Method for realizing automatic patrol intelligent vehicle path planning based on environment sensor

Abstract

The present application relates to the technical field of non-electric variable control system, and particularly relates to a method for realizing automatic patrol intelligent vehicle path planning based on environment sensors, comprising: planning a global path of an automatic patrol task on a patrol area grid map, obtaining a plurality of environment dimension time series when the intelligent vehicle moves along the global path, obtaining a significant attention factor of each key attention moment on each environment dimension, thereby obtaining an environment anomaly detection weight of each moment, performing weighted prediction anomaly detection on each environment dimension time series, obtaining an abnormal point position on the global path, determining a circular temporary obstacle area, thereby performing local path re-planning, generating a new local path, guiding the intelligent vehicle to return to the global path after bypassing the circular temporary obstacle area, and until the automatic patrol task is completed. The present application improves the efficiency and safety of the intelligent vehicle automatic patrol by accurately identifying the point position with environment anomaly and performing local path re-planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of non-electric variable control systems, and in particular to a method for planning a path of an intelligent vehicle for automatic inspection based on environmental sensors. BACKGROUND

[0002] The method for planning a path of an intelligent vehicle for automatic inspection based on environmental sensors first collects topography, static obstacle distribution, etc. of an inspection area through multi-source environmental sensors such as laser radars, visual cameras, and millimeter wave radars carried by the intelligent vehicle, so as to construct a grid map. Then, an initial globally optimal inspection path is generated on the grid map. In the inspection process, environmental data collected by the sensors is monitored in real time. When an environmental anomaly (such as local high temperature, gas leakage, or excessive radiation) is identified, a local path re-planning mechanism based on a dynamic window approach (DWA) is triggered to generate a new local path for obstacle avoidance, which is smoothly connected with the global path. The present method is aimed at the automatic inspection requirements of industrial plants, agricultural parks, power inspection, etc. and realizes dynamic planning and real-time adjustment of the path relying on the sensing data of environmental sensors, thereby solving the problems of poor flexibility and weak environmental adaptability of traditional fixed path inspection, and having both global coverage and local obstacle avoidance capability, so as to improve the efficiency and safety of the intelligent vehicle inspection.

[0003] Existing problems: During the movement of the intelligent vehicle, the environment itself will have normal dynamic changes. When the intelligent vehicle moves along the global path through different sub-scenarios (such as tunnels, intersections, and edges of industrial areas), the normal environmental parameter baseline values of different sub-scenarios differ. For example, the baseline concentration of carbon monoxide near an industrial area is higher than that in a residential area. The existing environmental anomaly judgment usually uses a unified global threshold. When the intelligent vehicle enters the industrial area, the normal increase in carbon monoxide concentration will be misjudged as an anomaly, triggering unnecessary alarms and causing false local path re-planning, which will reduce the efficiency and safety of the intelligent vehicle inspection. SUMMARY

[0004] The present application provides a method for planning a path of an intelligent vehicle for automatic inspection based on environmental sensors to solve the existing problems.

[0005] The method for planning a path of an intelligent vehicle for automatic inspection based on environmental sensors provided by the present application adopts the following technical scheme:

[0006] One embodiment of the present application provides a method for planning a path of an intelligent vehicle for automatic inspection based on environmental sensors, which comprises the following steps:

[0007] Planning a global path of an automatic inspection task on a grid map of an inspection area, obtaining a time sequence of several environmental dimensions when the intelligent vehicle moves along the global path, and a position coordinate at each time;

[0008] According to the difference between the data values of adjacent time points in each environmental dimension time sequence and the distance between the position coordinates of adjacent time points, a number of key attention time points are screened out from each environmental dimension time sequence, and a significant attention factor of each key attention time point on each environmental dimension is obtained;

[0009] On all environmental dimensions, according to the number of all environmental dimensions of each time point being a key attention time point and the significant attention factor, an environmental anomaly detection weight of each time point is obtained; according to the environmental anomaly detection weights of all time points, a weighted prediction anomaly detection is performed on each environmental dimension time sequence to obtain an abnormal point position on the global path;

[0010] A circular temporary obstacle region is generated with the abnormal point position on the global path as the center, a new local path is generated by starting from the abnormal point position and aiming to bypass the circular temporary obstacle region for local path re-planning, and the intelligent vehicle is guided to return to the global path after bypassing the circular temporary obstacle region, until the automatic patrol task is completed.

[0011] Further, the steps of screening a number of key attention time points from each environmental dimension time sequence include the following specific steps:

[0012] For the xth environmental dimension time sequence, the absolute value of the difference between the data values of the tth time point and the (t-1)th time point is obtained, denoted as the fluctuation of the tth time point, the Euclidean distance between the position coordinates of the tth time point and the (t-1)th time point is obtained, denoted as the movement of the tth time point, and the inverse proportional normalized value of the movement of the tth time point is obtained, denoted as the attention degree of the tth time point on the xth environmental dimension.

[0013] According to the size of the attention degree of each time point on each environmental dimension, a number of key attention time points on each environmental dimension are obtained.

[0014] Further, the steps of obtaining a number of key attention time points on each environmental dimension according to the size of the attention degree of each time point on each environmental dimension include the following specific steps:

[0015] For any one environmental dimension, all time points with an attention degree greater than a preset attention threshold are recorded as key attention time points on the any one environmental dimension.

[0016] Further, the steps of obtaining a significant attention factor of each key attention time point on each environmental dimension include the following specific steps:

[0017] On the xth environmental dimension time sequence, a time period formed by adjacent key attention time points at time is recorded as a key attention time period.

[0018] According to the difference between adjacent data values in the yth focus period on the xth environmental dimension time sequence and the length of the yth focus period, an environmental problem focus degree of the yth focus period is obtained;

[0019] According to the focus degree of each focus moment on each environmental dimension time sequence and the environmental problem focus degree of the focus period in which each focus moment is located, a significant focus factor of each focus moment on each environmental dimension is obtained.

[0020] Further, the specific steps of obtaining the environmental problem focus degree of the yth focus period include the following:

[0021] In the yth focus period on the xth environmental dimension time sequence, a difference value of a previous data value minus a next data value is obtained, if the difference value is negative, a label -1 is given, if the difference value is positive, a label 1 is given, and if the difference value is zero, a label 0 is given, and a -101 sequence composed of labels corresponding to all adjacent data values is obtained in time sequence;

[0022] In the -101 sequence, the number of all -1 and 0 is counted and recorded as a first number value, the number of all 0 and 1 is counted and recorded as a second number value, the ratio of the maximum value in the first number value and the second number value to the length of the -101 sequence is obtained and recorded as a trend uniformity, the two elements corresponding to the maximum value in the first number value and the second number value are recorded as target elements, the sum value of the number of elements between all adjacent target elements is recorded as a first sum value, the inverse ratio value of the first sum value to the length of the -101 sequence is recorded as a same trend continuity, and the mean value of the trend uniformity and the same trend continuity is recorded as a trend continuity uniformity.

[0023] The mean value of the inverse ratio value of the length of the yth focus period to the trend continuity uniformity is recorded as the environmental problem focus degree of the yth focus period.

[0024] Further, the specific steps of obtaining the significant focus factor of each focus moment on each environmental dimension according to the focus degree of each focus moment on each environmental dimension time sequence and the environmental problem focus degree of the focus period in which each focus moment is located include the following:

[0025] On the xth environmental dimension time sequence, the mean value of the focus degree of the zth focus moment and the environmental problem focus degree of the focus period in which the zth focus moment is located is recorded as the significant focus factor of the zth focus moment on the xth environmental dimension.

[0026] Further, the specific steps of obtaining the environmental anomaly detection weight of each moment include the following:

[0027] On each environmental dimension time sequence, the sum value of the environmental problem attention degree of all key attention periods is obtained, denoted as the abnormal performance of each environmental dimension;

[0028] On all environmental dimensions, all environmental dimensions at the bth moment are obtained, denoted as the attention environmental dimension;

[0029] According to the abnormal performance of the attention environmental dimension at the bth moment and the significant attention factor of the attention environmental dimension at the bth moment, the instantaneous important performance at the bth moment is obtained;

[0030] The attention environmental dimension set composed of all attention environmental dimensions at each moment is obtained;

[0031] According to the difference between the number of all attention environmental dimensions at the bth moment and the b-1th moment and the difference between the attention environmental dimension set at the b-1th moment and the bth moment, the important performance adjustment coefficient at the bth moment is obtained;

[0032] The sum value of the important performance adjustment coefficient at the bth moment and the mean value of the instantaneous important performance at the bth moment and the preset constant is denoted as the environmental anomaly detection weight at the bth moment.

[0033] Further, the specific steps of obtaining the instantaneous important performance at the bth moment include:

[0034] For the bth moment, the sum value of the abnormal performance of all attention environmental dimensions is obtained, denoted as a second sum value, the sum value of the abnormal performance of all environmental dimensions is obtained, denoted as a third sum value, the ratio of the second sum value to the third sum value is denoted as a first ratio, the mean value of the significant attention factors of all attention environmental dimensions at the bth moment is obtained, denoted as a comprehensive significant attention factor, the maximum value of the significant attention factors of all attention environmental dimensions at the bth moment is obtained, denoted as a limit significant attention factor, the mean value of the comprehensive significant attention factor and the limit significant attention factor is denoted as a first mean value, and the mean value of the first mean value and the first ratio is denoted as the instantaneous important performance at the bth moment.

[0035] Further, the specific steps of obtaining the important performance adjustment coefficient at the bth moment include:

[0036] The difference between the number of all environmental dimensions under attention at time b and the number of all environmental dimensions under attention at time b-1 is denoted as the first difference. The intersection and union of the sets of environmental dimensions under attention at time b-1 and time b are denoted as the target intersection and target union. The difference between the number of all elements in the target union and the number of all elements in the target intersection is denoted as the second difference. The ratio of the second difference to the number of all elements in the target union is denoted as the change amount of environmental dimension type. The difference between the number of all elements in the set of environmental dimensions under attention at time b and the number of all elements in the target intersection is denoted as the third difference. The ratio of the third difference to the number of all elements in the set of environmental dimensions under attention at time b is denoted as the increase amount of environmental dimension type. The average of the change amount of environmental dimension type and the increase amount of environmental dimension type is denoted as the change coefficient of environmental dimension type. The normalized value of the first difference and the average of the change coefficient of environmental dimension type are denoted as the important performance adjustment coefficient at time b.

[0037] Furthermore, the specific steps for obtaining the anomaly locations on the global path are as follows:

[0038] Obtain the sum of the environmental anomaly detection weights at all times, denoted as the fourth sum. Obtain the ratio of the environmental anomaly detection weight to the fourth sum at each time, denoted as the prediction weight at each time.

[0039] Based on the prediction weights of all times, a weighted autoregressive moving average model is used to perform weighted prediction on the time series of the xth environmental dimension to obtain a predicted value. If the absolute value of the difference between the data value at the current time and the predicted value in the time series of the xth environmental dimension is greater than a preset abnormal threshold, then the xth environmental dimension at the current time is determined to be an abnormal environmental dimension.

[0040] If there is an abnormal environment dimension among all environment dimensions at the current moment, then the current position coordinates are determined to be an abnormal point on the global path.

[0041] The beneficial effects of the technical solution of the present invention are:

[0042] In the embodiment of the present application, the global path of the automatic patrol task is planned on the patrol area grid map, the time sequence of several environment dimensions when the intelligent vehicle moves along the global path is obtained, the significant attention factor of each key attention moment on each environment dimension is obtained, and the environment anomaly detection weight of each moment is obtained, so as to perform weighted prediction anomaly detection on each environment dimension time sequence. Thus, by assigning a greater prediction weight to the moment with suspected environmental anomaly, the fitting degree of the prediction value to the suspected environmental anomaly data change rule is improved in the weighted prediction, so that the residual more accurately reflects the real anomaly, and finally the false detection rate and the missed detection rate of the anomaly detection are reduced, and the safety of subsequent automatic patrol is ensured. The abnormal point position on the global path is obtained, the circular temporary obstacle area is determined, and the local path is re-planned to generate a new local path to guide the intelligent vehicle to return to the global path after bypassing the circular temporary obstacle area until the automatic patrol task is completed. Thus, the present application can accurately identify the point position with environmental anomaly, re-plan the local path, realize real-time environmental danger perception and active avoidance, and ensure the safety of the intelligent vehicle itself. Thus, the environmental perception information is deeply integrated into the path planning decision-making closed loop, realizes the leap from "passive execution" to "active response", and according to the detected abnormal point position, updates the patrol point position list of the next automatic patrol, ensures the continuous and efficient monitoring of the abnormal area, avoids the cost and risk of manual review, and finally improves the efficiency and safety of the intelligent vehicle automatic patrol. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0044] Figure 1 The steps flow chart of the method for planning the path of the automatic patrol intelligent vehicle based on the environmental sensor of the present application;

[0045] Figure 2 The schematic diagram for screening the environment dimension at each moment. DETAILED DESCRIPTION

[0046] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the following describes in detail the specific implementation, structure, features and effects of the method for realizing automatic patrol intelligent vehicle path planning based on environmental sensors according to the present application, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0048] The specific scheme of the method for realizing automatic patrol intelligent vehicle path planning based on environmental sensors provided by the present application is described in detail below in combination with the accompanying drawings.

[0049] Please refer to Figure 1 which shows the step flowchart of the method for realizing automatic patrol intelligent vehicle path planning based on environmental sensors provided by one embodiment of the present application, which includes the following steps:

[0050] Step S001: planning a global path of the automatic patrol task on a patrol area grid map, obtaining a plurality of environmental dimension time series when the intelligent vehicle moves along the global path, and a position coordinate at each time.

[0051] In this embodiment, the intelligent vehicle is equipped with an environmental sensor array (including but not limited to temperature, humidity, carbon dioxide, carbon monoxide, sulfur dioxide, radiation sensors), a positioning module (GPS), a laser radar, a camera and a central processing unit. First, the central processing unit loads a high-precision grid map of the patrol area constructed in advance through SLAM technology. Then, the user sets a plurality of patrol points that need to be passed through in turn for each patrol task. Finally, the A algorithm is used to calculate the optimal global path connecting all the patrol points.

[0052] Among them, the SLAM technology and the A algorithm are all known technologies, and the specific method is not introduced here. SLAM (Simultaneous Localization and Mapping) is a core technology for realizing autonomous navigation in the fields of mobile robots, autonomous driving, unmanned aerial vehicles, etc. A algorithm (A-star algorithm) is a classic heuristic search algorithm, which is widely used in intelligent vehicle path planning, robot navigation, game map routing and other scenes.

[0053] Thus, the global path of the automatic patrol task is planned on the patrol area grid map, the time sequence of the environment dimensions when the intelligent vehicle moves along the global path is obtained, and the position coordinates at each time are obtained.

[0054] The environment dimensions collected by the environment sensor array carried by the intelligent vehicle in real time include temperature, humidity, carbon dioxide concentration, carbon monoxide concentration, sulfur dioxide concentration, and radiation intensity. The position coordinates of the intelligent vehicle on the patrol area grid map are collected by the GPS in real time. The collection frequency is once per second. In this embodiment, a sliding average filtering algorithm is used to filter each environment dimension time sequence, to filter out electronic noise and electromagnetic interference received by the environment sensor during data collection. A min-max normalization method is used to normalize each environment dimension time sequence to 0 to 1 after filtering, to unify the dimension. The sliding average filtering algorithm and the min-max normalization method are both known technologies, and the specific method is not described here.

[0055] Step S002: According to the difference between the data values of adjacent time points in each environment dimension time sequence and the distance between the position coordinates of adjacent time points, a plurality of key attention time points are selected from each environment dimension time sequence, and a significant attention factor of each key attention time point on each environment dimension is obtained.

[0056] It should be noted that during the automatic patrol, the first minute after the start of the intelligent vehicle is the preparation stage of the intelligent vehicle. The preparation stage does not perform environment anomaly judgment, and all data collected during the preparation stage are used as basic data for subsequent environment anomaly analysis.

[0057] Preferably, in an embodiment of the present application, the method for obtaining the significant attention factor of each key attention time point on each environment dimension comprises:

[0058] For the xth environment dimension time sequence, the absolute value of the difference between the data values of the tth time point and the (t-1)th time point is obtained, denoted as the fluctuation of the tth time point. The Euclidean distance between the position coordinates of the tth time point and the (t-1)th time point is obtained, denoted as the movement of the tth time point. The inverse proportional normalized value of the movement of the tth time point and the average value of the fluctuation of the tth time point are denoted as the attention degree of the tth time point on the xth environment dimension.

[0059] It is required to be explained: t is greater than 1, guaranteeing that the above analysis is established, and the attention degree of the first moment on each environmental dimension is the attention degree of the second moment on each environmental dimension. In this embodiment, the minimum maximum specification method is used to normalize the movement amount of all moments to between 0 and 1, and then the difference between 1 and the normalized value of the movement amount of the tth moment is taken as the inverse proportional normalized value of the movement amount of the tth moment. Wherein, the smaller the movement amount is, and the greater the fluctuation amount is, indicating that the intelligent vehicle moves a small distance in the patrol area, and then a large change occurs in this environmental dimension, so this environmental dimension needs to be focused on at this time.

[0060] The preset attention threshold is 0.7, and this is described as an example.

[0061] For any one environmental dimension, all moments with an attention degree greater than the preset attention threshold are recorded as key attention moments on the any one environmental dimension.

[0062] On the xth environmental dimension time sequence, the period formed by adjacent key attention moments at the moment is recorded as a key attention period.

[0063] Among them, a single key attention moment which is not a key attention moment at adjacent moments is also a key attention period.

[0064] It is required to be explained: when the environmental dimension appears a large fluctuation, it may be caused by the intelligent vehicle normally passing through different sub-scenes, for example: when entering and leaving the tunnel, the environmental difference between the tunnel and the outside of the tunnel is large, and at the transition of the tunnel entrance and exit, the environmental dimension will appear a large fluctuation. In addition, it may also be caused by the intelligent vehicle encountering environmental pollution, and the environmental pollution anomaly (such as waste gas leakage) is mostly a continuous emission source. When the intelligent vehicle approaches the pollution source, the spatial distribution of the pollutant concentration is not uniform due to the influence of factors such as emission intensity, air flow disturbance, and terrain shielding, and the sensor sampling value will appear high frequency and large fluctuation. For example: the intelligent vehicle moves to a radiation exceeding standard area, resulting in a sudden increase in radiation intensity, or the intelligent vehicle moves to a high-purity gas leakage area, such as nearby industrial pipeline leakage of inert gases such as nitrogen and argon, which dilutes the carbon dioxide in the air, resulting in a sudden drop in concentration. Therefore, it is further required to analyze whether each key attention period is a transition phase of the intelligent vehicle normally passing through different sub-scenes or a suspected encounter with environmental pollution.

[0065] In the yth key attention period on the xth environmental dimension time sequence, the difference between the previous data value and the next data value is obtained, if the difference is negative, the label -1 is given, if the difference is positive, the label 1 is given, and if the difference is zero, the label 0 is given. According to the time sequence, a -101 sequence formed by the labels corresponding to all adjacent data values is obtained.

[0066] In the -101 sequence, the number of all -1 and 0 is counted as a first number value, the number of all 0 and 1 is counted as a second number value, the maximum value of the first number value and the second number value is obtained, and the ratio of the length of the -101 sequence is recorded as the trend uniformity, the two elements corresponding to the maximum value of the first number value and the second number value are recorded as target elements, the number of elements between the adjacent two target elements (excluding target elements, if there is no other element between the adjacent two target elements, the number of elements is 0) is obtained, the sum value of the number of elements between all adjacent target elements is recorded as a first sum value, and the ratio of the first sum value and the length of the -101 sequence is recorded as the inverse proportional value of the trend uniformity . The average value of the trend uniformity and the same trend continuity is recorded as the trend continuity uniformity .

[0067] It should be noted that if the length of a certain focus period is less than 6, the nearest other time of the focus period is merged to obtain an updated focus period, and the length of the updated focus period is 6, for example, the focus period is {5th, 6th, 7th time}, and the new focus period is {3rd, 4th, 5th, 6th, 7th, 8th time}, and the length of the -101 sequence is at least 5, and the number of target elements is at least 3, and the trend continuity uniformity of the updated focus period is obtained in the above manner as the trend continuity uniformity of the focus period. If the maximum value of the first number value and the second number value is the first number value, -1 and 0 are target elements, if the maximum value of the first number value and the second number value is the second number value, 0 and 1 are target elements, if the first number value and the second number value are equal, the same trend continuity when -1 and 0 are target elements and the same trend continuity when 0 and 1 are target elements are obtained respectively, and the maximum value of the two is taken.

[0068] It should be further noted that when the intelligent vehicle normally passes through different sub-scenes, the environmental dimension data value gradually transitions from one scene to another, the environmental dimension data value gradually increases or decreases, and the data change trend is continuous and uniform. When the intelligent vehicle encounters suspected environmental pollution, due to the strong non-uniformity of the spatial distribution of pollution, the change trend of the environmental dimension data value is chaotic, and the change of the environmental dimension data value is first increasing and then decreasing, or first decreasing and then increasing (driving away from the pollution). Therefore, the smaller the trend continuity uniformity is, the more likely it is to encounter suspected environmental pollution, the more non-uniform the spatial distribution of pollution is, and the longer the length of the focus period is, the more it needs to be focused on to ensure the accuracy of subsequent environmental anomaly judgment.

[0069] On the xth environmental dimension time sequence, the normalized value of the length of the yth focus period and the inverse proportional value of the trend continuity uniformity the average of the environmental problem attention degrees of the key attention time periods, is recorded as the environmental problem attention degree of the yth key attention time period.

[0070] The length of each key attention time period is normalized to 0-1 using the min-max normalization method.

[0071] On the xth environmental dimension time sequence, the attention degree of the zth key attention time is recorded as the significant attention factor of the zth key attention time on the xth environmental dimension, where the attention degree of the zth key attention time and the average of the environmental problem attention degrees of the key attention time periods in which the zth key attention time is located.

[0072] It should be noted that for the xth environmental dimension, the greater the attention degree of the key attention time and the greater the environmental problem attention degree of the key attention time period in which the key attention time is located, the more important the key attention time is to the anomaly detection of the xth environmental dimension.

[0073] Step S003: On all environmental dimensions, the environmental anomaly detection weight of each time is obtained according to the number of all environmental dimensions for which each time is a key attention time and the significant attention factor; and the anomaly detection of each environmental dimension time sequence is weighted and predicted according to the environmental anomaly detection weights of all times, to obtain the abnormal point position on the global path.

[0074] It should be noted that there is a strong correlation between different environmental parameters, for example, an increase in temperature will cause a decrease in humidity, and an increase in carbon monoxide concentration is often accompanied by a simultaneous change in sulfur dioxide concentration. Therefore, it is further necessary to analyze the comprehensive performance of all environmental dimensions at the same time to determine the environmental anomaly attention degree of each time.

[0075] Preferably, in an embodiment of the present application, the method for obtaining the abnormal point position on the global path comprises:

[0076] On each environmental dimension time sequence, the sum of the environmental problem attention degrees of all key attention time periods is obtained, which is recorded as the abnormal performance of each environmental dimension.

[0077] The greater the abnormal performance, the greater the possibility that the environmental dimension is suspected to be abnormal in the current patrol process, and the more attention is needed for the environmental dimension in anomaly detection.

[0078] On all environmental dimensions, all environmental dimensions for which the bth time is a key attention time are obtained, which are recorded as the attention environmental dimensions.

[0079] The schematic diagram for screening the attention environmental dimensions at each time is shown in FIG. 2. Figure 2 Figure 2 ​The environment dimension concerned at each moment is marked as 1, and the environment dimension not concerned is marked as 0. For example, the 1st moment of temperature and sulfur dioxide concentration is the key moment concerned, so the environment dimension concerned at the 1st moment is temperature and sulfur dioxide concentration; the 2nd moment of temperature and sulfur dioxide concentration is the key moment concerned, so the environment dimension concerned at the 2nd moment is temperature and sulfur dioxide concentration; the 3rd moment of temperature, humidity, carbon monoxide concentration and sulfur dioxide concentration is the key moment concerned, so the environment dimension concerned at the 3rd moment is temperature, humidity, carbon monoxide concentration and sulfur dioxide concentration; the 4th moment of temperature, humidity, carbon dioxide concentration, carbon monoxide concentration and radiation intensity is the key moment concerned, so the environment dimension concerned at the 4th moment is temperature, humidity, carbon dioxide concentration, carbon monoxide concentration and radiation intensity; the 5th moment of temperature, carbon dioxide concentration, carbon monoxide concentration and radiation intensity is the key moment concerned, so the environment dimension concerned at the 5th moment is temperature, carbon dioxide concentration, carbon monoxide concentration and radiation intensity; the 5th moment of carbon dioxide concentration and radiation intensity is the key moment concerned, so the environment dimension concerned at the 5th moment is carbon dioxide concentration and radiation intensity.

[0080] For the bth moment, the sum value of the abnormal performance of all the environment dimensions concerned is obtained, denoted as a second sum value, the sum value of the abnormal performance of all the environment dimensions is obtained, denoted as a third sum value, the ratio of the second sum value to the third sum value is obtained, denoted as a first ratio, the average value of the significant attention factors on all the environment dimensions concerned at the bth moment is obtained, denoted as a comprehensive significant attention factor, the maximum value of the significant attention factors on all the environment dimensions concerned at the bth moment is obtained, denoted as a limit significant attention factor, the average value of the comprehensive significant attention factor and the limit significant attention factor is obtained, denoted as a first average value, and the average value of the first average value and the first ratio is obtained, denoted as an instantaneous important performance at the bth moment.

[0081] It should be noted that the more the types of environment dimensions concerned at the bth moment, the greater the abnormal performance of the environment dimensions concerned, and the greater the significant attention factor on the environment dimensions concerned at the bth moment, the more important the environment data at the bth moment.

[0082] It should be further noted that the outbreak of environmental abnormalities often has a gradual change process, and the instantaneous important performance may be in the “abnormal budding stage”, so it is necessary to adjust the instantaneous important performance by trend analysis to capture the development law of the abnormality in advance, thereby obtaining the comprehensive important performance at each moment.

[0083] The set of environment dimensions concerned at each moment is obtained.

[0084] Obtain the difference between the number of all concerned environmental dimensions at the bth moment and the number of all concerned environmental dimensions at the (b-1) th moment, denoted as the first difference value, obtain the intersection and union of the concerned environmental dimension sets at the (b-1) th moment and the bth moment, denoted as the target intersection and the target union, obtain the difference between the number of all elements in the target union and the number of all elements in the target intersection, denoted as the second difference value, obtain the ratio of the second difference value to the number of all elements in the target union, denoted as the concerned environmental dimension type transformation amount, obtain the difference between the number of all elements in the concerned environmental dimension set at the bth moment and the number of all elements in the target intersection, denoted as the third difference value, obtain the ratio of the third difference value to the number of all elements in the concerned environmental dimension set at the bth moment, denoted as the concerned environmental dimension type addition amount, obtain the average of the concerned environmental dimension type transformation amount and the concerned environmental dimension type addition amount, denoted as the concerned environmental dimension type change coefficient, obtain the average of the normalized value of the first difference value and the concerned environmental dimension type change coefficient, denoted as the important performance adjustment coefficient at the bth moment.

[0085] It should be noted that b is greater than 1, which ensures that the above analysis is valid, and the important performance adjustment coefficient at the first moment is the important performance adjustment coefficient at the second moment. The min-max normalization method is used to normalize all adjacent first difference values to between 0 and 1. The larger the first difference value, the more the number of environmental dimension types that need to be concerned increases, and the larger the transformation amount and the addition amount of the concerned environmental dimension type over time, the more likely it is that the instantaneous important performance at the bth moment is the budding moment of suspected environmental anomalies, and thus needs to be increased.

[0086] The preset constant is 1, which is used as an example for description.

[0087] The sum of the average of the important performance adjustment coefficient at the bth moment and the instantaneous important performance at the bth moment and the preset constant is denoted as the environmental anomaly detection weight at the bth moment.

[0088] Thus, the environmental anomaly detection weight at each moment is obtained. If there is no concerned environmental dimension at the bth moment, the environmental anomaly detection weight at the bth moment is set to the preset constant.

[0089] The preset anomaly threshold is 0.5, which is used as an example for description.

[0090] Obtain the sum of the environmental anomaly detection weights at all moments, denoted as the fourth sum value, obtain the ratio of the environmental anomaly detection weight at each moment to the fourth sum value, denoted as the prediction weight at each moment.

[0091] According to the prediction weight of all moments, a prediction value is obtained by using a weighted autoregressive moving average model to perform weighted prediction on the xth environmental dimension time sequence, and if the absolute value of the difference between the data value at the current moment (the last moment) in the xth environmental dimension time sequence and the prediction value is greater than a preset abnormal threshold, it is determined that the xth environmental dimension at the current moment is an abnormal environmental dimension.

[0092] It should be noted that the weighted autoregressive moving average model is a known technology, and the specific method is not introduced here. In this embodiment, by giving a larger prediction weight to the moment with suspected environmental anomaly, the fitting degree of the prediction value to the suspected environmental anomaly data change rule is improved during weighted prediction, so that the residual more accurately reflects the real anomaly, and finally reduces the false detection rate and the missed detection rate of anomaly detection, thereby completing real-time environmental anomaly detection.

[0093] In the above manner, it is determined whether each environmental dimension at the current moment is an abnormal environmental dimension.

[0094] If there is an abnormal environmental dimension among all environmental dimensions at the current moment, it is determined that the position coordinate at the current moment is an abnormal point position on the global path.

[0095] It should be noted that if the position coordinate at the current moment is an abnormal point position on the global path, a multimedia recording is triggered, a camera is controlled to take a photo and record a high-definition video of the abnormal area for a preset time length (such as 10 seconds), and then an early warning report is generated and uploaded, and the abnormal environmental dimension type, position coordinate, timestamp, and multimedia file are packaged and uploaded to the monitoring center through a wireless network.

[0096] Step S004: A circular temporary obstacle area is generated with the abnormal point position on the global path as the center, and a new local path is generated by starting from the abnormal point position and performing local path re-planning with the goal of bypassing the circular temporary obstacle area, guiding the intelligent vehicle to return to the global path after bypassing the circular temporary obstacle area, and completing the automatic patrol task.

[0097] The preset radius length M is 3 meters, and this is used as an example for description.

[0098] A circular temporary obstacle area with a radius of M is generated with the abnormal point position on the global path as the center, and a new local path is generated by starting from the abnormal point position and performing local path re-planning using the dynamic window method (DWA) with the goal of bypassing the circular temporary obstacle area, guiding the intelligent vehicle to return to the global path after bypassing the circular temporary obstacle area, and continuing to perform abnormal point position detection, and performing local path re-planning again after detecting a new abnormal point position, until the automatic patrol task is completed.

[0099] Among them, the dynamic window approach (DWA) is a well-known technology, and the specific method is not introduced here. It is a classic algorithm for real-time obstacle avoidance and path planning of mobile robots, which belongs to the category of local path planning. The core is to search for the optimal motion command within the kinematic constraints of the robot through a rolling window to achieve safe and efficient navigation in a dynamic environment.

[0100] It should be noted that after the completion of this automatic patrol task, all abnormal point positions and their related information (all abnormal environmental dimension types, position coordinates, and timestamps) constitute a key patrol point position list . In the next patrol task, the patrol point position list set by the user in this patrol task is merged with the key patrol point position list in the last patrol task to form a new patrol point position list. In the new patrol point position list, the weight of each patrol point position in is set to 1, and the weight of each abnormal point position in is set to , where is the number of all environmental dimensions, is the number of all abnormal environmental dimensions corresponding to each abnormal point position. Then, a priority Traveling Salesman Problem (TSP) model is used for weighted global path planning of the new patrol point position list, and the objective function is to find the shortest path to visit all points in the new patrol point position list, but at the same time, higher weights are given to the points in to ensure that the path generated by the algorithm will pass through these abnormal point positions preferentially and more "close" to them, achieving key patrol. Finally, a Genetic Algorithm (GA) is used to solve this optimization problem to obtain a new global key patrol path.

[0101] Among them, the priority Traveling Salesman Problem (TSP) model and the Genetic Algorithm (GA) are both well-known technologies, and the specific methods are not introduced here. TSP is a classic model for "shortest path to visit all points" in global path planning, and priority constraints can be achieved by modifying the objective function after introducing weights. GA is a commonly used intelligent optimization algorithm for solving TSP.

[0102] It should be further noted that after the intelligent vehicle completes the patrol of the new global key patrol path, the abnormal point positions in are detected again. If the environmental data of a certain abnormal point position returns to normal and stable, it is removed from the list , and subsequent patrols will no longer focus on it. The above operations are continuously executed to achieve continuous and adaptive intelligent patrol.

[0103] Thus, the present application is completed. ​

[0104] To sum up, in the embodiment of the present application, the global path of the automatic patrol task is planned on the patrol area grid map, a plurality of environment dimension time sequences and the position coordinates of each moment when the intelligent vehicle moves along the global path are obtained, a plurality of key attention moments are screened out from each environment dimension time sequence, the significant attention factors of each key attention moment on each environment dimension are obtained, the environment anomaly detection weight of each moment is obtained, each environment dimension time sequence is weighted and predicted for anomaly detection, the abnormal point position on the global path is obtained, the circular temporary obstacle area is determined, and the local path is re-planned to generate a new local path to guide the intelligent vehicle to return to the global path after bypassing the circular temporary obstacle area until the automatic patrol task is completed. The present application accurately identifies the point position with environmental anomaly, re-plans the local path, and improves the efficiency and safety of the automatic patrol of the intelligent vehicle.

[0105] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for automatic patrol intelligent vehicle path planning based on environmental sensors, characterized in that, The method includes the following steps: Plan the global path of the automatic patrol task on the grid map of the patrol area, and obtain the time sequence of several environmental dimensions when the intelligent vehicle moves along the global path, as well as the position coordinates at each moment. Based on the differences in data values ​​between adjacent moments in the time series of each environmental dimension, and the distance between the location coordinates of adjacent moments, several key moments of interest are selected from the time series of each environmental dimension, and the significant interest factors for each key moment of interest in each environmental dimension are obtained. Across all environmental dimensions, based on the number of all environmental dimensions at each time point of focus and the aforementioned significant attention factors, the environmental anomaly detection weight for each time point is obtained. Based on the environmental anomaly detection weights at all times, a weighted prediction anomaly detection is performed on the time series of each environmental dimension to obtain the anomaly points on the global path. A circular temporary obstacle area is generated centered on the abnormal point on the global path. Starting from the abnormal point, a local path replanning is performed with the goal of bypassing the circular temporary obstacle area to generate a new local path. The intelligent vehicle is guided to bypass the circular temporary obstacle area and then return to the global path until the automatic inspection task is completed.

2. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 1, characterized in that, The specific steps for selecting several key moments of interest from the time series of each environmental dimension are as follows: For the x-th environmental dimension time series, obtain the absolute value of the difference between the data values ​​at time t and time t-1, and denote it as the fluctuation at time t. Obtain the Euclidean distance between the position coordinates at time t and time t-1, and denote it as the movement at time t. Denote the inverse proportional normalized value of the movement at time t and the mean of the fluctuation at time t as the attention level at time t in the x-th environmental dimension. Based on the level of attention at each moment in each environmental dimension, several key moments of attention are obtained for each environmental dimension.

3. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 2, characterized in that, The specific steps for obtaining several key attention moments for each environmental dimension based on the level of attention at each moment in each environmental dimension are as follows: For any given environmental dimension, all moments when the attention level exceeds a preset attention threshold are recorded as key attention moments for that given environmental dimension.

4. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 2, characterized in that, The specific steps for obtaining the significant attention factors at each key attention moment in each environmental dimension are as follows: In the x-th environmental dimension time series, the time period consisting of adjacent key attention moments is denoted as the key attention period; Based on the differences between adjacent data values ​​within the y-th key attention period in the x-th environmental dimension time series, and the duration of the y-th key attention period, the environmental issue attention level for the y-th key attention period is obtained; Based on the attention given to each key moment in the time series of each environmental dimension and the attention given to environmental issues during the key period in which each key moment is located, significant attention factors for each key moment in each environmental dimension are obtained.

5. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 4, characterized in that, The specific steps involved in obtaining the environmental issue attention level for the y-th key attention period are as follows: In the y-th key attention period of the x-th environmental dimension time series, obtain the difference between the previous data value and the next data value. If the difference is negative, assign the label -1; if the difference is positive, assign the label 1; if the difference is zero, assign the label 0. In chronological order, obtain the -101 sequence composed of the labels corresponding to all adjacent data values. In the -101 sequence, count the number of all -1s and 0s, and denote it as the first quantity value. Count the number of all 0s and 1s, and denote it as the second quantity value. Take the ratio of the maximum value between the first and second quantity values ​​to the length of the -101 sequence, and denote it as the trend uniformity. Denote the two elements corresponding to the maximum value between the first and second quantity values ​​as the target elements. Denote the sum of the number of elements between all adjacent target elements as the first sum. Denote the inverse proportional value of the ratio of the first sum to the length of the -101 sequence as the same trend continuity. Denote the average of the trend uniformity and the same trend continuity as the trend continuity uniformity. The average of the normalized value of the duration of the y-th key focus period and the inverse proportional value of the trend continuity is denoted as the environmental issue attention level of the y-th key focus period.

6. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 4, characterized in that, The specific steps for obtaining the significant attention factor for each key attention moment in each environmental dimension based on the attention level of each key attention moment in the time series of each environmental dimension and the attention level of environmental issues in the key attention period in which each key attention moment is located are as follows: In the x-th environmental dimension time series, the average of the attention level at the z-th key attention moment and the attention level of environmental issues during the key attention period in which the z-th key attention moment is located is denoted as the significant attention factor at the z-th key attention moment in the x-th environmental dimension.

7. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 4, characterized in that, The specific steps for obtaining the environmental anomaly detection weights at each moment are as follows: For each environmental dimension time series, obtain the sum of the environmental issues attention levels for all key attention periods, and record it as the abnormal performance of each environmental dimension; Across all environmental dimensions, obtain all environmental dimensions that are of primary concern at time b, and denote them as the environmental dimensions of concern. Based on the abnormal performance of the environmental dimension at time b and the significant attention factors in the environmental dimension at time b, the instantaneous importance performance at time b is obtained. Obtain the set of all concerned environmental dimensions at each time point; Based on the difference in the number of all environmental dimensions of concern between time b and time b-1, and the difference in the set of environmental dimensions of concern between time b-1 and time b, the important performance adjustment coefficient at time b is obtained. The sum of the important performance adjustment coefficient at time b and the mean of the instantaneous important performance at time b and the preset constant is denoted as the environmental anomaly detection weight at time b.

8. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 7, characterized in that, The specific steps for obtaining the instantaneous significance at time b are as follows: For time b, obtain the sum of the anomalous performance of all environmental dimensions and denote it as the second sum. Obtain the sum of the anomalous performance of all environmental dimensions and denote it as the third sum. Denote the ratio of the second sum to the third sum as the first ratio. Obtain the mean of the significant attention factors in all environmental dimensions at time b and denote it as the comprehensive significant attention factor. Obtain the maximum value among the significant attention factors in all environmental dimensions at time b and denote it as the limiting significant attention factor. Denote the mean of the comprehensive significant attention factor and the limiting significant attention factor as the first mean. Denote the mean of the first mean and the first ratio as the instantaneous importance performance at time b.

9. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 7, characterized in that, The specific steps for obtaining the important performance adjustment coefficient at time b are as follows: The difference between the number of all environmental dimensions under attention at time b and the number of all environmental dimensions under attention at time b-1 is denoted as the first difference. The intersection and union of the sets of environmental dimensions under attention at time b-1 and time b are denoted as the target intersection and target union. The difference between the number of all elements in the target union and the number of all elements in the target intersection is denoted as the second difference. The ratio of the second difference to the number of all elements in the target union is denoted as the change amount of environmental dimension type. The difference between the number of all elements in the set of environmental dimensions under attention at time b and the number of all elements in the target intersection is denoted as the third difference. The ratio of the third difference to the number of all elements in the set of environmental dimensions under attention at time b is denoted as the increase amount of environmental dimension type. The average of the change amount of environmental dimension type and the increase amount of environmental dimension type is denoted as the change coefficient of environmental dimension type. The normalized value of the first difference and the average of the change coefficient of environmental dimension type are denoted as the important performance adjustment coefficient at time b.

10. The method for automatic patrol intelligent vehicle path planning based on environmental sensors according to claim 1, characterized in that, The specific steps for obtaining the anomaly locations on the global path are as follows: Obtain the sum of the environmental anomaly detection weights at all times, denoted as the fourth sum; obtain the ratio of the environmental anomaly detection weight to the fourth sum at each time, denoted as the prediction weight at each time. Based on the prediction weights of all times, a weighted autoregressive moving average model is used to perform weighted prediction on the time series of the xth environmental dimension to obtain a predicted value. If the absolute value of the difference between the data value at the current time and the predicted value in the time series of the xth environmental dimension is greater than a preset abnormal threshold, then the xth environmental dimension at the current time is determined to be an abnormal environmental dimension. If there is an abnormal environment dimension among all environment dimensions at the current moment, then the current position coordinates are determined to be an abnormal point on the global path.

Images