Automatic charging processing method and device for charging robot

Abstract

The invention discloses an automatic charging processing method and device for a charging robot, and relates to the technical field of robot energy management, and the method comprises the following steps: constructing an acoustic array, collecting high-frequency background sound wave signals in a working area of the charging robot in real time, and extracting sound spectrum energy distribution; and identifying an interference frequency band coinciding with the working frequency of the ultrasonic sensor, and calculating the intensity level of the interference frequency band according to the sound spectrum energy intensity and the duration time in the interference frequency band. According to the method, a closed-loop system of interference perception-risk assessment-control response is constructed, and dynamic adjustment of perception confidence and output of blindness risk indexes are realized through multi-modal identification consistency analysis and historical deviation tracking. Once the threshold value is exceeded, the system executes frequency hopping and strategy optimization, frequency band shielding is effectively avoided, the sensing stability, path accuracy and operation safety of the charging robot in the sound interference environment are improved, and good application prospects are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot energy management, and in particular to an automatic charging processing method and device for a charging robot. Background Art

[0002] Automatic charging by charging robots refers to the use of intelligent robots with autonomous navigation and recognition capabilities to automatically complete the entire charging process, from target identification, path planning, docking calibration, to energy transfer, without human intervention. This process typically includes multiple steps, including environmental perception, charging target positioning, precise docking interface identification and adjustment, charging connection establishment, charging parameter control, and charging status monitoring. It aims to improve charging efficiency, reduce labor costs, and enhance the flexibility and intelligence of charging operations. It is widely used in automatic charging scenarios such as electric vehicles, unmanned delivery vehicles, and mobile devices.

[0003] The existing technology has the following shortcomings: During the automatic charging process of the charging robot, obstacle detection and path avoidance are usually performed by ultrasonic sensor modules. However, in some typical industrial operation scenarios, such as processing workshops and fan rooms with large fans, pneumatic equipment or high-frequency mechanical vibration sources, high-frequency sound wave signals with a frequency range that highly overlaps with the operating frequency of the ultrasonic obstacle avoidance module are generated. Because this type of sound wave interference has the characteristics of non-directionality, continuity and environmental coupling, it is easy to cause noise superposition or frequency band shielding in the signal receiving channel of the robot's ultrasonic sensor, resulting in the sensor module being unable to accurately receive the ultrasonic echo signal emitted by itself, and then the phenomenon of perception blindness occurs, in which the environmental perception function temporarily fails. This problem is not easy to be discovered in time during actual operation, especially when the robot is making fine identification of the edges of complex obstacles or positioning the path of high-precision docking interfaces. If the above-mentioned interference phenomenon persists, the robot may misjudge the position of the obstacle and directly cause a high-speed collision without effective deceleration or path adjustment. This may not only easily cause damage to the outer shell structure of the charging robot, but may also cause serious problems such as structural dislocation of the high-voltage charging interface, insulation damage, and even arc breakdown, which seriously threatens the operational stability and safety of the charging system.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The present invention aims to provide a method and device for automated charging of a charging robot. By constructing a closed-loop "interference perception-risk assessment-control response" system, the system dynamically adjusts perception confidence and outputs a blindness risk index through multimodal recognition consistency analysis and historical deviation tracking. Once a threshold is exceeded, the system performs frequency hopping and strategy optimization to effectively avoid frequency band masking, improving the charging robot's perception stability, path accuracy, and operational safety in acoustic interference environments. This method offers promising application prospects and addresses the aforementioned technical challenges.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a charging robot automatic charging processing method, comprising the following steps:

[0007] S1: Build an acoustic array to collect high-frequency background sound wave signals in the charging robot's operating area in real time, extract the sound spectrum energy distribution, identify the interference frequency band that coincides with the ultrasonic sensor's operating frequency, and calculate the interference frequency band intensity level based on the sound spectrum energy intensity and duration within the interference frequency band;

[0008] S2, based on the identified interference frequency band and interference frequency band intensity level, combined with the current position of the charging robot and path node parameters, establishes an acoustic interference prediction model and generates an acoustic interference spatial projection map;

[0009] S3, obtains the obstacle recognition results of the lidar, visual recognition, and ultrasonic sensors, calculates the recognition consistency deviation index based on the spatial deviation between the recognition results, and records the number of recognition deviation events to obtain the historical number of recognition deviations;

[0010] S4, integrates the recognition consistency deviation index with the acoustic interference space projection map for analysis, constructs the perception confidence level adjustment factor, and dynamically adjusts the credibility weight of each recognition result;

[0011] S5, based on the perception confidence level adjustment factor, the interference frequency band intensity level and the number of historical recognition deviations, builds a multi-factor perception blindness risk assessment model and outputs the perception blindness risk index;

[0012] S6 compares the perceived blindness risk index with the preset risk threshold. When the perceived blindness risk index is greater than the preset risk threshold, frequency hopping control is executed to switch the ultrasonic sensor operating frequency to a frequency range that does not include the current interference frequency band. The echo signal processing strategy is synchronously adjusted based on the frequency parameters after the hopping to achieve continuous and stable output of obstacle recognition results.

[0013] Preferably, step S1 includes:

[0014] Construct an acoustic array consisting of multiple microphone units and collect high-frequency background sound wave signals in the charging robot's current operating area in real time;

[0015] The high-frequency background sound wave signal is subjected to high-pass filtering and time-domain windowing, and a fast Fourier transform is performed within a fixed time sampling window to obtain the spectrum energy distribution;

[0016] The spectrum is divided into multiple sub-bands, and the energy peak, mean, and variance indicators are extracted to determine whether there is an interference frequency band that overlaps with the operating frequency of the ultrasonic sensor;

[0017] The interference frequency band intensity level is calculated based on average energy density, peak duration, spectrum fluctuation stability and multi-channel coupling consistency.

[0018] Preferably, step S2 includes:

[0019] Based on the interference frequency band and the interference frequency band intensity level, obtain the equivalent radiation range of the interference source and the duration of the interference source impact;

[0020] Combine the charging robot's current position, movement speed, orientation angle, and path node parameters to construct a path trajectory prediction segment set;

[0021] Map the equivalent radiation range of the interference source to the set of path trajectory prediction segments, and calculate the duration of the interference source impact in each trajectory segment;

[0022] Based on the path trajectory, the segment set and the duration of interference source impact are predicted, and an acoustic interference spatial projection map with interference intensity labels and high failure-sensitive node identifiers is generated.

[0023] Preferably, step S3 includes:

[0024] Obtain the obstacle recognition results output by the lidar module, visual recognition module, and ultrasonic sensor module, and uniformly convert each recognition result into the reference coordinate system of the charging robot body;

[0025] Match each recognition result according to the target fusion window parameters, and calculate the deviation value of each recognized target in spatial position and boundary contour to obtain the recognition consistency deviation index;

[0026] Determine whether the recognition consistency deviation index exceeds the preset confidence interval, and record it as a recognition deviation event if it exceeds the confidence interval;

[0027] The cumulative number of identification deviation events is counted within the set operation cycle to obtain the historical number of identification deviations.

[0028] Preferably, step S4 includes:

[0029] Based on the acoustic interference spatial projection map and the recognition consistency deviation index, a local grid is constructed within the charging robot's path prediction area, and the perception anomaly weight of each grid is calculated.

[0030] In each grid, the confidence adjustment rule table is called to adjust the confidence of each type of sensor based on the perception anomaly weight, acoustic interference level, recognition deviation status and historical recognition deviation count;

[0031] Integrate the sensor confidence adjustment results in each grid into the path node to generate a perception confidence level adjustment factor that includes the three types of sensor confidence weights;

[0032] The perception confidence level adjustment factor is updated in each navigation control cycle to guide the weighted fusion output of multimodal obstacle recognition results.

[0033] 6. The automatic charging processing method of a charging robot according to claim 1, wherein step S5 comprises:

[0034] Obtain the perception confidence level adjustment factor, interference frequency band intensity level, and historical identification deviation count of the path node, and construct a perception vulnerability label;

[0035] For path nodes marked as high concern, a risk score is calculated, including the degree of confidence imbalance, the proportion of interference frequency band coverage time, and the trend of identification deviation changes;

[0036] Cluster analysis was performed on the risk scores of the path nodes to extract continuous high-risk segments and calculate the probability factor of perceptual blindness;

[0037] The risk scores of all path nodes are weighted and integrated with the perceptual blindness probability factor to output the perceptual blindness risk index.

[0038] Preferably, when the perception blindness risk index output by the charging robot's perception system is greater than a preset risk threshold, the ultrasonic sensor frequency hopping control is immediately executed, and the echo signal processing strategy is synchronously and adaptively adjusted to achieve stable perception output in a complex acoustic interference environment. The specific steps are as follows:

[0039] Based on the risk index exceeding the limit, the frequency hopping response strength is calculated as the frequency offset amplitude of this frequency hopping. The parameters are defined as follows:

[0040] , where is the frequency hopping response strength, is the frequency hopping sensitivity coefficient, is the perception blindness risk index, which indicates the risk level of perception function breakdown calculated by the current system. is the preset risk threshold, is the acoustic interference time coverage ratio;

[0041] Based on the relative position of the current ultrasonic sensor operating frequency and the center frequency of the interference frequency band, the new operating frequency after the jump is calculated. The calculation formula is as follows:

[0042] , where is the new operating frequency after the jump, is the current operating frequency of the ultrasonic sensor, is the center frequency of the interference band, is a sign function;

[0043] By constraints:

[0044] ,in, is the lowest available frequency supported by the ultrasonic sensor, is the highest available frequency supported by the ultrasonic sensor, is a collection variable, It is a logical condition;

[0045] After successfully completing the frequency hopping switch, based on the new operating frequency after the hopping Dynamically adjust the echo signal processing strategy, including: echo sampling window length, echo gain adjustment coefficient and reconstruction of matched filter bandwidth parameters. The calculation expression is as follows:

[0046] , where is the echo sampling window length, is the sampling scale constant, is the echo gain adjustment coefficient, is the standard gain value at the original operating frequency, is the gain adjustment sensitivity factor, is the matched filter bandwidth adjustment value, is the bandwidth response expansion factor.

[0047] An automatic charging processing device for a charging robot includes an acoustic interference perception module, an acoustic interference prediction module, a multimodal recognition consistency assessment module, a perception confidence level adjustment module, a perception blindness risk assessment module, and a frequency hopping control and recognition strategy adaptation module:

[0048] The acoustic interference perception module builds an acoustic array to collect high-frequency background sound wave signals in the charging robot's operating area in real time, extract the sound spectrum energy distribution, identify the interference frequency band that coincides with the ultrasonic sensor's operating frequency, and calculate the interference frequency band intensity level based on the sound spectrum energy intensity and duration within the interference frequency band;

[0049] The acoustic interference prediction module establishes an acoustic interference prediction model based on the identified interference frequency band and interference frequency band intensity level, combined with the charging robot's current position and path node parameters, and generates an acoustic interference spatial projection map;

[0050] The multimodal recognition consistency evaluation module obtains the obstacle recognition results of lidar, visual recognition, and ultrasonic sensors, calculates the recognition consistency deviation index based on the spatial deviation between the recognition results, and records the number of recognition deviation events to obtain the historical number of recognition deviations;

[0051] The perception confidence level adjustment module integrates the recognition consistency deviation index with the acoustic interference space projection map for analysis, constructs the perception confidence level adjustment factor, and dynamically adjusts the credibility weight of each recognition result;

[0052] The perceptual blindness risk assessment module builds a multi-factor perceptual blindness risk assessment model based on the perceptual confidence level adjustment factor, interference frequency band intensity level, and historical recognition deviation times, and outputs a perceptual blindness risk index.

[0053] The frequency hopping control and recognition strategy adaptive module compares the perception blindness risk index with the preset risk threshold. When the perception blindness risk index is greater than the preset risk threshold, it performs frequency hopping control, switches the ultrasonic sensor operating frequency to a frequency range that does not include the current interference frequency band, and synchronously adjusts the echo signal processing strategy based on the frequency parameters after hopping to achieve continuous and stable output of obstacle recognition results.

[0054] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0055] The present invention constructs a real-time, quantitative perception confidence level adjustment mechanism through multimodal recognition consistency analysis and historical deviation tracking, and ultimately outputs a perception blindness risk index that can be used for control decisions. Once the risk index exceeds the threshold, the system will immediately perform frequency hopping and adaptive optimization of the echo strategy, thereby effectively avoiding the frequency band masking phenomenon and maintaining the continuity and accuracy of obstacle recognition capabilities. This method constructs an overall "interference perception-risk assessment-control response" closed-loop system, which significantly improves the robot's charging positioning success rate, path execution accuracy, and system safety fault tolerance in acoustic interference environments, and has good industrial application prospects and engineering promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0057] Figure 1 This is a flow chart of a method for automatic charging processing of a charging robot according to the present invention.

[0058] Figure 2This is a module schematic diagram of an automatic charging processing device for a charging robot according to the present invention. DETAILED DESCRIPTION

[0059] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0060] The present invention provides Figure 1 and Figure 2 The automatic charging processing method of a charging robot shown includes the following steps:

[0061] An acoustic array is constructed to collect high-frequency background sound wave signals in the charging robot's current operating area in real time. Based on the collected data, the sound spectrum energy distribution is extracted to identify the interference frequency band that coincides with the ultrasonic sensor's operating frequency. The intensity level of the interference frequency band is calculated based on the sound spectrum energy intensity and duration within the interference frequency band.

[0062] To proactively identify and respond to high-frequency acoustic interference that may exist in the charging environment, an acoustic perception system based on a multi-element microphone array was constructed. This system then performs real-time acquisition of high-frequency background acoustic signals and identifies and processes interference frequency bands. This includes the following steps:

[0063] A circular distributed acoustic array consisting of at least six microphone units is embedded in the main structure of the charging robot. The array is evenly distributed on the front and sides of the robot to form a 360-degree acquisition structure with spatial coverage. The microphone array uses digital MEMS microphones with a bandwidth range of 20kHz to 80kHz and a signal-to-noise ratio greater than 70dB. Each microphone is connected to the robot's main control chip via the I2S bus to support parallel transmission of high-frequency sampling data. The system uses a 10-ms time sampling window to continuously record the raw amplitude data of all frequency component signals in the environment and cache it in the on-chip RAM for subsequent processing. To improve signal quality, the collected signal is first passed through a digital high-pass filter to remove the baseline of the frequency band below 20kHz before entering the analysis module, and a Heiman window function is used for time domain windowing to reduce the impact of spectral leakage.

[0064] The main control chip uses the embedded Fast Fourier Transform (FFT) module to perform a 1024-point FFT spectrum analysis on the sampled data within a single time window, obtaining a spectral energy distribution curve for the current ambient sound wave signal within the 20kHz to 80kHz range. To improve the accuracy of interference identification, a "dynamic energy threshold segmentation method" is introduced. This method divides the target spectrum into several sub-bands with a step size of 2kHz. Within each sub-band, the mean, peak, and variance of the spectral energy are calculated. The fluctuation trend within the continuous time window determines whether the band constitutes a candidate interference segment. If a sub-band exhibits a peak amplitude exceeding three times the static environmental mean of the corresponding sub-band within three consecutive sampling periods (30ms), and the band has similar amplitude characteristics across multiple channels in the array (with a fluctuation difference of less than 15%), it is marked as a "potential interference band" and proceeds to the next step of processing.

[0065] For sub-bands identified as potential interference bands, the interference intensity level is further calculated. This paper proposes an "interference band intensity grading model" that comprehensively considers the average energy density (in dB / Hz), peak duration (in milliseconds), spectral fluctuation stability, and multi-channel coupling consistency within the band. The average energy density is categorized as low, medium, and high, based on the standard average energy value of the ultrasonic sensor echo reflection. Peak duration is measured using a sliding time window method, with durations exceeding 50 milliseconds considered strong interference. Coupling consistency is measured using the multi-channel spectral cross-correlation coefficient, with a coefficient greater than 0.9 considered high consistency. Finally, a rule-based decision tree algorithm is used to weight these indicators and output a corresponding "interference band intensity level" label (e.g., Level-1 to Level-4). This level and the corresponding frequency band's start and end frequencies are recorded for use in subsequent acoustic interference spatial modeling and risk modeling modules.

[0066] To ensure data continuity and real-time response, the system has established a cyclical "dynamic monitoring task for environmental acoustic interference." This task repeats the aforementioned acquisition and identification process every 10ms, comparing the previous cycle's results with the current one. If the same frequency band is identified as a high-level interference band (Level 3 or higher) for three consecutive cycles, the acoustic interference prediction engine is automatically triggered. In practice, this dynamic acoustic interference identification module works in conjunction with the robot's path planning module to predict the acoustic interference area the robot is about to enter before path nodes are updated. This allows for dynamic adjustments to the perception system and optimization of protection strategies in advance, effectively preventing environmental perception failures caused by interference blocking the ultrasonic sensor's echo signal.

[0067] Based on the identified interference frequency bands and interference frequency band intensity levels, combined with the current position of the charging robot and path node parameters, an acoustic interference prediction model is established to predict the spatial distribution range of acoustic interference in the future period and generate an acoustic interference spatial projection map;

[0068] To enable the charging robot to spatially predict high-frequency acoustic interference that may be encountered in its future work path and ensure perception stability, an acoustic interference prediction model was constructed based on interference frequency band identification results, charging robot path parameters, and spatial propagation modeling. Based on this, an acoustic interference spatial projection map was generated to guide subsequent perception confidence correction and frequency dynamic adjustment. Specifically, the model includes the following four steps:

[0069] After completing the collection of high-frequency background sound waves and identifying the interference frequency band and intensity level, the center frequency, frequency band range, and corresponding interference frequency band intensity level of the currently identified interference frequency band are passed as input variables to the acoustic interference prediction module. Based on the physical characteristics of sound propagation, the prediction module defines two intermediate parameters: "interference source equivalent radiation range" and "interference source impact duration": The interference source equivalent radiation range refers to the maximum distance range within which the sound energy of this frequency band can effectively propagate in the air and affect the identification of ultrasonic echoes at the current sound intensity level. This is specifically obtained through a table matching sound intensity levels with environmental structure types. The interference source impact duration refers to the continuous time period during which the robot is under the influence of this sound energy interference during movement. Its calculation method is based on the spatial span of the interference area and the current operating speed of the robot. It is used to reflect the expected length of time the robot will be exposed to the interference area, providing time dimension support for subsequent risk assessment and control strategy selection.

[0070] Using the charging robot's current spatial coordinates, movement speed, heading angle, and the geometric positions of consecutive path nodes along a pre-set path, a "path prediction segment set" for the future is constructed. This is the set of spatial positions the robot may occupy within a pre-set prediction time window (e.g., 2-3 seconds). This path segment set serves as the core reference path for interference modeling, projecting the spatial distribution of acoustic interference. The equivalent radiation range of the interference source is mapped onto a spherical or ellipsoidal surface, and the acoustic interference level attribute of that spatial region is determined based on the intensity level of the identified frequency band. If at least one trajectory point in the path prediction segment set falls within the interference coverage area, the robot is considered to enter the acoustic interference area within that time window. Furthermore, by traversing the total number of consecutive trajectory points within the interference coverage area and combining it with the robot's displacement speed, the predicted "interference source impact duration" of the robot in the interference area is calculated. If this duration exceeds a preset threshold (e.g., 500 milliseconds), the system marks the path segment as a "continuous interference exposure zone" and assigns a higher-weighted risk level label in the projection map.

[0071] To improve the accuracy and real-time performance of spatial distribution prediction, environmental map information (such as 2D SLAM mapping results or static obstacle distribution maps) is further incorporated to correct the spatial propagation of acoustic interference. Taking into account the varying degrees of reflection and scattering effects of sound waves on metal walls, concrete structures, and glass surfaces, a "structural reflection coupling parameter" is introduced. This parameter dynamically adjusts the spatial shape of acoustic interference based on the regional material type and spatial shape characteristics. For example, when the identified area is an enclosed fan room or a narrow corridor surrounded by concrete walls, the system increases the sound propagation radius by a certain multiple and also increases the duration of the acoustic reverberation. Based on the duration of the interference source's impact, this correction mechanism also dynamically adjusts the width and buffer boundary of the acoustic interference coverage time window to ensure that interference signals that persist in the reflective environment for a long time are not misclassified as transient events. Furthermore, the equivalent propagation area after structural reflection correction is integrated with the charging robot's predicted trajectory segment set to generate acoustic interference spatial prediction data consisting of a "time-space-intensity" triple.

[0072] The structural reflection coupling parameter is a technical parameter used in this paper to describe and adjust the impact of environmental structures on the propagation characteristics of high-frequency sound waves. Essentially, it is a quantitative indicator used to model and compensate for the propagation path, reflection behavior, and interference persistence of sound waves within a specific spatial structure. Its primary function is to dynamically adjust the propagation radius, reflection superposition interval, and reverberation time of the interference wave in space when generating the acoustic interference space prediction model. This avoids underestimating the actual interference range due to ignoring sound wave reflection and structural coupling effects, thereby improving the prediction accuracy and stability of the acoustic interference space projection map.

[0073] The determination of this parameter is based on a comprehensive evaluation of the following three factors:

[0074] (1) Spatial structural material type: Materials with high acoustic reflection coefficients, such as metal walls, glass windows, and high-density concrete, will be assigned higher reflection coupling weights;

[0075] (2) Spatial geometric features: structures such as narrow and long passages, closed-loop pipes, and box-shaped rooms are prone to multiple echo superposition, which increases the residence time of sound waves and requires an increase in the coupling parameter level;

[0076] (3) Historical acoustic signal reverberation characteristics: By analyzing the waveform attenuation curve of the collected background acoustic signal, if a high-persistence tail wave signal is detected, it means that the structure has a strong effect on the storage and reflection of acoustic energy, and the parameter level needs to be adjusted upward.

[0077] In summary, the structural reflection coupling parameters dynamically adjust the coverage range and interference persistence prediction results of the acoustic interference spatial model by mapping material, spatial geometry and acoustic reverberation data, thereby achieving a modeling capability that is closer to the acoustic interference propagation characteristics in real industrial scenarios. It is an important compensation mechanism to ensure the accuracy and robustness of the spatial prediction model of the present invention.

[0078] The acoustic interference space prediction data is graphically projected into the local map coordinate system of the charging robot to construct an acoustic interference space projection map. This map, centered on the robot's current position, covers all path points within a certain distance in the future. Each path node is labeled as to whether it is in a potential acoustic interference impact zone, along with an impact level label (e.g., low, medium, high, or very high). When a path point is continuously located in an interference zone, and the duration of the corresponding interference source impact exceeds a set threshold, the node is dynamically labeled as a "highly failure-sensitive node," providing a time-weighted input for the perception confidence adjustment module. This map is updated in a rolling manner as a dynamic data structure, with each prediction result retaining the most recent 3 seconds of coverage prediction, which is used to drive subsequent perception confidence factor adjustments, risk assessment model calculations, and trigger criteria input for frequency hopping control.

[0079] Obtain the obstacle recognition results output by the laser radar, visual recognition, and ultrasonic sensors, and calculate the recognition consistency deviation index based on the spatial deviation between the recognition results. Record the number of recognition deviation events that occur in continuous operation cycles to obtain the historical number of recognition deviations.

[0080] To assess the accuracy of multimodal sensor perception results and identify potential mismatch risks in advance, a recognition deviation monitoring mechanism based on consistency analysis of obstacle recognition results was constructed. This mechanism analyzes the spatial deviations between the obstacle recognition results output by the lidar module, visual recognition module, and ultrasonic sensor module, extracts consistency deviation indicators, and counts the frequency of deviation events within continuous operation cycles, thereby obtaining a historical number of recognition deviations that reflects the long-term stability of the perception system. This implementation includes the following steps:

[0081] First, the laser radar module, visual recognition module and ultrasonic sensor module configured on the charging robot are called separately to obtain the obstacle recognition results output by each of them. Each type of sensor will generate a perception data set about the obstacle position in a single scanning cycle, including the two-dimensional or three-dimensional coordinates, boundary contours, reflection characteristics, etc. of the target. In order to achieve alignment between the outputs of different sensors, the system first converts the respective coordinate information into the reference coordinate system of the robot body to ensure that all recognition results are compared under the same spatial reference. At the same time, taking into account the differences in the perception accuracy of different sensors for the same obstacle, the system defines a unified "target fusion window" parameter, which is used to preliminarily associate each identified target within the spatial neighborhood, that is, the identification points within the preset distance threshold of the three types of sensors are regarded as candidate matching groups for the same target.

[0082] The "target fusion window" parameter is a key spatial threshold control parameter defined to achieve spatial matching and fusion of targets identified by different sensors when processing multimodal obstacle recognition results. Its function is to determine a spatial neighborhood with a single identified target as the reference center within the robot's unified coordinate system. This is used to determine whether obstacles identified by other sensors can be considered the same object as the identified target. The following steps are involved: First, a three-dimensional spherical neighborhood with a fixed radius (e.g., 0.3 meters) is constructed, using the center point of the identified target output by the lidar as the reference point, as the initial fusion window. Second, this neighborhood is checked to see if the center points of the identified targets from the visual recognition module and the ultrasonic sensor module fall within this area. Third, if both fall within the fusion window, the three recognition results are matched into the same obstacle recognition group. Finally, the window radius is dynamically adjusted based on historical fusion accuracy and recognition error distribution, allowing for a more relaxed matching tolerance in high-interference scenarios while maintaining fusion accuracy in delicate work areas. By setting the "target fusion window", the tolerance mechanism can be used to handle small deviations between different sensors, effectively improving the stability of obstacle recognition and the consistency of data fusion, and providing a basic premise for the subsequent extraction of recognition consistency deviation indicators.

[0083] For multimodal recognition targets that have completed preliminary matching, the spatial deviation of each recognition result on the target center point coordinates is calculated, and based on this, the recognition consistency deviation index is constructed. This deviation index includes two core calculation dimensions: one is the target center coordinate deviation value, which measures the difference in Euclidean distance between the same target center point detected by the three sensors; the second is the recognition boundary shape overlap, which measures the degree of matching between different recognition results in the spatial coverage range. The present invention introduces the concept of "recognition deviation confidence interval" to calibrate whether the deviation value exceeds the allowable error range. When any deviation in the above two dimensions exceeds the corresponding confidence interval, the system determines that there is a consistency anomaly in the recognition event, records it as a recognition deviation event, and records the corresponding deviation value, target position and timestamp.

[0084] The process of constructing a recognition consistency deviation index aims to quantitatively evaluate the spatial matching quality of obstacle recognition results output by the three perception modules of LiDAR, visual recognition, and ultrasonic sensors, thereby determining whether there is potential recognition bias in the multi-source perception system. The construction of this index includes the following two steps:

[0085] The center coordinates of the obstacle targets identified by the three sensors within the same sampling period are extracted and uniformly converted to the charging robot's reference coordinate system. A preliminary matching process is then performed based on spatial proximity rules to confirm whether the three sensors have identified the same target. After matching, the Euclidean distance difference between the target center points of each pair of sensors is calculated, and the maximum value among the three pairs is taken as the target center coordinate deviation. This step measures the consistency of each sensor's recognition of the same object's position. Larger deviations indicate inconsistent positional recognition results, indicating potential spatial recognition errors.

[0086] The boundary range information of each target recognition outline output by the three types of sensors is extracted and uniformly mapped to a two-dimensional spatial grid. Using the area overlap ratio calculation method, the ratio of the boundary intersection to the boundary union between each pair of sensor perception areas is analyzed. The overlap ratio of the three pairs of combinations is calculated, and the minimum value is taken as the overlap ratio of the recognition boundary shape. This step is used to evaluate the consistency of obstacle shapes recognized by different sensors. In particular, it can reveal whether there is visual bias or edge recognition failure in the case of blurred edges, size changes, or partial occlusion of objects.

[0087] The target center coordinate deviation and the overlap of the identified boundary shape are combined as components of the recognition consistency deviation index, and preset threshold intervals are set. When any of these indicators exceeds the corresponding trustworthy range, the system determines that the recognition event has a consistency anomaly, records it as a recognition deviation event, and enters the historical recognition deviation count process. The establishment of this index provides a key judgment basis for subsequent perception confidence factor adjustment and perception mismatch area identification, effectively enhancing the robot's perception reliability and failure response capabilities in multi-source environments.

[0088] To determine the stability trend of the recognition system during continuous operation, the system continuously monitors the frequency of recognition deviation events within a set operating cycle window (e.g., the past 60 seconds) and counts the cumulative number of recognition deviations that occurred during that time period to obtain the historical number of recognition deviations. As a key indicator for measuring perception stability and the impact of environmental interference, the historical number of recognition deviations has both time continuity and trend indicative properties. Furthermore, the system categorizes and records the historical number of recognition deviations, including deviations caused by specific sensors (e.g., persistently low ultrasonic perception, large visual recognition offsets, etc.), to assist in determining whether a sensor module's perception capability has declined due to external interference or its own fault.

[0089] Based on the recognition consistency deviation index and the number of historical recognition deviations, the perception status assessment table is dynamically updated, and this data is used as the basic input for constructing subsequent perception confidence level adjustment factors. In particular, within the area covered by the acoustic interference space projection map, once recognition consistency deviations occur frequently and the cumulative number of historical recognition deviations exceeds the threshold (e.g., more than 10 consecutive deviations occur in the same sensor-dominated path segment), the area is determined to have a high mismatch risk. The subsequent strategy control module will increase the redundant perception trigger frequency or switch the main perception channel based on this result. Through this mechanism, the charging robot can perceive the operational consistency status of the multi-source recognition system in real time, and promptly identify local perception collapse trends caused by high-frequency interference, obstructions, lighting changes, etc., thereby providing accurate data basis for dynamic perception adjustment, path planning correction, and frequency hopping control.

[0090] The recognition consistency deviation index is integrated with the acoustic interference space projection map for analysis to construct a perception confidence level adjustment factor and dynamically adjust the credibility weight of each recognition result.

[0091] To address the dynamic imbalance in confidence levels among multi-source obstacle recognition results for charging robots in complex acoustic interference environments, a dynamic adjustment mechanism for perception confidence levels was designed that integrates the recognition consistency deviation index with the acoustic interference spatial projection map. This mechanism constructs a local grid structure, combines spatial recognition consistency with acoustic interference level assessment results, and generates a confidence adjustment factor that dynamically adjusts multimodal perception weights, thereby improving the system's perception robustness and safety response capabilities in unstable interference areas. Specifically, it includes the following steps:

[0092] Based on the current location, posture, and predicted path of the charging robot, the system establishes a set of local reference grids. Using these grids as the base regional unit, the system binds the acoustic interference spatial projection map information and the recognition consistency deviation index to each grid cell. A parameter called "perception anomaly weight" is introduced for each grid cell. This weight is a scalar quantity that describes the degree to which the perception system within that cell is affected by the interference. Its calculation is based on the following:

[0093] (1) Whether the current grid is covered by the acoustic interference space projection map;

[0094] (2) The spatial deviation amplitude of each sensor output in the identification consistency deviation index within the region;

[0095] (3) The cumulative value of the number of historical recognition deviations within the range of the grid. The system assigns weighting factors to the above three factors respectively and sets dynamic thresholds. When the perception anomaly weight of a grid exceeds the set range, it is judged as a "perception high-risk area" and serves as the key input unit for the next step of credibility weight correction.

[0096] The perception anomaly weight is a key intermediate parameter used to assess the stability and interference risk of the charging robot's perception system within a local path area. It primarily drives the subsequent calculation of the perception confidence level adjustment factor. This weight, expressed numerically, represents the comprehensive degree of perception uncertainty within each local grid cell in the current environment. Higher values ​​indicate greater perception mismatch risk. Its setting involves the following specific steps: First, the system determines whether the current grid cell is within the interference coverage area based on the real-time acoustic interference spatial projection map. If so, an initial interference weight is assigned. Second, the recognition consistency deviation index is used to extract the spatial deviation between the outputs of various sensors within the grid cell area and determine whether it exceeds a preset deviation threshold. If so, the weight is increased. Third, the frequency of recognition anomalies in the grid cell over a certain period of time is calculated based on the number of historical recognition deviations. This frequency is then mapped into a historical stability index, which is then added to the current weight. Finally, the system integrates these three factors (interference status, real-time recognition deviation, and historical stability) according to a preset weighting factor to output the final perception anomaly weight for the grid cell. The perception anomaly weight, as one of the input parameters of the subsequent confidence adjustment rule table, directly affects the weight ratio adjustment of various sensors in the perception fusion process, which helps to achieve dynamic fault-tolerant control of areas with high mismatch risk, thereby enhancing the recognition stability and adaptability of the entire system in complex interference environments.

[0097] The recognition results of various sensors in the current local grid are uniformly spatially mapped and classified into corresponding grid cells. Based on this, the system calls a preset "confidence adjustment rule table" to dynamically adjust the confidence assigned to each sensor type under different combinations of interference levels and recognition deviations. This confidence adjustment rule table is a nonlinear configuration matrix proposed in this invention. Its input variables include: the perception anomaly weight level, the current acoustic interference intensity level, whether the recognition consistency deviation exceeds a threshold, and the proportion of historical recognition deviations for that sensor type. The rule table outputs a floating-point correction factor, representing the adjustment factor for the current perception confidence. For example, if the ultrasonic sensor in a grid cell is located in a high-intensity acoustic interference area and the spatial deviation exceeds a set threshold, the rule table may output a correction factor of 0.3, reducing its current confidence to 30% of its original value. Meanwhile, if the lidar module is located in a low-interference area and has high recognition stability, a correction factor of 1.2 may be output to increase its weight in the fusion calculation.

[0098] The preset confidence adjustment rule table is a decision matrix used in this invention to dynamically adjust the perception confidence weights of multiple sensor types. It aims to provide targeted confidence corrections to perception results based on the level of interference and recognition stability in the current environment. This rule table, in the form of a table or mapping structure, takes inputs including multiple quantifiable parameters, such as the perception anomaly weight level, the interference intensity level in the acoustic interference spatial projection map, whether the current recognition consistency deviation exceeds a threshold, the number of historical recognition deviations, and the sensor type. Each input combination corresponds to an output value: a floating-point confidence correction factor, typically between 0 and 1.5, indicating whether the current sensor perception weight should be reduced (<1), maintained (=1), or appropriately enhanced (>1). For example, if a grid cell is located in a high-intensity interference area, the ultrasonic recognition results in that area exhibit significant deviations, and the number of historical recognition deviations is high, the rule table might output a correction factor of 0.3, indicating that the ultrasonic sensor weight in that area should be reduced to 30% of its original value. The role of this rule table is to provide a logical basis for the generation of perception confidence level adjustment factors, realize the hierarchical trust allocation of sensor perception capabilities in complex scenarios, avoid the decline in overall recognition accuracy due to local failure of a single sensor, and thus significantly improve the robustness and adaptability of the multimodal perception system.

[0099] Using a path node as an index, the adjustment results of various sensors within the grid cell are aggregated to construct a "perception confidence level adjustment factor" at the path node level. This adjustment factor is a vector parameter composed of the confidence weights of three types of sensors, which is used to describe the distribution of trust relationships within the perception system in the current node environment. The system recalculates and updates this factor during each navigation control cycle to ensure that the perception fusion weights used for path following and obstacle avoidance judgments are highly aligned with the actual environmental interference conditions. This adjustment factor is then passed to the multimodal perception fusion module, which uses a weighted approach to calculate the fusion output during the actual obstacle recognition process. This avoids undue trust in perception results with interfered signals, thereby reducing the probability of misjudgment.

[0100] To achieve closed-loop tuning and improve scene adaptation capabilities, a "confidence correction feedback mechanism" is introduced. After the charging robot completes the path execution task, the offset between the actual obstacle avoidance action results and the perception fusion prediction results is collected, and an error reverse calibration channel is established. Specifically, it includes:

[0101] (1) Compare the actual deviation trajectory of the robot with the deviation of the relative position of the obstacle in the fusion prediction;

[0102] (2) Determine whether a significant error occurs (e.g., the actual obstacle avoidance distance deviates from the predicted obstacle position by more than a threshold);

[0103] (3) When errors continue to occur, the weight coefficients of the calculation rules for the perception confidence level adjustment factor are backtracked and adjusted. By locating and analyzing the source of the error, the parameter configuration in the confidence adjustment rule table can be continuously optimized, so that it can form adaptive adjustment capabilities in long-term operation. Through this mechanism, even under complex changes in acoustic interference patterns or sudden environmental disturbances, relatively stable perception fusion judgment results can be maintained.

[0104] Based on the perception confidence level adjustment factor, interference frequency band intensity level, and historical recognition deviation times, a multi-factor perception blindness risk assessment model is constructed to output the perception blindness risk index of the charging robot under its current operating state.

[0105] To address the inability of traditional perception systems to quantify the risk of reduced recognition capabilities and predict perception failure scenarios, a multi-factor fusion-based perception loss risk assessment model was developed. This model provides real-time assessment of the stability of the charging robot's perception system under its current operating state and outputs a perception loss risk index. This risk index is used to drive downstream control strategies for early warning, perception reconstruction, or frequency hopping responses. This model includes the following four steps:

[0106] After calculating the perception confidence level adjustment factor, the confidence weight vector for each path node in the charging robot's current path prediction segment is obtained, and the interference frequency band intensity level and historical recognition deviation data are simultaneously called as basic input items for the risk modeling module. To achieve unified processing of feature fusion, the system formats these three types of data into a standard feature vector and introduces a "perception vulnerability label." This label is set based on the confidence weight extreme difference, the interference frequency band level threshold trigger state, and the cumulative value of the recognition deviation number. It is used to identify whether the current operating area has the characteristics of weak perception stability and high probability of misidentification. If the perception vulnerability label is triggered, the path node is marked as a "high-attention node" and enters the key risk calculation path.

[0107] The "perception vulnerability label" is a technical parameter introduced to identify and classify the stability of the local perception system of the charging robot in its current operating state in advance, thereby screening out key area nodes with perception degradation trends in the multi-factor perception blindness risk assessment model, and enhancing the risk assessment model's recognition sensitivity and early warning capabilities for high-risk areas.

[0108] The core significance of this label is that when the perception system exhibits multi-source perception imbalance, high interference intensity, and frequent historical misidentification in a certain area, it indicates that the perception capability in that area has entered an abnormally fluctuating state, known as "perception vulnerability." Without intervention, the robot is highly susceptible to "recognition blindness" or path misjudgment in that area. Therefore, the system prioritizes data from these areas in the risk assessment model and assigns them higher weight.

[0109] The label setting logic is based on the combined judgment of three input indicators:

[0110] Whether the range of the perception confidence level adjustment factor exceeds the set threshold: This is used to determine whether the confidence differences among the three types of sensors are significant, reflecting whether there is an imbalance within the perception system;

[0111] Whether the interference frequency band intensity level is continuously high: used to identify whether the environment where the current path node is located is affected by high-intensity acoustic interference for a long time;

[0112] Whether the number of historical recognition deviations exceeds the threshold within the set time window: reflects whether the actual error rate of the perception system is high in the recent period.

[0113] When at least two of the three conditions mentioned above are met simultaneously, a "perceived vulnerability label" is automatically added to the path node. This label is essentially a Boolean flag signal that triggers the risk model to perform a "focused assessment" processing path for the node.

[0114] By setting this label, the system can identify potential perceptual degradation areas in advance, providing strong feature constraints for the subsequent calculation of the "perceptual blindness risk index", thereby improving the risk prediction accuracy and real-time response capabilities of the entire model.

[0115] A multi-factor risk calculation structure is constructed for high-attention nodes. This structure, based on the combined effects of an expert rule model and a weight fusion mechanism, utilizes information from three dimensions: First, the maximum deviation value in the perception confidence level adjustment factor, representing the degree of confidence imbalance between the three sensor types; second, the proportion of time with continuous coverage of high-level frequency bands within the interference frequency intensity level, representing the robot's exposure time to the interference field; and third, the changing trend of the number of historical recognition deviations within the time window, used to measure the degree of degradation of the recognition system. These three types of data are jointly scored using a weight factor mapping table to generate a continuous risk score value, which is assigned to the path nodes to construct a node-level risk score map. This map, with the path time series as the horizontal axis and the risk value as the vertical axis, clearly shows the fluctuation distribution trend of the robot's perception ability along the future path.

[0116] Generating continuous risk scores and assigning them to path nodes to construct a node-level risk score map involves the entire process from multi-source data fusion, weighted scoring to path space mapping, specifically including the following steps:

[0117] Three types of input data corresponding to each path node in the current path prediction segment are extracted and normalized into numerical feature vectors: the perception confidence adjustment factor, the interference frequency band intensity level, and the number of historical recognition deviations. This step unifies multi-source heterogeneous data (with different units and dimensions) into a standard scoring dimension that can be used in the calculation, ensuring that all features have equivalent expressive power in the scoring model.

[0118] The three standardized features are entered into a weight factor mapping table for mapping calculations, generating a raw risk score for each path node. This mapping table assigns weights to each feature's impact on the overall risk of perceptual blindness based on prior empirical rules and field data training. For example, recognition bias may be weighted higher than interference intensity. This step extracts a basic quantitative result of the comprehensive risk level for each path node through weighted summation or rule-based judgment.

[0119] The raw risk score is smoothed and normalized to generate a continuous risk index. The system uses a sliding average window or multi-node interpolation method to locally smooth the risk score to eliminate jumps between path nodes caused by single-point anomalies. The final risk score is then normalized to a uniform range (e.g., 0-1). This step improves the continuity and trend representation of the risk distribution, making the risk score map more suitable for forward-looking judgments and dynamic control triggering.

[0120] The processed risk index value is assigned to each corresponding node in the path prediction structure, and a node-level risk score map is constructed. This map is represented as a visual or structured matrix, with the path node index as the horizontal axis and the continuous risk score as the vertical axis. This step "localizes" the abstract risk assessment results to actual path points, achieving spatial risk localization, which can be used and responded to in real time by subsequent risk modeling modules, frequency hopping control modules, or path reconstruction modules.

[0121] To enhance the dynamic responsiveness and adaptability of the overall model, local clustering and trend analysis are performed on the node-level risk score map to extract continuous high-risk segments. A parameter called a "perception loss probability factor" is introduced. This factor, calculated by combining the length, fluctuation amplitude, and maximum risk value of the local continuous high-risk segment, represents the comprehensive probability of losing stable perception along the continuous path. The risk scores of all path nodes are weighted and integrated with the perception loss probability factor to generate a unified output parameter, the "perception loss risk index." This index, output as a numerical scalar, reflects the overall failure risk level of the perception system in the current robot operation area in real time, demonstrating immediacy, trend-based, and measurable characteristics.

[0122] The "Perception Blindness Probability Factor" is a comprehensive probabilistic indicator used to quantify the risk of a charging robot's perception system failing within a continuous path segment. Its calculation process is based on high-risk continuous segments extracted from the path node-level risk scoring map and is weighted and integrated using the following core dimensions:

[0123] In the node-level risk scoring map, path node segments whose continuous risk scores exceed the set high-risk threshold are identified, and the length of their continuous high-risk segments, that is, the number of nodes that are continuously in a high-risk state, is extracted. This parameter reflects the time span that the perception system is in an unstable state for a long time and is the primary basis for judging potential failure trends.

[0124] The risk value fluctuation range within this continuous high-risk segment is analyzed. By calculating the difference between the maximum and minimum risk scores within this segment, as well as the risk change rate between adjacent nodes, we determine whether the perceived risk exhibits significant fluctuations. Large fluctuations indicate a high degree of uncertainty in the perception system due to interference, increasing the likelihood of system prediction errors.

[0125] The maximum risk score within this high-risk segment is used to measure the system's risk of blindness at the most extreme node. A higher score indicates that the perception system at a particular node in this path segment is close to complete failure.

[0126] Based on these three parameters, a fusion scoring model is constructed. The length of the high-risk segment is used as the primary factor, and the fluctuation amplitude and maximum risk value are used as adjustment factors. This comprehensive calculation results in a perception loss probability factor ranging from 0 to 1. The closer this factor is to 1, the higher the probability of the robot experiencing complete perception failure in the current path segment. This factor directly contributes to the final output of the perception loss risk index, serving as a critical bridge between local risk fluctuations and system-level judgments.

[0127] To ensure the model's long-term effectiveness and environmental adaptability, a dynamic learning mechanism was designed for the risk assessment module. This mechanism analyzes the discrepancies between the perception execution results and the risk prediction results after each task is completed, records the prediction error distribution, and adjusts the weight mapping table parameters in the risk model based on the actual misjudgment rate. For example, if the perception blindness risk index is low across multiple predictions but recognition error actually occurs, the impact weight of the interference frequency band intensity level will be recalibrated to increase its risk contribution. Through this feedback mechanism, the system continuously learns and adjusts the assessment model during operation, enhancing its ability to predict perception degradation trends in diverse industrial environments.

[0128] The perception blindness risk index is compared with a preset risk threshold. When the perception blindness risk index exceeds the preset risk threshold, frequency hopping control is executed to switch the operating frequency of the ultrasonic sensor to a frequency range that does not include the current interference frequency band. The echo signal processing strategy is synchronously adjusted based on the frequency parameter after hopping to achieve continuous and stable output of obstacle recognition results in an acoustic interference environment.

[0129] This step addresses the issue of ultrasonic sensor performance degradation or even failure in complex industrial environments caused by high-frequency acoustic interference, ensuring that charging robots can consistently and stably complete obstacle detection and path correction tasks along the charging path. In the previous step, the system calculated a perception blindness risk index for the current operating state using a multi-factor perception blindness risk assessment model. This index reflects the likelihood and severity of ultrasonic perception blindness in the robot's future path. This step compares this index with a preset risk threshold, acting as a trigger to determine whether the system needs to enter an emergency adjustment state. If the current risk index exceeds the threshold, indicating that the ultrasonic sensor at the current operating frequency is highly likely to be unable to properly receive echo signals due to interference, the system immediately implements a frequency hopping control strategy, switching the operating frequency to a frequency range that does not overlap with the current interference frequency band to proactively mitigate the effects of acoustic interference. Furthermore, to ensure robust recognition at the new frequency, the system also adjusts echo signal recognition processing parameters, such as echo gain, time threshold, and matched filter window, to achieve immediate adaptive recognition after the frequency adjustment. Through this step, the system has the ability to adaptively adjust in scenarios where acoustic interference is continuous or sudden. It can not only restore the perception ability, but also maximize the continuity and accuracy of the recognition results, thereby effectively improving the operating stability and safety of the charging robot in changing environments.

[0130] When the perception blindness risk index output by the charging robot's perception system exceeds the preset risk threshold, the ultrasonic sensor frequency hopping control is immediately executed, and the echo signal processing strategy is synchronously and adaptively adjusted to achieve stable perception output in complex acoustic interference environments. The specific steps are as follows:

[0131] Based on the risk index exceeding the limit, the frequency hopping response strength is calculated as the frequency offset amplitude of this frequency hopping. The parameters are defined as follows:

[0132] , where It is the frequency hopping response strength, which indicates the frequency amplitude that the ultrasonic sensor needs to jump from the current operating frequency. It is used to control the distance of the frequency hopping target. The larger the value, the stronger and wider the interference area needs to be avoided. is the frequency hopping sensitivity coefficient, which is used to control the response amplitude of the system to the risk index difference and adjust the "intensity" of the frequency hopping reaction. is the perception blindness risk index, which indicates the risk level of perception function breakdown calculated by the current system. It is the preset risk threshold and the dividing line for judging whether frequency hopping behavior is triggered. It is the acoustic interference time coverage ratio, with a value range of 0-1, indicating whether the robot will be exposed to the interference environment for a long time in the future. The closer it is to 1, the more persistent the interference is, and the stronger the frequency hopping response should be. For example: : represents that the path is in interference coverage only 20% of the time;

[0133] This step directly links the risk level, interference time and frequency hopping control amplitude, achieving a linkage between the response degree and the severity of the interference.

[0134] Based on the relative position of the current ultrasonic sensor operating frequency and the center frequency of the interference frequency band, the new operating frequency after the jump is calculated. The calculation formula is as follows:

[0135] , where It is the new operating frequency after the jump, which is used to replace the frequency of the current interfered frequency band to avoid the interference source and ensure the perception function. Is the current operating frequency of the ultrasonic sensor, as the starting point of frequency hopping, and is the reference benchmark for frequency hopping. It is the center frequency of the interference frequency band, which is used to determine whether the current operating frequency is close to the interference frequency band and determine the frequency hopping direction. is a sign function;

[0136] In this step, the core role of the sign function is to control the directionality of the frequency jump. Interference center frequency The system automatically decides whether the frequency hopping is "upward" (frequency increase) or "downward" (frequency decrease). For example, when the current operating frequency is lower than the interference frequency, that is, ,but , the system will hop down to a lower frequency band that is further away from the interference area; on the contrary, when , the system then hops upward. This eliminates the need for manual direction specification during the frequency hopping process. Instead, the system automatically determines the hopping direction based on the actual location of the interference, ensuring that the new frequency is as far away from the interference band as possible, improving the clarity of the echo signal and the reliability of obstacle identification. Therefore, the sign function plays a key role in direction determination and decision-making control in this step, and is one of the logical cores of the frequency hopping strategy's "adaptive interference avoidance."

[0137] By constraints:

[0138] ,in, It is the lowest available frequency supported by the ultrasonic sensor and is used to prevent the frequency from hopping beyond the frequency range allowed by the sensor hardware. It is the highest available frequency supported by the ultrasonic sensor and is the hardware upper limit protection for frequency hopping. It is a set variable, which represents the set of frequency intervals that have been identified as having high-intensity interference. It ensures that the newly selected frequency does not fall into any interfered frequency band interval, thus ensuring the perceived signal-to-noise ratio and stability. It is a logical condition to determine whether the new frequency falls into the interference frequency band. If it falls into the interference range, it needs to be readjusted. or implement a frequency scheduling fallback mechanism;

[0139] Ensure that the new frequency is within the sensor's available range and does not overlap with any of the identified interference band sets middle.

[0140] After successfully completing the frequency hopping switch, based on the new operating frequency after the hopping Dynamically adjust the echo signal processing strategy, including: echo sampling window length, echo gain adjustment coefficient and reconstruction of matched filter bandwidth parameters. The calculation expression is as follows:

[0141] , where It is the echo sampling window length, which indicates the sampling duration set by the receiving end in echo identification, and is used to intercept the effective echo signal area. is the sampling ratio constant (unit: Hz·s), which is adjusted according to the sensor detection distance, reflection delay and waveform period. The general value range is 20-50. It is used to adjust the strength of the inverse relationship between the window scale and frequency. It is the echo gain adjustment coefficient, which is used to control the amplifier in the receiving circuit to adjust the signal voltage or power multiples to compensate for the decrease in signal strength. It is the standard gain value at the original operating frequency, indicating the default signal amplification factor set when the system is initialized. It is usually determined based on the reflection characteristics at 40kHz. It is the gain adjustment sensitivity factor, which controls the response amplitude to frequency changes. It can be calibrated through test scenarios. The value range is generally between 0.3 and 1.0. It is the bandwidth adjustment value of the matched filter, which indicates the frequency domain range that the filter needs to cover after frequency hopping, and is used to adapt the signal spectrum characteristics at the new frequency. It is the bandwidth response expansion coefficient, which controls the degree of bandwidth expansion and has a value between 1.5 and 3.0.

[0142] By dynamically calculating and loading the above three parameters, the signal processing link in the receiving channel can be rebuilt according to the frequency hopping results, so that the ultrasonic echo signal maintains recognition accuracy and stability at the new frequency, effectively avoiding recognition drift, echo loss or obstacle misjudgment caused by interference.

[0143] This step achieves the adaptive reconstruction and high-stability operation of the ultrasonic sensing system to acoustic interference conditions by establishing a complete control chain of risk triggering → frequency switching → strategy synchronization adjustment, providing key guarantees for the autonomous charging of the charging robot in complex scenarios.

[0144] Through this approach, charging robots facing industrial environments with severe high-frequency acoustic interference can implement active interference recognition, adaptive risk assessment, and dynamic frequency response adjustment for their ultrasonic sensing systems, significantly improving their perception stability and operational safety in complex operating scenarios. Compared to traditional passive sensing systems, this approach not only proactively identifies interfering frequency bands that could cause ultrasonic echo failure, but also establishes a real-time, quantitative perception confidence level adjustment mechanism through multimodal recognition consistency analysis and historical deviation tracking. Ultimately, it outputs a perception blindness risk index that can be used for control decisions. Once the risk index exceeds a threshold, the system immediately implements frequency hopping and adaptive echo strategy optimization to effectively avoid frequency band masking and maintain the continuity and accuracy of obstacle recognition. This approach, which forms a closed-loop "interference perception-risk assessment-control response" system, significantly improves the robot's charging positioning success rate, path execution accuracy, and system safety and fault tolerance in acoustic interference environments, promising promising industrial application prospects and engineering promotion value.

[0145] The present invention provides Figure 2 The automatic charging processing device for a charging robot shown in the figure includes an acoustic interference perception module, an acoustic interference prediction module, a multimodal recognition consistency assessment module, a perception confidence level adjustment module, a perception blindness risk assessment module, and a frequency hopping control and recognition strategy adaptation module:

[0146] The acoustic interference perception module builds an acoustic array to collect high-frequency background sound wave signals in the charging robot's operating area in real time, extract the sound spectrum energy distribution, identify the interference frequency band that coincides with the ultrasonic sensor's operating frequency, and calculate the interference frequency band intensity level based on the sound spectrum energy intensity and duration within the interference frequency band;

[0147] The acoustic interference prediction module establishes an acoustic interference prediction model based on the identified interference frequency band and interference frequency band intensity level, combined with the charging robot's current position and path node parameters, and generates an acoustic interference spatial projection map;

[0148] The multimodal recognition consistency evaluation module obtains the obstacle recognition results of lidar, visual recognition, and ultrasonic sensors, calculates the recognition consistency deviation index based on the spatial deviation between the recognition results, and records the number of recognition deviation events to obtain the historical number of recognition deviations;

[0149] The perception confidence level adjustment module integrates the recognition consistency deviation index with the acoustic interference space projection map for analysis, constructs the perception confidence level adjustment factor, and dynamically adjusts the credibility weight of each recognition result;

[0150] The perceptual blindness risk assessment module builds a multi-factor perceptual blindness risk assessment model based on the perceptual confidence level adjustment factor, interference frequency band intensity level, and historical recognition deviation times, and outputs a perceptual blindness risk index.

[0151] The frequency hopping control and recognition strategy adaptive module compares the perception blindness risk index with the preset risk threshold. When the perception blindness risk index is greater than the preset risk threshold, it performs frequency hopping control, switches the ultrasonic sensor operating frequency to a frequency range that does not include the current interference frequency band, and synchronously adjusts the echo signal processing strategy based on the frequency parameters after hopping to achieve continuous and stable output of obstacle recognition results.

[0152] An embodiment of the present invention provides an automatic charging processing method for a charging robot, which is implemented by the above-mentioned automatic charging processing device for a charging robot. The specific method and process of the automatic charging processing device for a charging robot are detailed in the embodiment of the above-mentioned automatic charging processing method for a charging robot, and will not be repeated here.

[0153] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0154] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0155] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0156] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0157] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0158] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0159] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0160] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0161] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0162] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A charging robot automatic charging processing method, characterized in that: The following steps are involved: S1: Build an acoustic array to collect high-frequency background sound wave signals in the charging robot's operating area in real time, extract the sound spectrum energy distribution, identify the interference frequency band that coincides with the ultrasonic sensor's operating frequency, and calculate the interference frequency band intensity level based on the sound spectrum energy intensity and duration within the interference frequency band; S2, based on the identified interference frequency band and interference frequency band intensity level, combined with the current position of the charging robot and path node parameters, establishes an acoustic interference prediction model and generates an acoustic interference spatial projection map; S3, obtains the obstacle recognition results of the lidar, visual recognition, and ultrasonic sensors, calculates the recognition consistency deviation index based on the spatial deviation between the recognition results, and records the number of recognition deviation events to obtain the historical number of recognition deviations; S4, integrates the recognition consistency deviation index with the acoustic interference space projection map for analysis, constructs the perception confidence level adjustment factor, and dynamically adjusts the credibility weight of each recognition result; S5, based on the perception confidence level adjustment factor, the interference frequency band intensity level and the number of historical recognition deviations, builds a multi-factor perception blindness risk assessment model and outputs the perception blindness risk index; S6 compares the perceived blindness risk index with the preset risk threshold. When the perceived blindness risk index is greater than the preset risk threshold, frequency hopping control is executed to switch the ultrasonic sensor operating frequency to a frequency range that does not include the current interference frequency band. The echo signal processing strategy is synchronously adjusted based on the frequency parameters after the hopping to achieve continuous and stable output of obstacle recognition results.

2. The automatic charging processing method of a charging robot according to claim 1, characterized in that: Step S1 includes: Construct an acoustic array consisting of multiple microphone units and collect high-frequency background sound wave signals in the charging robot's current operating area in real time; The high-frequency background sound wave signal is subjected to high-pass filtering and time-domain windowing, and a fast Fourier transform is performed within a fixed time sampling window to obtain the spectrum energy distribution; The spectrum is divided into multiple sub-bands, and the energy peak, mean, and variance indicators are extracted to determine whether there is an interference frequency band that overlaps with the operating frequency of the ultrasonic sensor; The interference frequency band intensity level is calculated based on average energy density, peak duration, spectrum fluctuation stability and multi-channel coupling consistency.

3. The automatic charging processing method of a charging robot according to claim 1, characterized in that: Step S2 includes: Based on the interference frequency band and the interference frequency band intensity level, obtain the equivalent radiation range of the interference source and the duration of the interference source impact; Combine the charging robot's current position, movement speed, orientation angle, and path node parameters to construct a path trajectory prediction segment set; Map the equivalent radiation range of the interference source to the set of path trajectory prediction segments, and calculate the duration of the interference source impact in each trajectory segment; Based on the path trajectory, the segment set and the duration of interference source impact are predicted, and an acoustic interference spatial projection map with interference intensity labels and high failure-sensitive node identifiers is generated.

4. The automatic charging processing method of a charging robot according to claim 1, characterized in that: Step S3 includes: Obtain the obstacle recognition results output by the lidar module, visual recognition module, and ultrasonic sensor module, and uniformly convert each recognition result into the reference coordinate system of the charging robot body; Match each recognition result according to the target fusion window parameters, and calculate the deviation value of each recognized target in spatial position and boundary contour to obtain the recognition consistency deviation index; Determine whether the recognition consistency deviation index exceeds the preset confidence interval, and record it as a recognition deviation event if it exceeds the confidence interval; The cumulative number of identification deviation events is counted within the set operation cycle to obtain the historical number of identification deviations.

5. The automatic charging processing method of a charging robot according to claim 1, characterized in that: Step S4 includes: Based on the acoustic interference spatial projection map and the recognition consistency deviation index, a local grid is constructed within the charging robot's path prediction area, and the perception anomaly weight of each grid is calculated. In each grid, the confidence adjustment rule table is called to adjust the confidence of each type of sensor based on the perception anomaly weight, acoustic interference level, recognition deviation status and historical recognition deviation count; Integrate the sensor confidence adjustment results in each grid into the path node to generate a perception confidence level adjustment factor that includes the three types of sensor confidence weights; The perception confidence level adjustment factor is updated in each navigation control cycle to guide the weighted fusion output of multimodal obstacle recognition results.

6. The automatic charging processing method of a charging robot according to claim 1, characterized in that: Step S5 includes: Obtain the perception confidence level adjustment factor, interference frequency band intensity level, and historical identification deviation count of the path node, and construct a perception vulnerability label; For path nodes marked as high concern, a risk score is calculated, including the degree of confidence imbalance, the proportion of interference frequency band coverage time, and the trend of identification deviation changes; Cluster analysis was performed on the risk scores of the path nodes to extract continuous high-risk segments and calculate the probability factor of perceptual blindness; The risk scores of all path nodes are weighted and integrated with the perceptual blindness probability factor to output the perceptual blindness risk index.

7. The automatic charging processing method of a charging robot according to claim 1, characterized in that: When the perception blindness risk index output by the charging robot's perception system exceeds the preset risk threshold, the ultrasonic sensor frequency hopping control is immediately executed, and the echo signal processing strategy is synchronously and adaptively adjusted to achieve stable perception output in complex acoustic interference environments. The specific steps are as follows: Based on the risk index exceeding the limit, the frequency hopping response strength is calculated as the frequency offset amplitude of this frequency hopping; Based on the relative position of the current ultrasonic sensor operating frequency and the center frequency of the interference frequency band, the new operating frequency after the jump is calculated; Through the constraint conditions, it is determined whether the new frequency falls into the interference frequency band. If it falls into the interference range, it needs to be readjusted. Implement frequency scheduling fallback mechanism; After successfully completing the frequency hopping switch, based on the new operating frequency after the hopping Dynamically adjust the echo signal processing strategy, including: echo sampling window length, echo gain adjustment coefficient and reconstruction of matched filter bandwidth parameters.

8. An automatic charging processing device for a charging robot, used to implement the automatic charging processing method for a charging robot according to any one of claims 1 to 7, characterized in that: It includes an acoustic interference perception module, an acoustic interference prediction module, a multimodal recognition consistency assessment module, a perception confidence level adjustment module, a perception blindness risk assessment module, and a frequency hopping control and recognition strategy adaptation module: The acoustic interference perception module builds an acoustic array to collect high-frequency background sound wave signals in the charging robot's operating area in real time, extract the sound spectrum energy distribution, identify the interference frequency band that coincides with the ultrasonic sensor's operating frequency, and calculate the interference frequency band intensity level based on the sound spectrum energy intensity and duration within the interference frequency band; The acoustic interference prediction module establishes an acoustic interference prediction model based on the identified interference frequency band and interference frequency band intensity level, combined with the charging robot's current position and path node parameters, and generates an acoustic interference spatial projection map; The multimodal recognition consistency evaluation module obtains the obstacle recognition results of lidar, visual recognition, and ultrasonic sensors, calculates the recognition consistency deviation index based on the spatial deviation between the recognition results, and records the number of recognition deviation events to obtain the historical number of recognition deviations; The perception confidence level adjustment module integrates the recognition consistency deviation index with the acoustic interference space projection map for analysis, constructs the perception confidence level adjustment factor, and dynamically adjusts the credibility weight of each recognition result; The perceptual blindness risk assessment module builds a multi-factor perceptual blindness risk assessment model based on the perceptual confidence level adjustment factor, interference frequency band intensity level, and historical recognition deviation times, and outputs a perceptual blindness risk index. The frequency hopping control and recognition strategy adaptive module compares the perception blindness risk index with the preset risk threshold. When the perception blindness risk index is greater than the preset risk threshold, it performs frequency hopping control, switches the ultrasonic sensor operating frequency to a frequency range that does not include the current interference frequency band, and synchronously adjusts the echo signal processing strategy based on the frequency parameters after hopping to achieve continuous and stable output of obstacle recognition results.

Images