in a parking lot lighting area, if the light intensity suddenly drops by more than 40 Lux within one minute due to cloud cover or other reasons, and then quickly rises by 30 Lux, then this continuous and drastic change represents that the lighting environment is in a fluctuating state, and the sensing system needs to respond and adjust. Thus, the comparison between the rate of change of light intensity and the set benchmark interval determines whether a fluctuation has occurred, and outputs the light fluctuation recognition result accordingly.

[0075] The parameter adjustment submodule assesses whether the current sensitivity of the infrared sensor needs adjustment based on the illumination fluctuation recognition results. If the illumination fluctuation exceeds the standard configuration range, the infrared sensor parameters are adjusted to a high-sensitivity state using the following formula:

[0076] ;

[0077] Obtain infrared sensitivity adjustment results This represents the overall output of the infrared sensor after sensitivity adjustment under specific lighting conditions, where... It is the first Sub-infrared signal intensity sampling data, It is the initial reference for infrared signal strength, used as a comparison benchmark to help determine the amount of signal change. It is the first Sub-infrared band responsivity, described in the first... During the second sampling, the response strength of the infrared sensor to a signal in a specific frequency band, It is the first The sensitivity influence coefficient at the sampling frequency indicates the degree of influence of the sampling frequency on the sensitivity of the infrared sensor. It is the infrared signal response reference, the baseline response level of the infrared sensor, used for comparison with sampled data. No. The amplitude of the second sampling reflects the first sampling fluctuation. The magnitude of the fluctuation in the infrared signal during each sampling. This is the total number of samples.

[0078] There are three sampling data points, with infrared signal intensity data of 250, 270, and 230 respectively, and an initial reference is set. For the first sampling point 250, the infrared band responsivity The sensitivity influence coefficients are 1.1, 1.2, and 1.1 respectively. The values ​​are 0.8, 0.9, and 0.85 respectively, representing the infrared signal response benchmarks. Set to 100, sampling fluctuation amplitude The values ​​are 5, 8, and 7 respectively. Based on the data, the parameters are substituted into the formula for calculation. For the first sampling point:

[0079] Then multiply by the sampling fluctuation amplitude ;

[0080] ;

[0081] For the second sampling point:

[0082] Then multiply by the sampling fluctuation amplitude ;

[0083] ;

[0084] For the 3rd sampling point:

[0085] Then multiply by the sampling fluctuation amplitude ;

[0086] ;

[0087] Sum the results above:

[0088] ;

[0089] Infrared sensitivity adjustment results obtained Based on this result, if the system sensitivity adjustment threshold is set to... If sensitivity is increased only at this time, then at this time... A value less than zero indicates that the sensitivity of the infrared sensor does not need to be adjusted at present.

[0090] The power consumption control submodule calls the infrared sensitivity adjustment result, and adjusts the image sensor's processing frame rate, grayscale level number and compression encoding according to the current power consumption configuration status of the image sensor to optimize power consumption and obtain the sensor configuration status.

[0091] Based on the current power consumption configuration of the image sensor, the processing frame rate, grayscale level, and compression encoding of the image sensor were adjusted to optimize power consumption, and the infrared sensitivity adjustment results have been obtained. According to the system's preset power consumption adjustment rules, assuming the system's sensitivity adjustment threshold is set as follows: when When this happens, it enters high-sensitivity mode and adjusts the image sensor power consumption; when When this happens, it enters a low-power mode; therefore, based on this result... If the value is less than 0, the system determines that no increase in sensitivity is needed, thus triggering a power control mode. In low-power mode, the system reduces the frame rate of the image sensor from the normal 30fps to 20fps, the number of grayscale levels from 256 to 128, and the image compression coding switches from full-frame detail coding to region-priority compression. That is, only areas with significant changes in the scene are saved with high precision, while the rest are encoded with a high compression ratio. In this way, the power consumption of the entire system is reduced, and the working time of the device is extended. Specifically, the processing power of the image sensor is adjusted to adapt to the low-sensitivity working state, reducing the real-time processing requirements of image data, thereby reducing power consumption. The optimization of the image compression algorithm also helps to reduce the bandwidth requirements for storage and transmission, improving the overall system efficiency. This power control mechanism automatically adjusts the working state of the image sensor according to the changes in the infrared sensitivity adjustment results, ensuring that the system can balance performance and energy efficiency under various environmental conditions.

[0092] Please see Figure 4 The sensor performance optimization module includes:

[0093] The illumination frequency analysis submodule, based on the sensor configuration status, calls the current infrared and image sensor configurations, compares the illumination intensity with obstacle detection frequency measurement data in a continuous time period, analyzes the degree of joint change of illumination and frequency, determines the data fluctuation characteristics, and obtains illumination frequency fluctuation data.

[0094] The system reads the current sampling frequency value of the infrared sensor, which represents the number of times the sensor samples per second. Simultaneously, it reads the currently set exposure time value of the image sensor, representing the exposure duration of each image frame. Based on this, it extracts the light intensity data and obstacle detection frequency data recorded at a sampling interval of one second over a continuous 10 seconds. The two values ​​at the same time point are paired to form a data set. The light intensity value comes from the instantaneous measurement value of each photoelectric array sensor in the working area, measured in Lux. The obstacle detection frequency is the number of times an obstacle is detected in the same spatial unit per unit time. For example, in a practical application scenario, when the robot enters an area with unstable lighting, such as an underground parking garage, the light intensity is 300 in the first second and drops to 210 in the second, with the obstacle frequency increasing from 1.0 to 1.8. The differences between the two times are recorded sequentially and summed as a single-segment joint fluctuation value, and this process is continuously performed. After processing, a complete sequence of 10 joint fluctuations is formed. Then, all joint fluctuation values ​​in this sequence are summarized and their average level and dispersion are statistically analyzed. If the fluctuation trend deviates significantly from the overall average level, for example, if the change values ​​at multiple points are much higher than the standard range of average change, it is identified as a significant fluctuation. The specific threshold is manually set by the robot's operating environment. Usually, the combination of light change greater than 100 and detection frequency change greater than 1.5 is taken as the reference standard. When it is detected that the light value drops to 120 at the 6th second and the obstacle frequency rises to 2.0, compared with the case of light 250 and frequency 0.5 at the 5th second, the change in the two indicators is very significant. This fluctuation will be marked as a joint abnormal fluctuation. The temporal position of the joint fluctuation point, the corresponding light and frequency values, and the judgment result are encapsulated into a set of light frequency fluctuation data for subsequent sensor configuration analysis.

[0095] The sensor adjustment submodule analyzes the consistency between the current exposure time and the sampling frequency based on the light frequency fluctuation data, adjusts the adjustable range of the sensor based on the analysis results, updates the sensor range limit, matches real-time environmental changes, and obtains the sensor adjustment range interval.

[0096] First, the abnormal fluctuation segments are traversed to extract the infrared sensor sampling frequency and image sensor exposure time corresponding to each time point. These values ​​are then compared one by one with the recommended frequency and time range in the equipment standard. If the current configuration value is close to the recommended upper limit, such as an infrared frequency of 59 and an image exposure time of 145, it is determined that the performance limit is about to be reached, and the configuration capability must be expanded. Next, the intensity of each fluctuation segment is divided into levels: fluctuation below 25 is the lowest level, between 25 and 50 is the medium level, and above 50 is the high level. The corresponding transmission is determined based on the level. The sensor adjustment range is defined as follows: for example, if the change range is 68, it is determined to be a level 3 fluctuation. The sampling frequency of the infrared sensor is adjusted upward by 15, and the exposure time of the image sensor is adjusted upward by 30. After the adjustment is completed, it is determined whether the new parameters exceed the maximum boundary value allowed by the hardware. For example, whether the infrared frequency is greater than 100 and the image exposure is greater than 200. If they do not exceed, the configuration boundary value is directly updated. If they exceed, they are kept at the maximum allowable upper limit, and a new configuration range for the infrared and image sensors is generated, which represents the adjustment range allowed by the sensors to adapt to the combined change trend of illumination and frequency in the current detection environment.

[0097] Please see Figure 5 The environmental data repair module includes:

[0098] The range determination submodule analyzes the sensor adjustment range, extracts the numerical variation amplitude in adjacent time periods based on light intensity and obstacle recognition data, determines whether it exceeds the upper and lower limits of the normal fluctuation range, marks the time points that deviate from the fluctuation range, and generates an abnormal fluctuation time series.

[0099] First, the upper and lower limits of the sensor parameters are read, namely the minimum and maximum sampling frequencies of the infrared sensor and the minimum and maximum exposure times of the image sensor. Combined with the current light intensity data and obstacle recognition data, the light intensity value and obstacle recognition frequency value are extracted for a continuous time period in 1-second increments. The changes in light intensity and recognition frequency between two adjacent time points are calculated. The sum of these two changes represents the numerical fluctuation range for that time period. Next, it is determined whether this range is within the normal fluctuation range. The normal fluctuation range is set based on statistical results from historical operating environments. The normal range for light intensity fluctuation is set to 0 to 80, and the normal range for obstacle recognition frequency fluctuation is set to 0 to... Therefore, when the fluctuation range within any time period exceeds 80 for illumination or 1.0 for recognition, it is considered to be outside the normal fluctuation range. This time point is recorded as an abnormal time point and added to the abnormal record sequence. The entire time series is then traversed. For example, if the illumination value increases from 120 to 240 and the recognition frequency increases from 0.3 to 1.5, the illumination change range is 120 and the recognition change range is 1.2. Both exceed the upper limit of the corresponding fluctuation range, and this time point is marked as an abnormal fluctuation point. After repeating the above operation, an abnormal fluctuation time series is formed, which contains all the time points that are judged to be outside the normal fluctuation range and their corresponding illumination and recognition change values.

[0100] The anomaly extraction submodule calls the abnormal fluctuation time series, collects light intensity data and obstacle recognition data within the corresponding time period, sorts the two data sequences in ascending order by value, removes the first and last values, takes the median of the remaining data, and generates a reference median set.

[0101] Read the time points marked as abnormal one by one, and retrieve the light intensity data of that time point and the three moments within one second before and after it, along with the obstacle recognition data to form a data segment. Sort the light intensity values ​​in each data segment in ascending order, and then sort the recognition data values ​​in the same ascending order. After sorting, remove the first and last data points in the sorted sequence, and keep the data in the middle segment. Then, extract the median from the retained data, that is, select the value in the middle position from the remaining data as the median representative value. If the number of remaining values ​​is even... The average of the two middle values ​​is taken as the median representative value. The median of illumination and recognition at each abnormal time point is collected and stored separately to form a set of medians. For example, if the abnormal time point is the 7th second, the corresponding illumination intensity values ​​are 220, 180, and 160. After sorting, they are 160, 180, and 220. After removing the first and last values, the median value is 180. The corresponding recognition frequency values ​​are 0.6, 0.9, and 1.8. After sorting, they are 0.6, 0.9, and 1.8. After removing the first and last values, the median value is 0.9. All abnormal time periods are processed in sequence to generate a reference median set.

[0102] The data replacement submodule, based on a reference median set, calculates the difference between the original illumination intensity data and obstacle recognition data within the abnormal fluctuation period, and calculates the deviation from the median using the formula:

[0103] ;

[0104] Get the Data correction bias for different time periods Replace unstable data fragments and generate anomaly correction logs, among which, Representing the The raw illumination intensity data for a given time period represents the initial measured values ​​of illumination intensity recorded during that time period. Representing the The reference median for the time period is calculated from the filtered light intensity and obstacle recognition data, and is used as a stabilizing and reference value to replace outlier data points. Representing the Within the time period The fluctuation range of light intensity indicates the value of the first group within that time period. The magnitude of the fluctuation in light intensity of this data set compared to other data points. Representing the Within the time period The fluctuation range of obstacle recognition reflects the obstacle recognition data in the first group. The range of change within the group Representing the The raw obstacle recognition data within a time period is the initial data on obstacle recognition recorded during that time period. It represents the total number of data points analyzed within each time period.

[0105] Light intensity data (unit lx) is collected from the robot's array photoelectric sensor at a rate of 1 time per second to form data segments.

[0106] Obstacle recognition data (per instance) is collected from the robot's infrared and camera data, and statistics are compiled within each time period.

[0107] The data is as follows:

[0108] The original light intensity data is 460;

[0109] Light intensity sequence (unit: lx):

[0110] ;

[0111] Obstacle recognition sequence:

[0112] ;

[0113] The absolute value of the deviation of each illumination value from the median is Take the maximum value ;

[0114] The absolute value of the obstacle recognition deviation from the median is Take the maximum value ;

[0115] Substitute parameters:

[0116] , , , , , ;

[0117] Calculate the summation term in the denominator:

[0118] ;

[0119] calculate :

[0120] ;

[0121] The results indicate that the illumination data in the first time period deviated from the median by 2.18 units of standardized deviation. When this value is higher than the threshold (e.g., set to 1.5), the data point can be identified as an outlier and replaced.

[0122] Please see Figure 6 The task reordering module includes:

[0123] The obstacle position correction submodule, based on the abnormal situation correction log, compares the old coordinates of obstacles in the original path map with the newly updated coordinates, identifies the difference between the two coordinates, and detects whether the offset exceeds the target range. If it does, it updates the obstacle position in the path map and generates the obstacle position offset range.

[0124] The system reads the current and corrected positions of all obstacles recorded in the anomaly correction log. It compares the original and new coordinates of each obstacle, determining the coordinate difference by calculating the straight-line distance between the two points. This distance is calculated using the Pythagorean theorem, by adding the squares of the differences in the horizontal and vertical coordinates and taking the square root. For example, if the original coordinates are (10, 20) and the new coordinates are (15, 25), the coordinate difference is calculated as √((15-10)² + (25-20)²) = √(25+25) = √50 ≈ 7.07. Then, based on a set offset threshold (usually determined by the robot...), the system... The spatial layout complexity of the work scenario determines the offset threshold. For example, in a spacious factory environment, the offset threshold is set to 5 units of length, while in a narrow warehouse environment, the threshold is set to 3 units of length. It is determined whether the calculated offset exceeds this threshold. If it does, the position of the obstacle in the path map will be updated, and the obstacle will be marked as having been corrected. All marked obstacle positions will be redrawn in the path map. For example, in a factory environment, the calculated offset is 7.07, which exceeds the threshold of 5. Therefore, the obstacle position in the original path map is updated from (10, 20) to (15, 25), and the obstacle position offset is generated.

[0125] The path priority recalculation submodule calls the obstacle position offset magnitude, recalculates the optimal path of the key task points associated with the obstacle, analyzes the impact of path length changes on task execution, and rearranges the path priorities according to the task number and task execution time to obtain the navigation priority configuration.

[0126] First, iterate through all path nodes with updated obstacle positions. Based on the new obstacle positions, recalculate the paths between critical task points associated with the obstacles. The calculation method relies on A* or Dijkstra's algorithm. The shortest path selection algorithm calculates the path length based on the distance between nodes and the transfer cost. For example, in a small factory with 5 critical task points, if the obstacle moves from point (10, 20) to (15, 25), the paths from the entry point (0, 0) to each task point will be recalculated. It is found that the original path to task point (30, 30) is no longer the shortest due to the new obstacle position. Therefore, a new shortest path is found, bypassing through (20, 20). The impact of the new path length on the execution of each task will then be analyzed. A longer path will increase the task execution time. This information, combined with the task number and the scheduled execution time, will be used to rearrange the execution order and priority of all tasks. The priority adjustment will be based on the increase or decrease in path length and the urgency of critical tasks. For example, if the execution time of task 1 is delayed due to the increased path length, the priority of task 2 will be advanced. The generated navigation priority configuration will reflect the changes, ensuring that all tasks are executed in the optimal order in the adjusted obstacle environment. The new task execution sequence and priority will be applied to the robot's navigation system to guide its actual movement path and task execution strategy.

[0127] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An AI modular perception robot service system based on precision sensing, characterized in that, The system includes: The environmental perception module acquires the light intensity and number of obstacles in the robot's work area, uses a photoelectric sensor array to collect light data in real time, counts the number of times obstacles are identified in the omnidirectional wheel navigation path grid, determines the fluctuation range of the two data types, and generates a perception data fluctuation index. The sensing sensitivity adjustment module analyzes the fluctuation of light intensity based on the sensing data fluctuation index. If the fluctuation amplitude exceeds the standard configuration range, the infrared sensor parameters are adjusted to a high-sensitivity state, and the power consumption of the image sensor is optimized to obtain the sensing configuration state. Based on the sensor configuration state, the sensor performance optimization module analyzes the fluctuation data of light intensity and obstacle detection frequency, and adjusts the exposure time and signal sampling frequency according to the analysis results to obtain the sensor adjustment range; The sensor performance optimization module includes: The illumination frequency analysis submodule, based on the sensor configuration status, calls the current infrared and image sensor configurations, compares the illumination intensity and obstacle detection frequency measurement data in a continuous time period, analyzes the degree of joint change of illumination and obstacle detection frequency, determines the data fluctuation characteristics, and obtains illumination frequency fluctuation data. The sensor adjustment submodule analyzes the conformity between the current exposure time and the sampling frequency based on the light frequency fluctuation data, adjusts the adjustable range of the sensor according to the analysis results, updates the sensor range limit, matches real-time environmental changes, and obtains the sensor adjustment range. The environmental data repair module analyzes the sensor adjustment range, identifies data sequences that deviate from the normal fluctuation range based on the fluctuation range of light intensity and obstacle recognition data, processes abnormal data segments by data sorting and median extraction, replaces unstable data segments, and generates an anomaly correction log.

2. The AI ​​modular perception robot service system based on precision sensing according to claim 1, characterized in that, The sensing data fluctuation indicators include illumination level and obstacle recognition frequency; the sensing configuration status includes sensitivity level and power consumption standard; the sensing adjustment range includes exposure parameter adjustment information and frequency setting; and the abnormal situation correction log includes data stability information and correction details.

3. The AI ​​modular perception robot service system based on precision sensing according to claim 1, characterized in that, The environment sensing module includes: The illumination acquisition submodule acquires the illumination intensity and the number of obstacles in the robot's working area. It uses the illumination data collected by the photoelectric sensor array to compare with the array trigger boundary to determine whether the illumination intensity exceeds the boundary range, and filters out the number of illumination points that deviate from the boundary to obtain the illumination offset. The obstacle recognition submodule collects the trigger frequency of obstacles at path nodes based on the omnidirectional wheel navigation path grid, determines whether the number of triggers meets the obstacle recognition standard based on the boundary trigger position data of the photoelectric sensor, and extracts the number of path nodes occupied by the identified obstacle to obtain the obstacle trigger count. The fluctuation statistics submodule, based on the number of illumination offsets and the number of obstacle triggers, calls statistical data of the two types of data over multiple sensing cycles, analyzes the amplitude, direction differences and frequency fluctuations in the time series, and generates a sensing data fluctuation index.

4. The AI ​​modular perception robot service system based on precision sensing according to claim 1, characterized in that, The sensing sensitivity adjustment module includes: The fluctuation recognition submodule obtains light intensity detection data based on the perceived data fluctuation index, calculates the rate of change of light intensity within a continuous time period, compares it with the light fluctuation reference interval, determines whether there is a fluctuation shift, and obtains the light fluctuation recognition result. Based on the light fluctuation recognition results, the parameter adjustment submodule evaluates whether the current sensitivity of the infrared sensor needs to be adjusted. If the light fluctuation exceeds the standard configuration range, the infrared sensor parameters are adjusted to a high-sensitivity state to obtain the infrared sensitivity adjustment results. The power consumption control submodule calls the infrared sensitivity adjustment result, and adjusts the image sensor's processing frame rate, grayscale level number and compression encoding according to the current power consumption configuration status of the image sensor to optimize power consumption and obtain the sensor configuration status.

5. The AI ​​modular perception robot service system based on precision sensing according to claim 1, characterized in that, The environmental data repair module includes: The range determination submodule analyzes the sensing adjustment range, extracts the numerical variation amplitude in adjacent time periods based on light intensity and obstacle recognition data, determines whether it exceeds the upper and lower bounds of the normal fluctuation range, marks the time points that deviate from the fluctuation range, and generates an abnormal fluctuation time series. The anomaly extraction submodule calls the abnormal fluctuation time series, collects light intensity data and obstacle recognition data within the corresponding time period, sorts the two data sequences in ascending order by value, removes the first and last values, takes the median of the remaining data, and generates a reference median set. The data replacement submodule calculates the difference between the original light intensity data and obstacle recognition data within the abnormal fluctuation period based on the reference median set, calculates the deviation from the median, and obtains the abnormal situation correction log.

6. The AI ​​modular perception robot service system based on precision sensing according to claim 1, characterized in that, The system also includes: The task reordering module corrects the updated obstacle position data in the log based on the abnormal situation, recalculates the associated coordinate points in the task path map, and rearranges the path priority according to the task number, execution time, and path length to obtain the navigation priority configuration. The navigation priority configuration includes priority sequence, coordinate point adjustment results, and time allocation results.

7. The AI ​​modular perception robot service system based on precision sensing according to claim 6, characterized in that, The task reordering module includes: The obstacle position correction submodule, based on the abnormal situation correction log, compares the old coordinates and the newly updated coordinates of the obstacles in the original path map, identifies the difference between the two coordinates, detects whether the offset exceeds the target range, and if it does, updates the obstacle position in the path map and generates the obstacle position offset range. The path priority recalculation submodule calls the obstacle position offset magnitude, recalculates the optimal path of the key task points associated with the obstacle, analyzes the impact of path length changes on task execution, and rearranges the path priorities according to the task number and task execution time to obtain the navigation priority configuration.

Images