Multi-sensor fusion navigation method and system of robot

Abstract

The invention relates to the technical field of robot navigation, and discloses a multi-sensor fusion navigation method and system of a robot. The method comprises the following steps: collecting point cloud data of a laser radar sensor and motion data of an IMU six-axis sensor; executing generalized likelihood ratio test and sliding window interval search on the point cloud data and the motion data to obtain a zero-speed abnormal interval, and creating a detection threshold value for the current ground environment according to the zero-speed abnormal interval; performing spatial superposition on the position coordinates of the zero-speed abnormal interval and a laser radar scanning blind area to obtain an obstacle distribution map; cleaning control parameters are adjusted based on the detection threshold value, a re-planning cleaning path is generated based on the obstacle distribution map and the cleaning control parameters, the method can actively identify the potential trapped area and re-plan the cleaning path, and the autonomous navigation ability and cleaning efficiency of the intelligent cleaning robot are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot navigation, and in particular to a multi-sensor fusion navigation method and system for a robot. BACKGROUND

[0002] Traditional sweeping robots mainly rely on laser radar sensors for environment perception and navigation positioning. However, the laser radar sensor is usually installed at a height of 15-20 cm from the ground, forming a horizontal scanning plane. This results in a serious detection blind area for low obstacles (such as threshold strips, carpet edges, and power lines) below the scanning plane, causing the robot to frequently encounter the problem of being stuck. At the same time, the existing technology generally uses a single sensor or a simple combination of sensors to determine the motion state, lacks deep mining of the periodic characteristics of the cleaning motion, and cannot accurately distinguish between normal cleaning mode and abnormal stuck state, affecting the cleaning efficiency and user experience. SUMMARY

[0003] The present application provides a multi-sensor fusion navigation method and system for a robot, which can actively identify potential stuck areas and re-plan the cleaning path, improving the autonomous navigation ability and cleaning efficiency of the intelligent cleaning robot.

[0004] In a first aspect, the present application provides a multi-sensor fusion navigation method for a robot, which comprises: collecting point cloud data of a laser radar sensor and motion data of an IMU six-axis sensor; performing generalized likelihood ratio test and sliding window interval search on the point cloud data and the motion data to obtain a zero-speed abnormal interval, and creating a detection threshold for the current ground environment according to the zero-speed abnormal interval; spatially superimposing the position coordinates of the zero-speed abnormal interval and the laser radar scanning blind area to obtain an obstacle distribution map; adjusting cleaning control parameters based on the detection threshold, and generating a re-planned cleaning path based on the obstacle distribution map and the cleaning control parameters.

[0005] In combination with the first aspect, in a first implementation manner of the first aspect of the present application, the collecting of the point cloud data of the laser radar sensor and the motion data of the IMU six-axis sensor comprises: setting the scanning frequency of the laser radar sensor in the intelligent cleaning robot and starting the environment scanning mode to obtain first collection configuration parameters; setting the sampling frequency of the IMU six-axis sensor in the intelligent cleaning robot and starting the accelerometer and gyroscope synchronous collection mode to obtain second collection configuration parameters; acquire distance measurement values and reflection intensity values of an environment around the intelligent cleaning robot based on the first acquisition configuration parameter to obtain point cloud data; synchronously acquire real-time measurement values of three-axis linear acceleration and three-axis angular velocity in the intelligent cleaning robot based on the second acquisition configuration parameter to obtain motion data.

[0006] In a second implementation manner of the first aspect, the generalized likelihood ratio test and the sliding window interval search are performed on the point cloud data and the motion data to obtain a zero-speed abnormal interval, and a detection threshold for a current ground environment is created according to the zero-speed abnormal interval, including: a change amount of a Euclidean distance of adjacent frame point cloud data is calculated as a point cloud change rate, and a resultant modulus value of three-axis angular velocity in the motion data is calculated; the point cloud change rate and the resultant modulus value are multiplied by corresponding weight coefficients respectively and then summed to obtain a motion state indicator of the intelligent cleaning robot; based on the motion state indicator, a normalized cumulative calculation is performed on an acceleration vector modulus length and a motor power mutation amount in the motion data to obtain a generalized likelihood ratio statistic; a preset time window is used to continuously scan the generalized likelihood ratio statistic and mark a continuous time period that exceeds a preset threshold to obtain a zero-speed abnormal interval containing a start time and an end time; a detection threshold for a current ground environment is created according to the zero-speed abnormal interval.

[0007] In a third implementation manner of the first aspect, the preset time window is used to continuously scan the generalized likelihood ratio statistic and mark a continuous time period that exceeds a preset threshold to obtain a zero-speed abnormal interval containing a start time and an end time, including: a time window is set and a scanning start position is located at a first data point of the generalized likelihood ratio statistic; the generalized likelihood ratio statistic in the time window is compared with a preset threshold to obtain a threshold exceeding state mark of each time point in the time window; time points that continuously exceed the preset threshold are sequentially marked and recorded in position coordinates on a time axis to obtain continuous sequence data; a first time point that exceeds the threshold in the continuous sequence data is detected as the start time and a last time point that exceeds the threshold is detected as the end time to obtain a zero-speed abnormal interval containing the start time and the end time.

[0008] In a fourth implementation manner of the first aspect, the detection threshold for the current ground environment is created according to the zero-speed abnormal interval, including: Calculate the mean and standard deviation of the generalized likelihood ratio statistic within the zero-velocity anomaly interval, and simultaneously calculate the duration and proportion of zero velocity within the zero-velocity anomaly interval to obtain the statistical characteristic parameters of the zero-velocity anomaly interval. The statistical feature parameters are matched with a preset stuck-out mode feature library to obtain the type identifier and corresponding learning rate coefficient of the current stuck-out event; The actual jamming rate is calculated based on the number of jamming events in the continuous cleaning task, and the actual jamming rate is corrected by comparing it with the preset target jamming rate to obtain the jamming rate deviation correction amount; Based on the type identifier of the current stuck event, the statistical feature parameters, the learning rate coefficient, and the stuck rate deviation correction amount are updated with thresholds to obtain the detection threshold for the current ground environment.

[0009] In conjunction with the first aspect, in the fifth implementation of the first aspect of the present invention, the step of spatially superimposing the position coordinates of the zero-velocity anomaly interval with the blind zone of the lidar scan to obtain an obstacle distribution map includes: Extract the coordinate data of the zero-speed abnormal position of the intelligent cleaning robot in the zero-speed abnormal range, and read the point cloud density information of the lidar sensor at the zero-speed abnormal position. Based on the number of obstacle points in the point cloud density information and the preset blind zone determination threshold, it is confirmed that there is a lidar scanning blind zone at the zero-velocity abnormal location. Based on the blind zone of the lidar scan, the height of low obstacles is estimated by using the IMU acceleration change amplitude within the zero-velocity anomaly range, and the height of low obstacles is data-bound with the coordinate data to obtain the spatial information of low obstacles; Based on the spatial information of the low-lying obstacles, a height attribute is added to the lidar grid map data structure, and the occupancy status and confidence of the corresponding grid are updated to obtain an obstacle distribution map.

[0010] In conjunction with the first aspect, in a sixth implementation of the first aspect of the present invention, the step of adjusting the cleaning control parameters based on the detection threshold and generating a replanning cleaning path based on the obstacle distribution map and the cleaning control parameters includes: The zero-speed detection intensity is calculated based on the generalized likelihood ratio statistic and the detection threshold. The zero-speed detection intensity is then multiplied by the power adjustment coefficient and added to the basic suction power to obtain the suction power. Calculate the first ratio of the generalized likelihood ratio statistic to the detection threshold, and trigger the corresponding movement speed adjustment parameter when the first ratio is in different intervals. monitoring real-time current when the dust suction motor is running, calculating a second ratio of the real-time current and a reference current, and reversely adjusting the dust suction power and a moving speed adjustment parameter to obtain a cleaning control parameter when the second ratio deviates from a normal range; generating a re-planned cleaning path based on the obstacle distribution map and the cleaning control parameter.

[0011] In a seventh implementation form of the first aspect, the generating a re-planned cleaning path based on the obstacle distribution map and the cleaning control parameter comprises: extracting height attributes, confidence and coordinate data of the zero-speed abnormal interval of each grid in the obstacle distribution map, and calculating spatial risk distribution data of each grid position; multiplying the spatial risk distribution data and a moving speed adjustment parameter in the cleaning control parameter, and linearly combining the result with a path distance cost according to a preset proportion coefficient to obtain a path total cost function; performing shortest path search from a current position of the intelligent cleaning robot to a target cleaning area by using the path total cost function to obtain a path set; comparing total cost values of each candidate path in the path set, selecting a path with the minimum cost as a main cleaning path, and marking a high-risk area as a delayed cleaning area to obtain a re-planned cleaning path.

[0012] In an eighth implementation form of the first aspect, the performing shortest path search from a current position of the intelligent cleaning robot to a target cleaning area by using the path total cost function to obtain a path set comprises: taking the current position of the intelligent cleaning robot as a path search starting point, reading coordinate ranges of the target cleaning area and determining a search ending position; dividing the obstacle distribution map into grid nodes according to a fixed interval, calculating a passing cost value of each grid node by substituting the grid node into the path total cost function, and generating a grid search space according to the passing cost value; starting to expand adjacent grid nodes from the path search starting point, automatically generating a detour branch path when detecting that a passing cost value of a grid node exceeds a safety threshold, and continuously searching until the search ending position is reached to obtain a path set.

[0013] In a second aspect, the present application provides a multi-sensor fusion navigation system of a robot, which comprises: a collection module configured to collect point cloud data of a laser radar sensor and motion data of an IMU six-axis sensor; a zero speed anomaly detection module configured to perform a generalized likelihood ratio test and a sliding window interval search on the point cloud data and the motion data to obtain a zero speed anomaly interval, and create a detection threshold for a current ground environment according to the zero speed anomaly interval; a spatial superposition module configured to perform spatial superposition on position coordinates of the zero speed anomaly interval and a laser radar scanning blind area to obtain an obstacle distribution map; a generation module configured to adjust a cleaning control parameter based on the detection threshold, and generate a re-planned cleaning path based on the obstacle distribution map and the cleaning control parameter.

[0014] In the technical solution provided in the present application, through multi-sensor asynchronous data deep fusion, laser radar point cloud data and IMU high-frequency motion data are processed in time and space synchronization to construct an accurate motion state evaluation system. The interval search adaptive zero speed detection mechanism extends the traditional "complete static" detection to "abnormal low speed motion" identification, which can accurately distinguish different types of entanglement such as wire entanglement and carpet edge entrapment. The adaptive threshold dynamic adjustment technology optimizes the detection parameters in real time according to the characteristics of the ground environment, effectively solving the adaptability problem of fixed threshold. The laser radar blind area intelligent compensation technology infers low obstacle information from the zero speed detection result, and the intelligent cleaning control driven by zero speed establishes a dynamic mapping relationship between detection intensity and cleaning parameters, realizing the coordinated optimization of cleaning effect and motion safety. The risk prediction model based on Bayesian network can actively identify potential entrapment areas and re-plan the cleaning path, improving the autonomous navigation ability and cleaning efficiency of the intelligent cleaning robot. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0016] Figure 1 A flowchart of a multi-sensor fusion navigation method of a robot provided in the embodiments of the present application; Figure 2 A structural schematic block diagram of a multi-sensor fusion navigation system of a robot provided in the embodiments of the present application. DETAILED DESCRIPTION

[0017] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only some of the embodiments of the present application, but not all of them. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts are within the scope of the present application.

[0018] The flowcharts shown in the drawings are only illustrative, and do not necessarily include all contents and operations / steps, nor are they necessarily executed in the described order. For example, some operations / steps can be further decomposed, combined or partially merged, so that the actual execution order can be changed based on actual situations.

[0019] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and the appended claims of the present application, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0020] It should be further understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0021] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments and features in the embodiments can be combined with each other without conflict.

[0022] Please refer to Figure 1 , Figure 1 The flowchart of the multi-sensor fusion navigation method of the robot provided in the embodiments of the present application is shown in FIG. 1. The multi-sensor fusion navigation method of the robot provided in the embodiments of the present application includes the following steps. Figure 1 Step S100, collecting point cloud data of a laser radar sensor and motion data of an IMU six-axis sensor; ​Specifically, in the intelligent cleaning robot, the working mode of the laser radar sensor is initialized and configured, the scanning frequency is set to a stable working frequency band between 5 and 10 Hz, and after the parameter loading is completed, the environment scanning mode is started to form the first collection configuration parameter, which is used to guide the radar to continuously collect the distance measurement value and the reflection intensity value of the surrounding obstacles in the set sector, and to construct the point cloud data with spatial density and reflection characteristics. At the same time, the control system synchronously samples and configures the IMU six-axis inertial measurement unit, sets the unified sampling frequency of the accelerometer and the gyroscope to 100 Hz, and enables the synchronous collection mechanism to ensure that the two types of sensors start the measurement process at the same time reference, forming the second collection configuration parameter, and the system real-time obtains the measurement data of three-axis linear acceleration and three-axis angular velocity, and constitutes the motion data set describing the motion state of the robot itself. In the data collection link, the hardware timing mechanism is used to align the time stamps of the laser radar and the IMU, and the point cloud data and the motion data are encapsulated into a unified time sequence data structure through the cache scheduling logic.

[0023] Step S200, performing generalized likelihood ratio test and sliding window interval search on the point cloud data and the motion data to obtain a zero speed abnormal interval, and creating a detection threshold for the current ground environment according to the zero speed abnormal interval; Specifically, the point cloud frame data continuously output by the laser radar sensor is compared frame by frame, the Euclidean distance change of the point cloud coordinates between adjacent frames is calculated to represent the dynamic change of the spatial geometric structure, and the point cloud change rate reflecting the environmental change sensitivity is obtained. At the same time, the square sum of the three-axis angular velocity output by the IMU six-axis sensor is calculated, and the resultant modulus value of the angular velocity is obtained to reflect the attitude characteristics of the robot in the turning, deflection or vibration state. The point cloud change rate and the resultant modulus value of the angular velocity are multiplied by the preset weight coefficient respectively, and the comprehensive motion state index is obtained by weighted summation. Based on the motion state index, the modulus of the three-axis acceleration vector in the motion data and the instantaneous mutation of the motor power are introduced, which are normalized and added to the motion state index for calculation, to construct a generalized likelihood ratio statistic. The change trajectory of the generalized likelihood ratio statistic can reveal whether the robot is in an abnormal zero speed state during operation. The sliding window mechanism is used to scan the generalized likelihood ratio statistic in the time domain, a fixed length time window is set and slid point by point, and when the statistic exceeds the preset threshold in the continuous time period, the starting time and the ending time of the time period are automatically marked and defined as the zero speed abnormal interval. According to the distribution, duration and intensity characteristics of the zero speed abnormal interval, a detection threshold suitable for the current ground material and environmental characteristics is created, so that the robot can identify the risk of being stuck in different environments in an adaptive manner.

[0024] Step S300, superimposing the position coordinates of the zero speed abnormal interval and the laser radar scanning blind area in space to obtain an obstacle distribution map; Specifically, the zero-speed abnormal interval is played back on the time axis, and the pose coordinates and timestamps of the intelligent cleaning robot appearing zero speed are extracted point by point from the zero-speed abnormal interval, and the point cloud density information around the corresponding position of the laser radar original frame or the cache frame at the same timestamp is read. A local sector detection area is established with the zero-speed position as the center and a radius of about 0.3 meters, the number of points marked as obstacles in the area is counted and compared with the preset blind area determination threshold, when the number of obstacle points is lower than the threshold (for example, less than five) and the zero-speed abnormal duration reaches the set requirement, it is determined that there is a laser radar scanning blind area at the zero-speed position. After completing the blind area confirmation, the vertical acceleration change amplitude of the IMU in the zero-speed abnormal interval and the rigid characteristics of the robot chassis are used to convert the equivalent height change caused by unit acceleration into the estimated height of the low obstacle according to the height conversion coefficient, the upper limit of the height is limited within the range of about 15 centimeters, and the estimated height is bound with the coordinates of the zero-speed position to generate the spatial information of the low obstacle containing position and height information. The map updating module writes the spatial information of the low obstacle into the grid map of the laser radar, increases the height attribute on the basis of the original two-dimensional occupancy state, and configures the confidence according to the duration and intensity of the zero-speed abnormality: the longer the duration and the greater the intensity, the higher the confidence. If the same position repeatedly triggers the zero-speed abnormality in multiple cleaning tasks, the risk level of the position is gradually increased through time decay accumulation, and the passage cost is increased in path planning. The generated obstacle distribution map contains plane occupancy information and has three attributes of height and confidence, forming a "2.5-dimensional" environment representation.

[0025] Step S400, adjust the cleaning control parameter based on the detection threshold, and generate a re-planned cleaning path based on the obstacle distribution map and the cleaning control parameter.

[0026] Specifically, the zero-speed detection intensity is calculated by comparing the generalized likelihood ratio statistic with the detection threshold, and the zero-speed detection intensity is input into the cleaning control module as a representation parameter of the environmental resistance. The zero-speed detection intensity is multiplied by the power adjustment coefficient, and on this basis, a preset basic suction value is superimposed to dynamically generate the current suction power. The suction power can be adaptively adjusted according to the complexity of the environment and the resistance of the robot. At the same time, the first ratio between the generalized likelihood ratio statistic and the detection threshold is calculated, and the corresponding moving speed adjustment parameter is triggered according to the interval in which the first ratio falls. For example, when the ratio is in a lower interval, the normal speed is maintained, and when the ratio increases and approaches the critical state of being stuck, the speed is gradually reduced to avoid the robot being further trapped in obstacles due to excessive inertia in complex terrain. During the operation of the suction motor, the real-time current is continuously monitored and compared with the preset reference current to obtain a second ratio. If the second ratio deviates significantly, such as an abnormally high current representing an increased resistance, the suction power is adjusted in the opposite direction and the moving speed parameter is simultaneously corrected to prevent excessive energy consumption or conflicts between suction force and speed affecting the escape and cleaning effect. The cleaning control parameters obtained through the above process include real-time adjustment results of suction force and speed, and also reflect the adaptability of the robot to environmental resistance. The navigation system combines the updated cleaning control parameters with the obstacle distribution map to recalculate the optimal cleaning route through a path planning algorithm, which reasonably matches the suction force and speed while considering the height, distribution density and risk level of obstacles, so that the robot can maintain high cleaning efficiency while reducing the probability of being stuck, and achieve a balance between energy saving and efficiency. The output of the re-planned cleaning path.

[0027] In a specific embodiment, the process of step S100 can specifically include the following steps: Set the scanning frequency of the laser radar sensor in the intelligent cleaning robot and start the environment scanning mode to obtain the first acquisition configuration parameter; Set the sampling frequency of the IMU six-axis sensor in the intelligent cleaning robot and start the synchronous acquisition mode of the accelerometer and gyroscope to obtain the second acquisition configuration parameter; Obtain the distance measurement value and the reflection intensity value of the environment around the intelligent cleaning robot based on the first acquisition configuration parameter to obtain point cloud data; Synchronously obtain the real-time measurement value of the three-axis linear acceleration and three-axis angular velocity in the intelligent cleaning robot based on the second acquisition configuration parameter to obtain motion data.

[0028] Specifically, in the initialization phase of the intelligent cleaning robot system, the functions of various sensors are activated and the parameters are configured. The laser radar sensor is the main means to obtain the spatial structure information of the environment. The scanning frequency is set between 5 Hz and 10 Hz according to the robot moving speed, the complexity of the environment and the operation ability of the processing system. On this basis, the environment scanning mode is enabled. Through instruction control, the radar system enters the continuous laser beam emission and reflected signal acquisition state, and the reflected intensity output flag bit is set, so that the distance information of the target point is returned in each scanning period, and the reflection energy index is attached. The point cloud data of the environment with position and material characteristics is obtained. The above parameter set constitutes the first acquisition configuration parameter. At the same time, the start and synchronization configuration of the IMU six-axis sensor are completed. The sampling frequencies of the three-axis accelerometer and the three-axis gyroscope are uniformly set to 100 Hz, so that the two types of measurement subsystems perform data acquisition tasks in the same sampling period. The internal sampling clock synchronization mechanism and data channel binding strategy are configured to ensure that the data collected at each moment contains complete information of the acceleration and angular velocity two physical quantities. The functions of constant temperature compensation, noise suppression and three-axis coupling calibration are enabled to eliminate the dynamic drift error between different sensors. The above sampling parameters and processing logic jointly constitute the second acquisition configuration parameter. During task execution, the laser radar performs periodic 360-degree or specified angle sector spatial scanning according to the first acquisition configuration parameter. Each frame of scanning data contains multiple points. The controller calculates it into a polar coordinate or rectangular coordinate point set, and maps the point cloud data to the robot body coordinate system according to the built-in external parameter model, and adds a frame-level timestamp to mark the data acquisition time. The continuous frame data is used to construct a dynamically changing spatial model to form complete point cloud data. In the same time domain, the IMU sensor outputs real-time measurement values of three-axis linear acceleration and three-axis angular velocity according to the second acquisition configuration parameter. The on-chip cache mechanism is used to package the data of the six channels into a six-dimensional vector sequence in time sequence, and the six-dimensional vector sequence is sent to the data processing module through the one-time write bus. In order to ensure the time consistency of the data in the point cloud fusion process, the control system adds a unified high-precision timestamp in the data packaging stage, and uses a ring buffer or a double-channel DMA channel to manage the point cloud frame and the inertial measurement batch in parallel, so that they are decoupled in the multi-thread processing architecture. Thus, a synchronous acquisition system of sensors is built.

[0029] In a specific embodiment, the process of performing step S200 can specifically include the following steps: Calculate the Euclidean distance change of adjacent frame point cloud data as the point cloud change rate, and calculate the synthetic modulus value of the three-axis angular velocity in the motion data; Multiply the point cloud change rate and the synthetic modulus value by the corresponding weight coefficients respectively, and perform summation operation to obtain the motion state index of the intelligent cleaning robot; Based on the motion state index, the acceleration vector module length in the motion data and the motor power mutation variable are normalized and accumulated to obtain a generalized likelihood ratio statistic; A preset time window is used to continuously scan the generalized likelihood ratio statistic and mark the continuous time period exceeding the preset threshold to obtain a zero speed anomaly interval containing a start time and an end time; A detection threshold for the current ground environment is created according to the zero speed anomaly interval.

[0030] Specifically, the point cloud frame data continuously output by the laser radar is subjected to differential analysis, the point sets in the same space region of the current frame and the previous frame are subjected to Euclidean distance calculation, and the distance difference values of all valid corresponding point pairs are subjected to squaring, summing and square root to obtain a point cloud change quantity reflecting the degree of change of the environment structure. The point cloud change quantity is divided by the inter-frame time interval to obtain a point cloud change rate as a spatial change index reflecting the motion speed of the robot in the environment coordinate system. At the same time, the three-axis angular velocity components are extracted from the data collected by the IMU six-axis sensor, and the synthetic module value is calculated by the method of squaring and square root to measure the overall angular velocity amplitude of the robot rotating around its center of gravity. According to a preset fusion model, the point cloud change rate and the angular velocity module value are multiplied by corresponding weight coefficients, the weight coefficients are calibrated through experiments, a higher weight is given to the point cloud change rate to enhance the dominance of the environment change, and the two weighted items are summed to obtain a motion state index reflecting the current overall motion state. Based on the motion state index, the three-axis linear acceleration module length, i.e. the acceleration vector is subjected to module value calculation to obtain the current linear acceleration amplitude of the robot, the power mutation amplitude of the motor in a unit time is extracted, and the acceleration module value and the power mutation variable are subjected to normalization processing to make their dimensions consistent and then participate in accumulation calculation to obtain a generalized likelihood ratio statistic. The generalized likelihood ratio statistic integrates multiple dynamic information such as environment change, posture rotation, linear acceleration fluctuation and driving energy change, and is used as a core basis for judging whether the robot is in an abnormal zero speed state. A sliding time window is used to continuously scan the generalized likelihood ratio statistic, the length of the time window is determined according to the control period and the environment response characteristics, and is set to be in the range of 0.5 seconds to 1 second. The mean value, standard deviation and peak value of the statistic in the window are compared with a preset threshold point by point, when a plurality of consecutive sampling points exceed the set threshold, the abnormal state is marked as starting, and when a plurality of consecutive points fall below the threshold, the abnormal state is marked as ending, so that a complete zero speed anomaly interval is determined and the start time and the end time are recorded. According to the statistical characteristics in the zero speed anomaly interval such as the duration, the peak intensity, the average amplitude and the oscillation frequency, a detection threshold is dynamically created according to the current ground reflection characteristics and the environment echo density.

[0031] In a specific embodiment, the execution step can specifically include the following steps: setting a time window and positioning the scanning start position at the first data point of the generalized likelihood ratio statistic; comparing the generalized likelihood ratio statistic within the time window with the preset threshold to obtain a threshold exceeding state marker for each time point within the time window; sequentially marking the time points continuously exceeding the preset threshold and recording the position coordinates on the time axis to obtain continuous sequence data; detecting the first time point exceeding the threshold in the continuous sequence data as the start time and the last time point exceeding the threshold as the end time to obtain the zero-speed anomaly interval containing the start time and the end time.

[0032] Specifically, a sliding time window structure is established, and the scanning start position of the time window is aligned with the first valid data point in the generalized likelihood ratio statistic sequence to ensure that no potential abnormal change is missed in the scanning process. The window length is set to be between 0.5 seconds and 1 second according to the sampling frequency and processing delay, and a fixed sliding step is configured, such as one sampling point per sliding, to realize fine-grained point-by-point evaluation of the statistic sequence. After initialization, the sliding scanning process is started, and the generalized likelihood ratio statistic contained in the time window is compared with the preset environmental detection threshold in the system. The judgment logic is executed for the statistic value of each time point: if the generalized likelihood ratio statistic is greater than the current detection threshold, it is determined to be “exceeding state”, otherwise it is marked as “non-exceeding state”. Each judgment result is marked as a binary state and recorded at the position corresponding to the time point, forming an exceeding state marker vector consistent with the length of the original statistic. The exceeding state marker vector is subjected to logical continuity analysis, and the entire sequence is traversed to find segments of “exceeding state” that appear continuously, and the start index and end index in each continuous state are recorded on the global time axis, corresponding to the specific sampling time coordinates in the original statistic sequence, forming multiple candidate continuous exceeding sequences. In each candidate sequence, the start point is extracted, the first time point exceeding the threshold is determined as the start time of the abnormal state, and the last time point still in the exceeding state is positioned as the end time, to constitute a zero-speed anomaly interval with a clear start time and end time and a clear boundary. At the same time, multiple anomaly intervals are merged, and when the interval between two zero-speed anomaly intervals is less than a set minimum recovery duration threshold (such as 0.2 seconds), they are automatically merged into a complete abnormal segment, preventing fragmentation of the abnormal region caused by temporary signal fluctuations, thereby enhancing the continuous identification ability of the stable stuck state.

[0033] In a specific embodiment, the process of creating a detection threshold for the current ground environment according to the zero-speed abnormal interval can specifically include the following steps: Calculate the mean and standard deviation of the generalized likelihood ratio statistic in the zero-speed abnormal interval, and simultaneously calculate the duration and zero-speed proportion of the zero-speed abnormal interval to obtain the statistical characteristic parameters of the zero-speed abnormal interval; Match the statistical characteristic parameters with the preset trapped mode feature library to obtain the type identification and corresponding learning rate coefficient of the current trapped event; Calculate the actual trapped rate based on the number of trapped events in the continuous cleaning task, and perform correction operation on the actual trapped rate and the preset target trapped rate to obtain the trapped rate deviation correction amount; According to the type identification of the current trapped event, update the threshold value of the statistical characteristic parameters, the learning rate coefficient and the trapped rate deviation correction amount to obtain the detection threshold for the current ground environment.

[0034] Specifically, a high-dimensional statistical operation is performed on each identified zero-speed abnormal interval to extract key parameters describing the behavior characteristics of the abnormal interval. The total generalized likelihood ratio statistic value within the time period is extracted in each zero-speed abnormal interval, and the mean and standard deviation thereof are calculated based on the generalized likelihood ratio statistic value, wherein the mean reflects the average intensity of the system deviating from the regular motion during the abnormal state, and the standard deviation reflects the fluctuation amplitude of the abnormal intensity; at the same time, the total duration of the zero-speed abnormal interval is recorded, and the number of sampling points in the interval that are judged to be in the zero-speed state is counted, and then the total sampling points in the interval are subjected to ratio operation to obtain the zero-speed proportion, which effectively distinguishes the strength of the zero-speed persistence under different entrapment situations. The above-mentioned mean, standard deviation, duration and zero-speed proportion form the statistical characteristic parameter set of the zero-speed abnormal interval. The built-in entrapment mode characteristic library is called to perform item-by-item matching comparison on the current statistical characteristic parameters. The statistical characteristic boundaries of a plurality of typical entrapment modes are preset in the entrapment mode characteristic library, for example, the statistical characteristic of the wire entanglement mode corresponds to a low duration, a medium zero-speed proportion and a large fluctuation; the edge of the carpet entrapment corresponds to a medium duration, a jumping statistical value and a zero-speed proportion close to the median level; and the threshold entrapment has a very high zero-speed proportion, a long duration and a stable high value of the statistical mean. The type identifier corresponding to the current entrapment event is output by the feature matching algorithm system, and the learning rate coefficient associated with the matching result is obtained from the matching result. The setting of the learning rate coefficient follows the principle of "the more significant the risk, the faster the adjustment". For example, a higher learning rate is configured for the threshold entrapment scene to accelerate the convergence of the threshold. In order to realize the environmental adaptability evolution in the long-term operation, a statistical feedback mechanism is introduced to accumulate the number of entrapment events in the continuous cleaning task execution process, and the actual entrapment rate is calculated in combination with the total number of cleaning tasks. The actual entrapment rate is compared with the preset target entrapment rate, and the entrapment rate deviation correction amount is obtained by difference operation. If the actual entrapment rate is significantly higher than the target value, it indicates that the current threshold setting is too loose, which leads to the entrapment not being blocked in time, and then the sensitivity threshold is reduced by correction; otherwise, it indicates that the system misjudges the normal fluctuation as entrapment, and the threshold tolerance needs to be improved to avoid false triggering. The corresponding threshold updating strategy is called by taking the current event type identifier as an index, and the linear incremental correction model is used for fusion update, that is, the original detection threshold is taken as the benchmark, the mean and standard deviation of the generalized likelihood ratio statistic are taken as dynamic weight factors, and the learning rate coefficient and the entrapment rate deviation correction amount are combined to participate in the correction calculation, and the new detection threshold is output. The new detection threshold takes effect in the next zero-speed identification or generalized likelihood ratio discrimination.

[0035] In a specific embodiment, the process of step S300 can specifically include the following steps: extracting coordinate data of the zero-speed abnormal position of the intelligent cleaning robot in the zero-speed abnormal interval, and reading the point cloud density information of the laser radar sensor at the zero-speed abnormal position; According to the number of obstacle points in the point cloud density information and the preset blind area determination threshold, it is confirmed that the zero speed abnormal position exists the laser radar scanning blind area; According to the laser radar scanning blind area, the height of the low obstacle is calculated by using the IMU acceleration change amplitude in the zero speed abnormal interval, and the height of the low obstacle is data bound with the coordinate data to obtain the low obstacle space information; Based on the low obstacle space information, the height attribute is added in the laser radar grid map data structure, and the occupation state and confidence of the corresponding grid are updated to obtain the obstacle distribution map.

[0036] Specifically, the coordinate data of the intelligent cleaning robot zero-speed abnormal position in the zero-speed abnormal interval is extracted, and the coordinate sources include the pose estimation results in the real-time SLAM mapping system and the time stamps corresponding to the recording time of the generalized likelihood ratio statistics, and each abnormal state is mapped to the spatial position in the laser radar coordinate system through the time indexing mechanism. At the same time, the point cloud data frame of the laser radar sensor at the corresponding coordinate position is read, and the point cloud density information in the fan-shaped area with the position as the center and the preset radius of 0.3 meters is counted, wherein the point cloud density is defined as the number of points marked as obstacles in a unit area or a unit angle range, which is used as a quantitative index reflecting the current radar perception integrity. The statistical point cloud density value is compared with the preset blind area determination threshold value, if the number of obstacle points is less than the threshold value (for example, less than 5) and the state overlaps with the zero-speed abnormal period for more than 2 seconds, it is determined that there is a laser radar scanning blind area at the current zero-speed abnormal position, and it is inferred that there is a low obstacle that the laser radar cannot identify due to the installation height or the viewing angle limitation in front of or below the robot. After confirming the existence of the blind area, the acceleration data of the IMU six-axis sensor in the abnormal interval is used, and the acceleration change amplitude in the vertical direction (usually the Z axis) is focused on; by analyzing the acceleration mutation value and the transfer function between the robot chassis structure, the height change corresponding to the unit acceleration mutation is calculated, and the height conversion coefficient is used to convert the acceleration change into the estimated height of the obstacle, and the upper limit is set to not more than 15 cm to avoid misjudgment of too high obstacles due to local non-contact vibration. The estimated height value is bound with the abnormal position coordinate to form the spatial information of the low obstacle with clear two-dimensional coordinate and corresponding height value, which constitutes the core compensation data in the map completion process. The existing ranging map data structure is expanded, and each position grid point in the original map only records the obstacle occupancy state, while in the current implementation, each grid point is expanded to three contents, namely the occupancy state, the obstacle height value and the confidence degree index. In the specific data writing process, when a position is marked as a laser blind area and has a clear height estimation value, the occupancy state is marked as occupied, the height value field is written with the height calculated by the acceleration mutation, and the confidence degree is valued according to the strength of the generalized likelihood ratio statistics and the stability of the acceleration mutation amplitude in this period. The confidence degree and the motion state intensity are positively correlated through function mapping, and the higher the intensity, the greater the confidence degree. In order to avoid local discontinuous labeling from causing interference to path planning, a buffer area is set around the low obstacle coordinate, and a certain confidence degree is assigned to the buffer grid point to construct a risk diffusion model. After all the above data is written, an obstacle distribution map with three information is output.

[0037] In a specific embodiment, the process of performing step S400 can specifically include the following steps: The zero-speed detection intensity is calculated according to the generalized likelihood ratio statistic and a detection threshold, and the zero-speed detection intensity is multiplied by a power adjustment coefficient and added to a basic suction force to obtain a suction power; A first ratio of the generalized likelihood ratio statistic and the detection threshold is calculated, and when the first ratio is in different intervals, corresponding movement speed adjustment parameters are triggered respectively; A real-time current when the suction motor is running is monitored, and a second ratio of the real-time current and a reference current is calculated, and when the second ratio deviates from a normal range, the suction power and the movement speed adjustment parameter are reversely adjusted to obtain cleaning control parameters; A re-planned cleaning path is generated based on an obstacle distribution map and the cleaning control parameters.

[0038] Specifically, a zero-speed detection response mechanism is established with the generalized likelihood ratio statistic as the core index. In each cleaning cycle, the navigation control module calculates the value of the current generalized likelihood ratio statistic in real time, and performs difference operation on the generalized likelihood ratio statistic and the updated detection threshold value to obtain a non-negative result, which is regarded as the current zero-speed detection intensity. The higher the zero-speed detection intensity value, the more likely it is that the robot is in an abnormal area with increased adhesion, blockage or resistance. Through the suction control module of the cleaning subsystem, the zero-speed detection intensity is multiplied by a set of preset power adjustment coefficients, which represent the contribution of unit intensity to the suction power increase amplitude. The product is added to the basic suction, which is 60% of the rated power of the suction motor, to obtain the suction power that should be output at the current time. The output suction power is converted by the PWM duty cycle conversion circuit to act on the suction motor, ensuring that the robot has completed the suction pressure boosting process before entering the high resistance area or being trapped in a static state, thereby improving the ability to escape and remove residual materials. At the same time, based on the generalized likelihood ratio statistic and the detection threshold value at the same time point, a first ratio is calculated, which is the dimensionless ratio obtained by dividing the statistic by the threshold value, reflecting the relative severity of the current abnormal state. When the first ratio is less than 1.0, it is considered to be in the normal operating interval, at which time the standard speed parameter is called to maintain the rated travel speed. When the ratio is between 1.0 and 2.0, it is identified as a medium resistance state, and the moving speed is reduced to 80% of the normal value to prolong the sampling coverage time of the laser radar and IMU on the environmental features. If the ratio exceeds 2.0, it enters a high resistance and high risk state, enabling the speed to be reduced to 50% or lower, and preparing to perform reciprocating cleaning or local path reconstruction operations in the control logic. Such speed adjustment parameters are directly transmitted by the motion control module to the drive motor control unit for dynamic limiting of current, voltage and PWM pulse period, ensuring that the speed response process is synchronized with the change in detection intensity. To prevent errors in the system adjustment of suction and speed process from deviating from the actual load situation, the real-time working current of the suction motor is continuously sampled, and the real-time current is divided by the baseline current recorded during no-load operation to obtain a second ratio, which reflects the actual air resistance or dust load on the current suction assembly. When the second ratio is stable within the set normal interval (e.g. 1.0 to 1.5), the predetermined suction and speed parameters remain unchanged. When the ratio is below the lower limit, it indicates that the suction is too strong but the load is insufficient, so the suction is lowered to save energy and reduce noise, and the speed is appropriately increased to prevent repeated coverage of local areas. If the ratio exceeds the upper limit, it indicates that the current suction is insufficient to overcome the environmental resistance or has entered a heavily polluted area, at which time the suction power upper limit is increased, and the speed upper limit is simultaneously lowered to ensure that the cleaning effect takes priority. The above adjustment process is completed by a feedback loop, and the generated suction-speed combination constitutes the cleaning control parameters for the current period, which are used to describe the optimal cleaning state response of the robot in the current environment. Based on the obstacle distribution map and the cleaning control parameters, a new path cost function is constructed.The path cost considers the traditional distance and turning cost terms, and incorporates the suction strength and speed lower limit in the cleaning control parameters into the constraint weight of the cost function. In particular, the passage cost is automatically increased in the low obstacle area with high confidence, prompting the system to preferentially bypass the high suction and high resistance area during path planning. Meanwhile, the system selects the area with low confidence and no significant cleaning pressure to pass through preferentially when the path allows, so as to generate a re-planned cleaning path that avoids known or estimated obstacle points while ensuring that the cleaning task is completed with minimum energy consumption and optimal suction matching degree on the premise of maintaining high coverage.

[0039] In a specific embodiment, the process of generating the re-planned cleaning path based on the obstacle distribution map and the cleaning control parameters can specifically include the following steps: Extract the height attribute, confidence and coordinate data of the zero-speed abnormal interval of each grid in the obstacle distribution map, and calculate the spatial risk distribution data of each grid position; Multiply the spatial risk distribution data by the movement speed adjustment parameter in the cleaning control parameter, and then linearly combine it with the path distance cost according to a preset proportion coefficient to obtain a path total cost function; Perform a shortest path search from the current position of the intelligent cleaning robot to the target cleaning area using the path total cost function to obtain a path set; Compare the total cost values of each candidate path in the path set, select the path with the minimum cost as the main cleaning path, and mark the high-risk area as a delayed cleaning area to obtain a re-planned cleaning path.

[0040] Specifically, the core attribute information of all updated grids is extracted from the current obstacle distribution map, including the estimated height value of the obstacle corresponding to each grid unit, the confidence level of the height data, and the coordinate position record accumulated in the historical zero-speed abnormal interval. These information respectively reflect the potential restriction degree of the grid in the vertical direction, the reliability of the information source, and the frequency density of the position being judged as an abnormal stationary area in the historical task. Based on the above basic data, the spatial risk modeling process is performed, and the spatial risk distribution value of each grid is independently calculated by constructing a spatial risk function containing three factors. The spatial risk function is designed as a linear amplification item of height value and an exponential weighting item of confidence, and is combined with the cumulative probability density of the position in the historical zero-speed abnormal data for normalization processing, so as to output a numerical index reflecting the overall traffic risk of the grid. At the same time, the parallel scanning method is used to perform grid-level risk update on the whole map, so as to ensure that the map risk distribution is updated in real time before each path reconstruction is called. The spatial risk distribution data and the movement speed adjustment parameter in the cleaning control parameter are interactively integrated, wherein the movement speed adjustment parameter represents the response state of the robot to the local resistance, and the lower the speed, the more complex the corresponding area may be. Therefore, the movement speed adjustment parameter is taken as an amplification factor of the risk value, so that the influence of the spatial risk on the path cost function is further enhanced in the area where the robot has lowered the speed. The risk value of each grid is combined with the speed adjustment parameter to generate a comprehensive risk correction factor. The risk correction factor and the traditional path cost are linearly combined, that is, on the basis of path length or displacement distance, the risk evaluation corresponding to the position is superimposed, to obtain a path total cost function. The path total cost function is used to perform a shortest path search algorithm based on the grid connection relationship between the current pose and the target cleaning area. Improved A*, Dijkstra or D* Lite heuristic path search methods are selected, and the total cost function cumulative value of each path passing through the area is evaluated in real time during the search process. All feasible paths are collected as a path set, and their total cost, number of inflection points, estimated cleaning coverage, and other information are recorded. After the search is completed, the path selection module traverses the path set, compares the total cost sums of the paths, and selects the path with the minimum total cost as the main cleaning path of the current task period. After the path selection is completed, the grid units outside the coverage range of the main path but marked as high-risk areas in the map are checked in reverse, and these areas are excluded from the current task cleaning priority queue and marked as "delayed cleaning areas". The output is a re-planned cleaning path.

[0041] In a specific embodiment, the process of performing a shortest path search on the current position of the intelligent cleaning robot to the target cleaning area using the path total cost function to obtain a path set can specifically include the following steps: The current position of the intelligent cleaning robot is taken as a path search starting point, the coordinate range of the target cleaning area is read, and a search end position is determined; The obstacle distribution map is divided into grid nodes at a fixed interval, the passing cost value of each grid node is calculated by substituting the path total cost function, and a grid search space is generated according to the passing cost value; From the path search starting point, adjacent grid nodes are expanded, when it is detected that the passing cost value of the grid node exceeds the safety threshold, a bypass branch path is automatically generated, and the search is continued until the search end position is reached, and a path set is obtained.

[0042] Specifically, the real-time pose data of the intelligent cleaning robot in the SLAM coordinate system is obtained, and the two-dimensional plane coordinates of the current position are taken as the starting node of the path search algorithm, i.e. the starting point of path search. The boundary description information of the target cleaning area is read from the task scheduling module. The boundary description information is a closed region composed of a group of coordinate points or a geographical range corresponding to a series of preset region numbers. The system needs to analyze and locate the center point or specified endpoint in the region as the search endpoint coordinates to ensure that the path search target has reachability and spatial directionality. The latest version of the obstacle distribution map is called, and regular grid processing is performed on the entire map according to the set spatial resolution, with a spacing of 5-10 cm. Each node in the grid division process is taken as a potential passing point, and its horizontal and vertical coordinates, occupancy state, obstacle height value, risk confidence, and map boundary constraint relationship are preserved. After completing node division, the path total cost function is taken as the evaluation model, and each grid node is substituted into the passing cost calculation one by one. The path total cost function integrates multiple factors, including path physical length, turning angle, obstacle distribution height and density, historical zero-speed anomaly frequency, sensor perception confidence, and moving speed correction factor under the current cleaning control parameters. After the calculation is completed, a two-dimensional search space graph containing all grid node cost values is generated. Each node in the two-dimensional search space graph has spatial coordinate information and passing cost attributes, and forms a cost heat distribution graph through color or gradient coding form for path algorithm visualization analysis and dynamic evaluation. In the generated search space, all impassable nodes (such as grids completely blocked by obstacles or with a confidence of 1) are marked as "closed nodes", while nodes with high cost but not locked are preserved as potential passable regions. Through dynamic cost adjustment strategy, the jump boundary processing is allowed in path search, enhancing the feasibility of dealing with narrow passages and complex geometric structure regions. From the path search starting point, the search process is carried out based on the improved A search algorithm or D Lite dynamic path adjustment algorithm. In the initial stage, the node with the smallest total cost in the adjacent grid around the starting point is selected as the first expansion direction, and the surrounding adjacent nodes are expanded layer by layer according to the eight-connected principle. In each node expansion process, the passing cost of the current candidate node is detected and compared with the safety threshold, which is dynamically set according to the current cleaning mode (such as standard mode, enhanced mode or energy-saving mode) and set as the critical sensitive value in the cost function. If the total cost of a node exceeds the safety threshold, it is considered as a potential risk area, and the detour logic is triggered immediately to automatically re-expand the adjacent nodes of the suboptimal cost direction around the current node and build a detour branch path, which is synchronized into the search path tree with the main path to ensure that the main path has backup path connection and access capability when encountering high-risk obstacles.This search-avoidance process will continue to develop throughout the path graph until a certain extended path is successfully connected to the end node for the first time, that is, the path is recorded as the first feasible path, and then the path algorithm still continues to explore other possible paths to generate a complete path set, ensuring that the system can quickly switch to the suboptimal path to complete the task when the main path is disturbed or misjudged. In the output path set, each path contains its cumulative total cost, path length, inflection point number and risk hit ratio and other indicators.

[0043] Please refer to Figure 2 , Figure 2 The structural schematic block diagram of the multi-sensor fusion navigation system of the robot provided in the embodiment of the present application is shown in Figure 2 The multi-sensor fusion navigation system of the robot comprises: The acquisition module 210 is configured to acquire point cloud data of the laser radar sensor and motion data of the IMU six-axis sensor. The zero speed anomaly detection module 220 is configured to perform generalized likelihood ratio test and sliding window interval search on the point cloud data and the motion data to obtain a zero speed anomaly interval, and create a detection threshold for a current ground environment according to the zero speed anomaly interval. The spatial superposition module 230 is configured to perform spatial superposition on position coordinates of the zero speed anomaly interval and a laser radar scanning blind area to obtain an obstacle distribution map. The generation module 240 is configured to adjust a cleaning control parameter based on the detection threshold, and generate a re-planned cleaning path based on the obstacle distribution map and the cleaning control parameter.

[0044] Through the cooperation of the above-mentioned components, the application constructs a comprehensive motion state evaluation system by fusing the 5-10Hz point cloud data of the laser radar and the 100Hz six-axis motion data of the IMU in time asynchronization but space synchronization, breaks through the single sensor perception limitation, can accurately identify normal motion modes such as straight cleaning, turning, edge following, etc., proposes a generalized likelihood ratio test statistic and a sliding window interval search mechanism, extends the zero speed detection from the traditional "complete static" concept to "abnormal low speed motion" detection, can accurately locate the start and end time boundaries of the zero speed anomaly, and effectively distinguishes different types of trapping events such as wire entanglement, carpet edge entrapment, threshold entrapment, etc. An adaptive threshold updating mechanism based on trapping type identification is established, the detection parameters are dynamically adjusted according to different ground environments and obstacle characteristics, the technical problem that the fixed threshold cannot adapt to complex environment changes is solved, the misjudgment rate is significantly reduced and the detection accuracy is improved. By fusing the zero speed detection result and the laser radar scanning data, a high blind area compensation mechanism is established, which can infer the position and height information of low obstacles, effectively solving the problem of low obstacle detection blind area. A intelligent mapping relationship between zero speed detection intensity and dust cleaning power, brush state is established, which can dynamically adjust the cleaning parameters according to the risk degree of entrapment, realizes the coordinated optimization of cleaning effect and motion safety, and automatically starts the anti-entanglement processing program when periodic zero speed mode is detected. The application constructs a Bayesian network trapping risk prediction model, which can predict potential trapping areas in the future and actively adjust the cleaning path, integrates the risk weight into the path cost function, realizes the technical transformation from passive obstacle avoidance to active prevention, and effectively improves the completion rate of the cleaning task and the autonomy of the robot.

[0045] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, system and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0046] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0047] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-sensor fusion navigation method for a robot, characterized in that, include: Collect point cloud data from the lidar sensor and motion data from the IMU six-axis sensor; Perform generalized likelihood ratio test and sliding window interval search on the point cloud data and the motion data to obtain the zero-velocity anomaly interval, and create a detection threshold for the current ground environment based on the zero-velocity anomaly interval; The location coordinates of the zero-velocity abnormal zone are spatially superimposed with the blind zone of the lidar scan to obtain an obstacle distribution map. The cleaning control parameters are adjusted based on the detection threshold, and a replanning cleaning path is generated based on the obstacle distribution map and the cleaning control parameters.

2. The multi-sensor fusion navigation method for robots according to claim 1, characterized in that, The acquisition of point cloud data from the lidar sensor and motion data from the IMU six-axis sensor includes: Set the scanning frequency of the lidar sensor in the intelligent cleaning robot and start the environmental scanning mode to obtain the first acquisition configuration parameters; Set the sampling frequency of the IMU six-axis sensor in the intelligent cleaning robot and start the synchronous acquisition mode of the accelerometer and gyroscope to obtain the second acquisition configuration parameters; Based on the first acquisition configuration parameters, the distance measurement value and reflection intensity value of the environment around the intelligent cleaning robot are obtained to obtain point cloud data; Based on the second acquisition configuration parameters, the real-time measurement values ​​of the three-axis linear acceleration and three-axis angular velocity of the intelligent cleaning robot are synchronously acquired to obtain motion data.

3. The multi-sensor fusion navigation method for robots according to claim 2, characterized in that, The process of performing a generalized likelihood ratio test and a sliding window interval search on the point cloud data and the motion data to obtain zero-velocity anomaly intervals, and creating a detection threshold for the current ground environment based on the zero-velocity anomaly intervals, includes: The change in Euclidean distance between adjacent frame point cloud data is calculated as the point cloud change rate, and the composite magnitude of the three-axis angular velocities in the motion data is also calculated. The motion state index of the intelligent cleaning robot is obtained by multiplying the point cloud change rate and the synthetic modulus by the corresponding weight coefficients and then summing them. Based on the motion state index, the generalized likelihood ratio statistic is obtained by normalizing and accumulating the acceleration vector magnitude and motor power mutation in the motion data. The generalized likelihood ratio statistic is continuously scanned using a preset time window, and continuous time periods exceeding a preset threshold are marked to obtain a zero-velocity anomaly interval including the start and end times. A detection threshold for the current ground environment is created based on the zero-velocity anomaly interval.

4. The multi-sensor fusion navigation method for robots according to claim 3, characterized in that, The method of continuously scanning the generalized likelihood ratio statistic using a preset time window and marking continuous time periods exceeding a preset threshold yields a zero-velocity anomaly interval including a start time and an end time, including: Set a time window and position the scan start point at the first data point of the generalized likelihood ratio statistic; The generalized likelihood ratio statistic within the time window is compared with a preset threshold to obtain the threshold exceedance status flag for each time point within the time window. The time points that continuously exceed the preset threshold are sequentially marked and their position coordinates are recorded on the time axis to obtain continuous sequence data; The first time point in the continuous sequence data that exceeds the threshold is taken as the start time and the last time point that exceeds the threshold is taken as the end time, thus obtaining a zero-velocity anomaly interval that includes the start time and the end time.

5. The multi-sensor fusion navigation method for robots according to claim 4, characterized in that, The step of creating a detection threshold for the current ground environment based on the zero-velocity anomaly interval includes: Calculate the mean and standard deviation of the generalized likelihood ratio statistic within the zero-velocity anomaly interval, and simultaneously calculate the duration and proportion of zero velocity within the zero-velocity anomaly interval to obtain the statistical characteristic parameters of the zero-velocity anomaly interval. The statistical feature parameters are matched with a preset stuck-out mode feature library to obtain the type identifier and corresponding learning rate coefficient of the current stuck-out event; The actual jamming rate is calculated based on the number of jamming events in the continuous cleaning task, and the actual jamming rate is corrected by comparing it with the preset target jamming rate to obtain the jamming rate deviation correction amount; Based on the type identifier of the current stuck event, the statistical feature parameters, the learning rate coefficient, and the stuck rate deviation correction amount are updated with thresholds to obtain the detection threshold for the current ground environment.

6. The multi-sensor fusion navigation method for robots according to claim 1, characterized in that, The step of spatially overlaying the position coordinates of the zero-velocity anomaly zone with the blind zone of the lidar scan to obtain an obstacle distribution map includes: Extract the coordinate data of the zero-speed abnormal position of the intelligent cleaning robot in the zero-speed abnormal range, and read the point cloud density information of the lidar sensor at the zero-speed abnormal position. Based on the number of obstacle points in the point cloud density information and the preset blind zone determination threshold, it is confirmed that there is a lidar scanning blind zone at the zero-velocity abnormal location. Based on the blind zone of the lidar scan, the height of low obstacles is estimated by using the IMU acceleration change amplitude within the zero-velocity anomaly range, and the height of low obstacles is data-bound with the coordinate data to obtain the spatial information of low obstacles; Based on the spatial information of the low-lying obstacles, a height attribute is added to the lidar grid map data structure, and the occupancy status and confidence of the corresponding grid are updated to obtain an obstacle distribution map.

7. The multi-sensor fusion navigation method for robots according to claim 1, characterized in that, The step of adjusting cleaning control parameters based on the detection threshold and generating a replanning cleaning path based on the obstacle distribution map and the cleaning control parameters includes: The zero-speed detection intensity is calculated based on the generalized likelihood ratio statistic and the detection threshold. The zero-speed detection intensity is then multiplied by the power adjustment coefficient and added to the basic suction power to obtain the suction power. Calculate the first ratio of the generalized likelihood ratio statistic to the detection threshold, and trigger the corresponding movement speed adjustment parameter when the first ratio is in different intervals. The real-time current of the vacuum cleaner motor during operation is monitored and a second ratio of the real-time current to the reference current is calculated. When the second ratio deviates from the normal range, the vacuuming power and the moving speed adjustment parameters are adjusted in reverse to obtain cleaning control parameters. A replanning cleaning path is generated based on the obstacle distribution map and the cleaning control parameters.

8. The multi-sensor fusion navigation method for robots according to claim 7, characterized in that, The process of generating a replanning cleaning path based on the obstacle distribution map and the cleaning control parameters includes: Extract the height attribute, confidence level, and coordinate data of the zero-velocity anomaly interval of each grid cell in the obstacle distribution map, and calculate the spatial risk distribution data of each grid cell location; Multiply the spatial risk distribution data by the moving speed adjustment parameter in the cleaning control parameters, and then linearly combine it with the path distance cost according to a preset proportional coefficient to obtain the total path cost function; The path total cost function is used to search for the shortest path from the current position of the intelligent cleaning robot to the target cleaning area, and a set of paths is obtained. The total cost values ​​of each candidate path in the path set are compared, and the path with the lowest cost is selected as the main cleaning path. At the same time, high-risk areas are marked as delayed cleaning areas, thus obtaining the replanned cleaning path.

9. The multi-sensor fusion navigation method for robots according to claim 8, characterized in that, The method employs the total path cost function to perform a shortest path search from the current position of the intelligent cleaning robot to the target cleaning area, resulting in a path set, including: The current position of the intelligent cleaning robot is used as the starting point for path search. The coordinate range of the target cleaning area is read and the search endpoint is determined. The obstacle distribution map is divided into grid nodes at fixed intervals. The total path cost function is substituted into each grid node to calculate the passage cost value, and a grid search space is generated based on the passage cost value. Starting from the path search starting point, adjacent grid nodes are expanded. When the passage cost value of a grid node is detected to exceed the safety threshold, a detour branch path is automatically generated. The search continues until the search endpoint is reached, resulting in a path set.

10. A multi-sensor fusion navigation system for a robot, characterized in that, A multi-sensor fusion navigation method for performing a robot as described in any one of claims 1-9 includes: The acquisition module is used to acquire point cloud data from the lidar sensor and motion data from the IMU six-axis sensor; The zero-velocity anomaly detection module is used to perform generalized likelihood ratio test and sliding window interval search on the point cloud data and the motion data to obtain the zero-velocity anomaly interval, and to create a detection threshold for the current ground environment based on the zero-velocity anomaly interval. The spatial overlay module is used to spatially overlay the position coordinates of the zero-velocity abnormal zone with the blind zone of the lidar scan to obtain an obstacle distribution map. The generation module is used to adjust the cleaning control parameters based on the detection threshold, and to generate a replanning cleaning path based on the obstacle distribution map and the cleaning control parameters.

Images