Key obstacle recognition method and device, unmanned vehicle and storage medium

Abstract

Embodiments of the present application disclose a key obstacle identification method and device, an unmanned vehicle and a storage medium. The method comprises detecting obstacles in the surrounding environment in real time during the driving of the unmanned vehicle, and obtaining description information of the obstacles; inputting the description information of the obstacles into a finite state machine model having preset invalid obstacle filtering rules to filter invalid obstacles, and obtaining candidate obstacles left after filtering; using a preset fuzzy logic algorithm to calculate an avoidance weight value corresponding to each candidate obstacle according to the lateral speed, longitudinal speed and longitudinal distance of the candidate obstacle; and obtaining key obstacles with an avoidance weight value greater than or equal to an avoidance threshold value from the candidate obstacles to effectively avoid the key obstacles. The technical solution of the embodiments of the present application provides a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions.

Description

Technical Field

[0001] The present invention relates to computer technology, and more particularly to autonomous driving technology, and more particularly to a method, apparatus, unmanned vehicle and storage medium for identifying key obstacles. Background Technology

[0002] With the development of vehicle intelligence technology, autonomous driving technology for driverless cars has gradually become a hot topic in the field of vehicle research. In real-world road environments, there are often stationary or low-speed obstacles blocking the path of driverless cars, affecting their driving efficiency. Therefore, driverless cars need to autonomously complete actions such as lateral obstacle avoidance and overtaking during operation.

[0003] Currently, effectively identifying key obstacles that vehicles truly need to avoid in complex and dynamically changing road conditions, and then making a reasonable obstacle avoidance and overtaking decision based on these key obstacles, has become a challenging problem in autonomous driving planning and decision-making algorithms. Summary of the Invention

[0004] This invention provides a method, device, autonomous vehicle, and storage medium for identifying key obstacles, thereby offering a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions.

[0005] In a first aspect, embodiments of the present invention provide a method for identifying key obstacles, the method comprising:

[0006] During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information about the obstacles is obtained.

[0007] The description information of the obstacle is input into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles, and the remaining candidate obstacles are obtained after filtering.

[0008] Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight values ​​corresponding to each candidate obstacle.

[0009] Among the candidate obstacles, the key obstacles with avoidance weight values ​​greater than or equal to the avoidance threshold value are selected for effective obstacle avoidance.

[0010] Secondly, embodiments of the present invention also provide a device for identifying key obstacles, the device comprising:

[0011] The obstacle description information acquisition module is used to detect obstacles in the surrounding environment in real time and acquire the description information of the obstacles during the operation of the autonomous vehicle.

[0012] The invalid obstacle filtering module is used to input the description information of the obstacle into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles and obtain the candidate obstacles left after filtering.

[0013] The avoidance weight value calculation module is used to calculate the avoidance weight value corresponding to each candidate obstacle based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle using a preset fuzzy logic algorithm.

[0014] The critical obstacle acquisition module is used to acquire critical obstacles from the candidate obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, so as to effectively avoid critical obstacles.

[0015] Thirdly, embodiments of the present invention also provide an unmanned vehicle, the unmanned vehicle comprising:

[0016] One or more processors;

[0017] Storage device for storing one or more programs;

[0018] When the one or more programs are executed by the one or more processors, the one or more processors implement the method for identifying key obstacles as described in any embodiment of the present invention.

[0019] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for identifying key obstacles as described in any embodiment of the present invention.

[0020] The technical solution of this invention involves real-time detection of obstacles in the surrounding environment during the operation of an autonomous vehicle, and acquisition of obstacle description information. This obstacle description information is then input into a finite state machine model with preset invalid obstacle filtering rules to filter out invalid obstacles, resulting in a list of candidate obstacles. Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight value corresponding to each candidate obstacle. From the candidate obstacles, key obstacles with avoidance weight values ​​greater than or equal to an avoidance threshold are identified for effective obstacle avoidance. This solves the problem that stationary or low-speed obstacles often obstruct the autonomous vehicle's path in real-world road environments, affecting its driving efficiency. It provides a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions. Attached Figure Description

[0021] Figure 1 This is a flowchart of a key obstacle identification method in an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of another key obstacle identification method in an embodiment of the present invention;

[0023] Figure 3 This is a flowchart of another key obstacle identification method in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of a key obstacle identification device according to an embodiment of the present invention;

[0025] Figure 5 This is a structural schematic diagram of an unmanned vehicle according to an embodiment of the present invention. Detailed Implementation

[0026] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0027] Figure 1 This is a flowchart of a method for identifying critical obstacles provided in an embodiment of the present invention. This embodiment is applicable to situations where critical obstacles that truly need to be avoided in front of a vehicle are identified in complex and dynamically changing road conditions. The method can be executed by a critical obstacle identification device and specifically includes the following steps:

[0028] S110. During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information of the obstacles is obtained.

[0029] The obstacle can be an object in front of the autonomous vehicle that affects or may affect its normal driving (such as a vehicle or pedestrian). The description information of the obstacle can refer to the description information of the object that affects or may affect the normal driving of the autonomous vehicle.

[0030] Optionally, the obstacle description information may include at least one of the following: speed information, location information, geographic location information, and obstacle type. Speed ​​information may refer to the relative speed of the obstacle with respect to the autonomous vehicle. Location information may refer to the relative position of the obstacle with respect to the autonomous vehicle. Geographic location information may refer to the absolute geographic location of the obstacle with respect to the Earth, and may include the obstacle's latitude and longitude. Obstacle type may refer to obstacles that cut into the autonomous vehicle's current lane from an adjacent lane, obstacles that cut into an adjacent lane from the autonomous vehicle's current lane, obstacles that cross the autonomous vehicle's current lane, and obstacles that merge into the same lane from an adjacent lane and the autonomous vehicle.

[0031] Specifically, during operation, autonomous vehicles can detect obstacles in the surrounding environment in real time through their own configuration and obtain relevant descriptive information about the obstacles, such as their speed, location, geographical location, and type.

[0032] In an optional embodiment of the present invention, during the driving of the autonomous vehicle, obstacles in the surrounding environment are detected in real time and descriptive information of the obstacles is obtained, including: during the driving of the autonomous vehicle, environmental perception data is collected in real time through a perception module configured on the vehicle; based on the environmental perception data, obstacles in the surrounding environment and descriptive information of the obstacles are identified.

[0033] The perception module can be a fusion module of various sensors (such as radar sensors, ultrasonic sensors, and infrared sensors) and can be used to collect data on the driving environment of autonomous vehicles in real time. Environmental perception data can include road data, data on static objects around the autonomous vehicle, and data on dynamic objects. Environmental perception data can serve as the basis for identifying obstacles.

[0034] During operation, autonomous vehicles can use their own sensor modules to collect real-time environmental perception data and filter the collected data to identify obstacles in the surrounding environment.

[0035] S120. Input the description information of the obstacle into the finite state machine model with the preset invalid obstacle filtering rules to filter invalid obstacles, and obtain the candidate obstacles left after filtering.

[0036] Invalid obstacles refer to obstacles that the autonomous vehicle clearly does not need to avoid, such as obstacles that are obviously traveling too fast or are outside the road network where the autonomous vehicle is traveling. Invalid obstacles can be filtered out by setting rules.

[0037] Invalid obstacle filtering rules can be rules that filter out obstacles that obviously do not need to be avoided from all obstacles identified by the autonomous vehicle. Optionally, invalid obstacle filtering rules can include at least one of the following: lateral movement intention obstacle filtering rules, off-network obstacle filtering rules, high-speed obstacle filtering rules, and long-distance obstacle filtering rules. Lateral movement intention obstacles can refer to obstacles that indicate an intention to change lanes in the normal direction of the autonomous vehicle's travel on the road, for example, vehicles with turn signals activated. Off-network obstacles can refer to obstacles outside the current travel path of the autonomous vehicle. High-speed obstacles can refer to obstacles traveling at high speed that do not affect the autonomous vehicle's travel; their classification as high-speed obstacles can be determined by judging whether their speed exceeds the speed limit of the current road. Long-distance obstacles can refer to obstacles that are far away from the autonomous vehicle and do not affect its travel; their classification as long-distance obstacles can be determined by setting a distance threshold. Obstacles that meet the invalid obstacle filtering rules can be identified as invalid obstacles, meaning the autonomous vehicle does not need to avoid invalid obstacles. The advantage of this setup is that, on the one hand, it can better handle the problem that the semantic analysis of obstacles is difficult to express with mathematical models, and on the other hand, it can directly filter out some obstacles that obviously do not need to be avoided, reducing the computational load of subsequent algorithms.

[0038] A finite state machine model can refer to a model that filters out obstacles (invalid obstacles) that reach the threshold by setting explicit rules and thresholds.

[0039] Alternate obstacles are those remaining after initially filtering out obstacles that clearly do not need to be avoided (invalid obstacles) from all obstacles identified by the autonomous vehicle. Further assessment is needed to determine whether obstacles in the alternative obstacles require avoidance.

[0040] Specifically, when an autonomous vehicle acquires description information of obstacles, it can input the description information of obstacles into a finite state machine model. The finite state machine model uses preset invalid obstacle filtering rules to filter out obstacles that obviously do not need to be avoided, thereby reducing the computational load of subsequent algorithms. The remaining obstacles after filtering are then identified as candidate obstacles.

[0041] In an optional embodiment of the present invention, inputting the description information of each obstacle into a finite state machine model with preset invalid obstacle filtering rules for filtering invalid obstacles may include at least one of the following:

[0042] The obstacle type is input into the finite state machine model. The obstacle type that falls into the lateral movement intention type set is filtered as an invalid obstacle by the lateral movement intention obstacle filtering rules in the finite state machine model.

[0043] The geographical location information of obstacles is input into the finite state machine model. The obstacle filtering rules outside the road network in the finite state machine model are used to filter out obstacles whose geographical location information does not fall into the road network as invalid obstacles.

[0044] The longitudinal velocity of obstacles is input into a finite state machine model. The high-speed obstacle filtering rules within the finite state machine model filter obstacles with longitudinal velocities greater than or equal to the road network limit as invalid obstacles.

[0045] The longitudinal distance of the obstacle is input into the finite state machine model. The long-distance obstacle filtering rules in the finite state machine model filter out obstacles whose longitudinal distance is greater than or equal to the distance threshold as invalid obstacles.

[0046] The obstacles in the lateral movement intent type set can include obstacles that cut into the lane where the vehicle is located from the adjacent lane, obstacles that cut into the adjacent lane from the lane where the vehicle is located, obstacles that cross the lane where the vehicle is located, and obstacles that merge into the same lane from the adjacent lane and the vehicle.

[0047] S130. Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight values ​​corresponding to the candidate obstacles.

[0048] In this context, lateral velocity can refer to the projected velocity of the obstacle in the normal direction of travel on the road network, longitudinal velocity can refer to the projected velocity of the obstacle in the tangential direction of travel on the road network, and longitudinal distance can refer to the projected distance between the obstacle and the current position of the autonomous vehicle on the road network. Fuzzy logic algorithm refers to an algorithm that obtains a precise decision result based on logic without clearly defined decision boundaries. The avoidance weight value can be the precise avoidance probability value for the current candidate obstacle obtained by using a fuzzy logic algorithm based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle. It should be noted that the avoidance conclusion obtained by the fuzzy logic algorithm can be a fuzzy conclusion value, such as ignoring, observing, or accepting. By setting weights for different avoidance conclusions, a weighted average of the different avoidance conclusions for the current candidate obstacle can be performed to obtain the avoidance weight value.

[0049] In this embodiment of the invention, the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle can be used as three attributes to further determine whether the candidate obstacle needs to be avoided. The avoidance probability value corresponding to each candidate obstacle can be calculated by a preset fuzzy logic algorithm.

[0050] S140. Among the candidate obstacles, identify the key obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, so as to effectively avoid the key obstacles.

[0051] The avoidance threshold can be the upper limit of the avoidance weight value calculated by a fuzzy logic algorithm. The key obstacle can be the obstacle that needs to be avoided from the candidate obstacles by using a preset fuzzy logic algorithm and the avoidance threshold.

[0052] Optionally, the avoidance weight value of the candidate obstacle is calculated. This avoidance weight value can be compared with an avoidance threshold value. Candidate obstacles with an avoidance weight value greater than or equal to the avoidance threshold value can be identified as critical obstacles and effectively avoided. Conversely, candidate obstacles with an avoidance weight value less than the avoidance threshold value can be identified as obstacles that do not need to be avoided.

[0053] The technical solution of this invention involves real-time detection of obstacles in the surrounding environment during the operation of an autonomous vehicle, and acquisition of obstacle description information. This obstacle description information is then input into a finite state machine model with preset invalid obstacle filtering rules to filter out invalid obstacles, resulting in a list of candidate obstacles. Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight value corresponding to each candidate obstacle. From the candidate obstacles, key obstacles with avoidance weight values ​​greater than or equal to an avoidance threshold are identified for effective obstacle avoidance. This solves the problem that stationary or low-speed obstacles often obstruct the autonomous vehicle's path in real-world road environments, affecting its driving efficiency. It provides a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions.

[0054] Figure 2 This is a flowchart of another method for identifying key obstacles provided by an embodiment of the present invention. Based on the above embodiments, this embodiment preferably further refines the operation of calculating avoidance weight values ​​corresponding to candidate obstacles using a preset fuzzy logic algorithm based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles. The method includes:

[0055] S210. During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information of the obstacles is obtained.

[0056] S220. Input the description information of the obstacle into the finite state machine model with the preset invalid obstacle filtering rules to filter invalid obstacles, and obtain the candidate obstacles left after filtering.

[0057] S230. Using the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle as attributes, and based on the preset fuzzy attribute values ​​corresponding to each attribute and the membership function corresponding to each fuzzy attribute value, calculate the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value.

[0058] Here, fuzzy attribute values ​​refer to the fuzzy values ​​of each attribute of each candidate obstacle. For example, lateral speed can include slow, medium, and fast; longitudinal speed can include slow, medium, and fast; and longitudinal distance can include near, medium, and far. The membership function can be a pre-designed function used to calculate the membership degree of each attribute of each candidate obstacle under each fuzzy attribute value. Membership degree refers to the degree to which each attribute of a candidate obstacle belongs to a certain fuzzy attribute value. It should be noted that lateral speed and longitudinal speed can be calculated using the same speed membership function. The speed membership function can be expressed as follows:

[0059]

[0060]

[0061]

[0062] Wherein, mf(v obj ) is the fast membership function, mm(v) obj ) is the medium-speed membership function, ms(v) obj ) represents the low-speed membership function. obj V represents the longitudinal / lateral velocity of the obstacle, where vf is the set threshold for the obstacle's high speed. s v is the set slow speed threshold for the obstacle. m For v f With v s The mean.

[0063] The distance membership function can be expressed as follows:

[0064]

[0065]

[0066]

[0067] Wherein, mf(d obj ) represents the long-distance membership function, mm(d) obj) is the mid-range membership function, ms(d obj ) represents the proximity membership function. d obj d represents the longitudinal velocity / lateral velocity of the obstacle. f d is the set threshold for the distance of obstacles. s d is the set obstacle proximity threshold. m For d f With d s The mean.

[0068] In this embodiment of the invention, the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle can be used as three attributes to further determine whether the candidate obstacle needs to be avoided. Based on preset fuzzy attribute values ​​(e.g., slow, medium, and fast, or near, medium, and far) corresponding to each attribute (e.g., lateral velocity, longitudinal velocity, and longitudinal distance), and membership functions (e.g., velocity membership function and distance membership function) corresponding to each fuzzy attribute value, the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value can be calculated. That is, a total of nine membership degrees can be calculated for the three attributes of each candidate obstacle (lateral velocity "slow", lateral velocity "medium", lateral velocity "fast", longitudinal velocity "slow", longitudinal velocity "medium", longitudinal velocity "fast", longitudinal distance "near", longitudinal distance "medium", and longitudinal distance "far"), which can be represented in numerical form.

[0069] S240. Match the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value with the preset fuzzy rule set to obtain the conclusion value of the fuzzy conclusion of each target hit by the candidate obstacle.

[0070] The fuzzy rule set can be a pre-defined table used to query fuzzy conclusions for avoiding candidate obstacles based on calculated membership degrees. The fuzzy rule set can be as shown in the table below. A fuzzy conclusion can be an ambiguous conclusion about avoiding a candidate obstacle, such as ignoring, observing, or accepting. The target fuzzy conclusion can be a conclusion determined from all fuzzy conclusions. For example, if a candidate obstacle has a lateral velocity of "fast," a longitudinal velocity of "fast," and a longitudinal distance of "far," the target fuzzy conclusion queried from the fuzzy rule set would be "ignore." The conclusion value can be the specific numerical value of the fuzzy conclusion determined based on membership degrees.

[0071]

[0072]

[0073] Specifically, based on the calculated membership degrees of the three attributes, the intersection rule (minimum value rule) can be used to obtain the numerical value corresponding to the fuzzy conclusion. That is, the minimum value of the membership degrees of the three attributes can be taken as the numerical value of the target fuzzy conclusion for the candidate obstacle. For example, for a candidate obstacle, the membership degree of "fast" for lateral speed is calculated to be 0.2, the membership degree of "fast" for longitudinal speed is 0.3, and the membership degree of "far" for longitudinal distance is 0.1. Using the fuzzy rule set, the target fuzzy conclusion for the current candidate obstacle is determined to be "ignore," and the value of "ignore" can be determined to be 0.1 according to the intersection rule (minimum value rule).

[0074] S250. Based on the conclusion weights corresponding to each type of fuzzy conclusion, perform a weighted average of the conclusion values ​​of each target fuzzy conclusion corresponding to the candidate obstacles to obtain the avoidance weight values ​​corresponding to the candidate obstacles.

[0075] Among them, the conclusion weight can refer to the proportion of importance of different fuzzy conclusions made when avoiding alternative obstacles.

[0076] Optionally, weight values ​​can be pre-set for each type of fuzzy conclusion. Then, a weighted average can be calculated based on these pre-set weight values ​​to determine the avoidance weight value for each candidate obstacle. For example, the weight for "ignore" can be set to -1.0, the weight for "observe" to 0.2, and the weight for "accept" to 1.0. If the fuzzy conclusion for a candidate obstacle is determined to be 0.1 for "ignore," 0.2 for "observe," and 0.3 for "accept," then its avoidance weight value can be determined as 0.1*(-1.0) + 0.2*(0.2) + 0.3*(1.0) = 0.24 based on the pre-set fuzzy conclusion weight values.

[0077] S260. Among the candidate obstacles, identify the key obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, so as to effectively avoid the key obstacles.

[0078] The technical solution of this invention involves obtaining descriptive information of obstacles by real-time detection of the surrounding environment during the operation of an autonomous vehicle; filtering candidate obstacles using invalid obstacle filtering rules; calculating the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value based on its lateral velocity, longitudinal velocity, longitudinal distance attributes, and membership function; matching the membership degree with a fuzzy rule set to obtain the conclusion value of the target fuzzy conclusion corresponding to the candidate obstacle; performing weighted averaging on the conclusion value to obtain the avoidance weight value of the candidate obstacle; and identifying candidate obstacles with avoidance weight values ​​greater than or equal to the avoidance threshold as key obstacles, thereby enabling effective obstacle avoidance. This solves the problem that stationary or low-speed obstacles often block the path of autonomous vehicles in real-world road environments, affecting their driving efficiency, and provides a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions.

[0079] Figure 3 A flowchart of another key obstacle identification method provided by an embodiment of the present invention. Based on the above embodiments, this embodiment preferably further adds operations, the method including:

[0080] S310. During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information of the obstacles is obtained.

[0081] S320. Input the description information of the obstacle into the finite state machine model with the preset invalid obstacle filtering rules to filter invalid obstacles, and obtain the candidate obstacles left after filtering.

[0082] S330. Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight values ​​corresponding to the candidate obstacles.

[0083] S340. Collect obstacle avoidance decisions from multiple vehicle drivers for multiple reference obstacles, the obstacle avoidance decisions including avoiding obstacles or not avoiding obstacles.

[0084] Obstacle avoidance decision-making refers to the decision made by the driver of a vehicle whether to avoid an obstacle.

[0085] Specifically, multiple human drivers' obstacle avoidance decisions for various reference obstacles can be collected in advance, such as avoiding or not avoiding the obstacle, and the obstacle avoidance decisions can be labeled. For example, the results of human drivers' decisions on whether to avoid a reference obstacle can be represented by AD, where AD = 1 indicates that the reference obstacle needs to be avoided in human driving behavior, and AD = 0 indicates that the reference obstacle does not need to be avoided in human driving behavior.

[0086] S350. Based on the obstacle avoidance decision corresponding to each of the reference obstacles and the avoidance weight value corresponding to each reference obstacle, the avoidance threshold value is obtained by comparison.

[0087] Optionally, an avoidance threshold value can be obtained by comparing the vehicle driver's obstacle avoidance decisions for each reference obstacle with the calculated avoidance weight value corresponding to each reference obstacle.

[0088] In an optional embodiment of the present invention, the avoidance threshold value is obtained by comparing the obstacle avoidance decision corresponding to each of the reference obstacles and the avoidance weight value corresponding to each reference obstacle, and may include: according to the formula: argmax α∈[0，1] f(α) = γ1F(output > α | AD = 1) + γ2F(output < α | AD = 0), and the avoidance threshold value α is obtained by comparison; where F(.) is the probability distribution function, AD = 1 means the vehicle driver's obstacle avoidance decision is to avoid the obstacle, AD = 0 means the vehicle driver's obstacle avoidance decision is not to avoid the obstacle, output is the avoidance weight value corresponding to each reference obstacle, and γ1 and γ2 can be 0.5 respectively.

[0089] The advantage of this setup is that by collecting data on human driving behavior, we can determine the obstacle avoidance threshold and ensure that the obstacle avoidance performance of the autonomous vehicle is more in line with human driving behavior.

[0090] S360. Among the candidate obstacles, identify the key obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, so as to effectively avoid the key obstacles.

[0091] The technical solution of this invention involves real-time detection of obstacles in the surrounding environment and acquisition of obstacle description information during the operation of an autonomous vehicle; filtering out candidate obstacles using invalid obstacle filtering rules; calculating avoidance weight values ​​corresponding to each candidate obstacle using a preset fuzzy logic algorithm based on the lateral speed, longitudinal speed, and longitudinal distance of the candidate obstacles; collecting obstacle avoidance decisions from multiple drivers for multiple reference obstacles; comparing the avoidance decisions and avoidance weight values ​​corresponding to each reference obstacle to obtain an avoidance threshold value; and identifying key obstacles among the candidate obstacles whose avoidance weight values ​​are greater than or equal to the avoidance threshold value for effective obstacle avoidance. This solves the problem that stationary or low-speed obstacles often obstruct the path of autonomous vehicles in real-world road environments, affecting their driving efficiency, and provides a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions.

[0092] In an optional embodiment of the present invention, after identifying key obstacles among candidate obstacles whose avoidance weight values ​​are greater than or equal to avoidance threshold values, the method may further include: inputting the description information of the key obstacle into the vehicle's decision module, and determining an obstacle avoidance strategy matching the key obstacle through the decision module; controlling the vehicle to execute the obstacle avoidance strategy to effectively avoid the key obstacle.

[0093] The decision-making module refers to the module that determines whether the autonomous vehicle should avoid a critical obstacle. The obstacle avoidance strategy refers to an appropriate obstacle avoidance method for a critical obstacle, such as lateral obstacle avoidance or overtaking.

[0094] After selecting key obstacles from the candidate obstacles, the description information of the key obstacles can be input into the decision module of the autonomous vehicle. The decision module determines the obstacle avoidance strategy that matches the key obstacle, thereby controlling the autonomous vehicle to execute the obstacle avoidance strategy and achieving effective obstacle avoidance of the key obstacle.

[0095] Figure 4 This is a schematic diagram of the structure of a key obstacle identification device provided in an embodiment of the present invention. The device may include: an obstacle description information acquisition module 410, an invalid obstacle filtering module 420, an invalid obstacle filtering module 430, and a key obstacle acquisition module 440. Wherein:

[0096] The obstacle description information acquisition module 410 is used to detect obstacles in the surrounding environment in real time and acquire the description information of the obstacles during the driving of the unmanned vehicle.

[0097] The invalid obstacle filtering module 420 is used to input the description information of the obstacle into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles, and to obtain the candidate obstacles left after filtering.

[0098] The invalid obstacle filtering module 430 is used to calculate the avoidance weight value corresponding to each candidate obstacle based on the lateral velocity, longitudinal velocity and longitudinal distance of the candidate obstacle using a preset fuzzy logic algorithm.

[0099] The critical obstacle acquisition module 440 is used to acquire critical obstacles from the candidate obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, so as to effectively avoid the critical obstacles.

[0100] The technical solution of this invention involves real-time detection of obstacles in the surrounding environment during the operation of an autonomous vehicle, and acquisition of obstacle description information. This obstacle description information is then input into a finite state machine model with preset invalid obstacle filtering rules to filter out invalid obstacles, resulting in a list of candidate obstacles. Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight value corresponding to each candidate obstacle. From the candidate obstacles, key obstacles with avoidance weight values ​​greater than or equal to an avoidance threshold are identified for effective obstacle avoidance. This solves the problem that stationary or low-speed obstacles often obstruct the autonomous vehicle's path in real-world road environments, affecting its driving efficiency. It provides a new technology for effectively identifying key obstacles in complex and dynamically changing road conditions.

[0101] Optionally, in the above-mentioned device, the obstacle description information acquisition module 410 can be used to: collect environmental perception data in real time through the perception module configured on the vehicle during the driving of the unmanned vehicle; identify obstacles in the surrounding environment and their description information based on the environmental perception data; wherein the obstacle description information includes at least one of the following: speed information, location information, geographical location information, and obstacle type.

[0102] Optionally, the invalid obstacle filtering rule in the above-mentioned device may include at least one of the following: lateral movement intention obstacle filtering rule, road network external obstacle filtering rule, high-speed obstacle filtering rule, and long-distance obstacle filtering rule.

[0103] Optionally, the invalid obstacle filtering module 420 in the above-described device may include at least one of the following units: a lateral movement intention obstacle filtering unit, used to input the obstacle type of an obstacle into a finite state machine model, and filter obstacles whose obstacle type falls into the lateral movement intention type set as invalid obstacles according to the lateral movement intention obstacle filtering rules in the finite state machine model; an off-network obstacle filtering unit, used to input the geographical location information of an obstacle into a finite state machine model, and filter obstacles whose geographical location information does not fall into the road network as invalid obstacles according to the off-network obstacle filtering rules in the finite state machine model; a high-speed obstacle filtering unit, used to input the longitudinal speed of an obstacle into a finite state machine model, and filter obstacles whose longitudinal speed is greater than or equal to the road network limit value as invalid obstacles according to the high-speed obstacle filtering rules in the finite state machine model; and a long-distance obstacle filtering unit, used to input the longitudinal distance of an obstacle into a finite state machine model, and filter obstacles whose longitudinal distance is greater than or equal to a distance threshold as invalid obstacles according to the long-distance obstacle filtering rules in the finite state machine model.

[0104] Optionally, the avoidance weight value calculation module 430 in the above device may include: a membership degree calculation unit, used to take the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle as attributes, and calculate the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value according to preset fuzzy attribute values ​​corresponding to each attribute and membership degree functions corresponding to each fuzzy attribute value of each attribute; a conclusion value acquisition unit, used to match the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value with a preset fuzzy rule set to obtain the conclusion value of each target fuzzy conclusion hit by the candidate obstacle; and an avoidance weight value calculation subunit, used to perform weighted averaging processing on the conclusion values ​​of each target fuzzy conclusion corresponding to the candidate obstacle according to the conclusion weights corresponding to each type of fuzzy conclusion to obtain the avoidance weight value corresponding to the candidate obstacle.

[0105] Optionally, the above-described device may further include an obstacle avoidance threshold acquisition module, which can be used to identify key obstacles among candidate obstacles whose avoidance weight values ​​are greater than or equal to the avoidance threshold value. This module may include: an obstacle avoidance decision collection unit, used to collect obstacle avoidance decisions from multiple vehicle drivers for multiple reference obstacles, the obstacle avoidance decisions including avoiding obstacles or not avoiding obstacles; and an obstacle avoidance threshold acquisition subunit, used to compare the avoidance threshold value with the obstacle avoidance decisions corresponding to each of the reference obstacles and the avoidance weight values ​​corresponding to each reference obstacle.

[0106] Optionally, in the above device, the avoidance threshold acquisition subunit can be used to obtain the value according to the formula: argmax α∈[0，1] f(α) = γ1F(output > α | AD = 1) + γ2F(output < α | AD = 0), and the avoidance threshold value α is obtained by comparison; where F(.) is the probability distribution function, AD = 1 means the vehicle driver's obstacle avoidance decision is to avoid the obstacle, AD = 0 means the vehicle driver's obstacle avoidance decision is not to avoid the obstacle, and output is the avoidance weight value corresponding to each reference obstacle.

[0107] Optionally, the above-mentioned device may also include an obstacle avoidance strategy execution module, which can be used to identify key obstacles among candidate obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, input the description information of the key obstacles into the vehicle's decision module, and determine an obstacle avoidance strategy matching the key obstacles through the decision module; control the vehicle to execute the obstacle avoidance strategy to effectively avoid the key obstacles.

[0108] The key obstacle identification device provided in the embodiments of the present invention can execute the key obstacle identification method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0109] Figure 5 This is a schematic diagram of the structure of an unmanned vehicle provided in an embodiment of the present invention, such as... Figure 5 As shown, the autonomous vehicle includes a processor 510, a storage device 520, an input device 530, and an output device 540; the number of processors 510 in the autonomous vehicle can be one or more. Figure 5 Taking a processor 510 as an example; the processor 510, storage device 520, input device 530, and output device 540 in an autonomous vehicle can be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0110] Storage device 520, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as program instructions / modules corresponding to the key obstacle identification method in this embodiment of the invention (e.g., obstacle description information acquisition module 410, invalid obstacle filtering module 420, avoidance weight value calculation module 430, and key obstacle acquisition module 440 in the key obstacle identification device). Processor 510 executes various functional applications and data processing of the autonomous vehicle by running the software programs, instructions, and modules stored in storage device 520, thereby implementing the aforementioned key obstacle identification method, which may include:

[0111] During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information about the obstacles is obtained.

[0112] The description information of the obstacle is input into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles, and the remaining candidate obstacles are obtained after filtering.

[0113] Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight values ​​corresponding to each candidate obstacle.

[0114] Among the candidate obstacles, the key obstacles with avoidance weight values ​​greater than or equal to the avoidance threshold value are selected for effective obstacle avoidance.

[0115] Storage device 520 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, storage device 520 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, storage device 520 may further include memory remotely located relative to processor 510, which can be connected to the device / terminal / server via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0116] Input device 530 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the autonomous vehicle. Output device 540 may include display devices such as a display screen.

[0117] This invention also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program package is used to perform a method for identifying a key obstacle. The method may include:

[0118] During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information about the obstacles is obtained.

[0119] The description information of the obstacle is input into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles, and the remaining candidate obstacles are obtained after filtering.

[0120] Based on the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacles, a preset fuzzy logic algorithm is used to calculate the avoidance weight values ​​corresponding to each candidate obstacle.

[0121] Among the candidate obstacles, the key obstacles with avoidance weight values ​​greater than or equal to the avoidance threshold value are selected for effective obstacle avoidance.

[0122] Of course, the computer program of the computer-readable storage medium provided in the embodiments of the present invention is not limited to the method operation described above, but can also perform related operations in the key obstacle identification method provided in any embodiment of the present invention.

[0123] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0124] It is worth noting that in the embodiments of the search device described above, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of the present invention.

[0125] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method of critical obstacle identification, the method comprising: include: During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information about the obstacles is obtained. The description information of the obstacle is input into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles, and the remaining candidate obstacles are obtained after filtering. Using the lateral velocity, longitudinal velocity, and longitudinal distance of candidate obstacles as attributes, and based on preset fuzzy attribute values ​​corresponding to each attribute and membership functions corresponding to each fuzzy attribute value, the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value is calculated. The membership degrees of each attribute of the candidate obstacle under each fuzzy attribute value are matched with a preset fuzzy rule set to obtain the conclusion values ​​of the fuzzy conclusions for each target hit by the candidate obstacle. Based on the conclusion weights corresponding to each type of fuzzy conclusion, the conclusion values ​​of the fuzzy conclusions for each target corresponding to the candidate obstacle are weighted and averaged to obtain the avoidance weight value corresponding to each candidate obstacle. The avoidance weight value refers to the accurate avoidance probability value for the current candidate obstacle. Wherein, the membership degree refers to the degree to which each attribute of the candidate obstacle belongs to the fuzzy attribute value; the membership degrees of the lateral velocity and the longitudinal velocity under the fuzzy attribute values ​​are calculated using the same velocity membership function, which is expressed as follows: ,in, For fast membership functions, For medium-speed membership functions, For low-speed membership functions, The longitudinal and lateral velocities of the obstacle. The set speed threshold for obstacles, The set slow speed threshold for obstacles, for and The mean of the longitudinal distance; the membership degree of the longitudinal distance under the fuzzy attribute value is calculated by the distance membership function, which is expressed as follows: ,in, For long-distance membership functions, For the mid-range membership function, For proximity membership function, The longitudinal velocity / lateral velocity of the obstacle. The set threshold for the distance to obstacles. The threshold for proximity to obstacles is set. for and The mean; Collect obstacle avoidance decisions from multiple vehicle drivers for multiple reference obstacles, the obstacle avoidance decisions including avoiding obstacles or not avoiding obstacles; obtain an avoidance threshold value by comparing the obstacle avoidance decisions corresponding to each of the reference obstacles and the avoidance weight value corresponding to each reference obstacle. Among the candidate obstacles, key obstacles with avoidance weight values ​​greater than or equal to the avoidance threshold value are selected. The description information of the key obstacles is input into the vehicle's decision module, and the decision module determines an obstacle avoidance strategy that matches the key obstacle. The vehicle is controlled to execute the obstacle avoidance strategy to effectively avoid the key obstacle. The avoidance threshold value is determined by collecting human driving behavior data. The avoidance threshold value is obtained by comparing the obstacle avoidance decision corresponding to each of the reference obstacles and the avoidance weight value corresponding to each reference obstacle, including: according to the formula: The avoidance threshold value is obtained by comparison. Where F(.) is the probability distribution function, AD=1 indicates that the vehicle driver's obstacle avoidance decision is to avoid the obstacle, AD=0 indicates that the vehicle driver's obstacle avoidance decision is not to avoid the obstacle, and output is the avoidance weight value corresponding to each reference obstacle.

2. The method according to claim 1, characterized in that, During the operation of the autonomous vehicle, obstacles in the surrounding environment are detected in real time, and descriptive information about the obstacles is obtained, including: During the operation of the driverless vehicle, environmental perception data is collected in real time through the perception module configured on the vehicle. Based on the environmental perception data, obstacles in the surrounding environment and their descriptive information are identified. The description information of the obstacle includes at least one of the following: speed information, location information, geographical location information, and obstacle type.

3. The method according to claim 2, characterized in that, include: The invalid obstacle filtering rule includes at least one of the following: Lateral movement intention obstacle filtering rules, road network external obstacle filtering rules, high-speed obstacle filtering rules, and long-distance obstacle filtering rules.

4. The method according to claim 3, characterized in that, The description information of each obstacle is input into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles, including at least one of the following: The obstacle type is input into the finite state machine model. The obstacle type that falls into the lateral movement intention type set is filtered as an invalid obstacle by the lateral movement intention obstacle filtering rules in the finite state machine model. The geographical location information of obstacles is input into the finite state machine model. The obstacle filtering rules outside the road network in the finite state machine model are used to filter out obstacles whose geographical location information does not fall into the road network as invalid obstacles. The longitudinal velocity of the obstacle is input into the finite state machine model. The high-speed obstacle filtering rules in the finite state machine model filter out obstacles whose longitudinal velocity is greater than or equal to the road network limit value as invalid obstacles. as well as The longitudinal distance of the obstacle is input into the finite state machine model. The long-distance obstacle filtering rules in the finite state machine model filter out obstacles whose longitudinal distance is greater than or equal to the distance threshold as invalid obstacles.

5. A device for identifying key obstacles, characterized in that, include: The obstacle description information acquisition module is used to detect obstacles in the surrounding environment in real time and acquire the description information of the obstacles during the operation of the autonomous vehicle. The invalid obstacle filtering module is used to input the description information of the obstacle into a finite state machine model with preset invalid obstacle filtering rules to filter invalid obstacles and obtain the candidate obstacles left after filtering. The avoidance weight value calculation module is used to take the lateral velocity, longitudinal velocity, and longitudinal distance of the candidate obstacle as attributes, and calculate the membership degree of each attribute of the candidate obstacle under each fuzzy attribute value according to preset fuzzy attribute values ​​and membership functions corresponding to each fuzzy attribute value. The module then matches the membership degrees of each attribute of the candidate obstacle under each fuzzy attribute value with a preset fuzzy rule set to obtain the conclusion values ​​of the fuzzy conclusions for each target hit by the candidate obstacle. Based on the conclusion weights corresponding to each type of fuzzy conclusion, the module performs a weighted average of the conclusion values ​​of each target fuzzy conclusion corresponding to the candidate obstacle to obtain the avoidance weight value corresponding to each candidate obstacle. The avoidance weight value refers to the precise avoidance probability value for the current candidate obstacle. Wherein, the membership degree refers to the degree to which each attribute of the candidate obstacle belongs to the fuzzy attribute value; the membership degrees of the lateral velocity and the longitudinal velocity under the fuzzy attribute values ​​are calculated using the same velocity membership function, which is expressed as follows: ,in, For fast membership functions, For medium-speed membership functions, For low-speed membership functions, The longitudinal velocity / lateral velocity of the obstacle. The set speed threshold for obstacles, The set slow speed threshold for obstacles, for and The mean of the longitudinal distance; the membership degree of the longitudinal distance under the fuzzy attribute value is calculated by the distance membership function, which is expressed as follows: ,in, For long-distance membership functions, For the mid-range membership function, For proximity membership function, The longitudinal velocity / lateral velocity of the obstacle. The set threshold for the distance to obstacles. The threshold for proximity to obstacles is set. for and The mean; The obstacle avoidance decision collection unit is used to collect obstacle avoidance decisions made by multiple vehicle drivers for multiple reference obstacles, the obstacle avoidance decisions including avoiding obstacles or not avoiding obstacles; the avoidance threshold value acquisition subunit is used to obtain the avoidance threshold value by comparing the obstacle avoidance decision corresponding to each of the reference obstacles and the avoidance weight value corresponding to each reference obstacle. A key obstacle acquisition module is used to acquire key obstacles from candidate obstacles whose avoidance weight value is greater than or equal to the avoidance threshold value, input the description information of the key obstacles into the vehicle's decision module, and determine an obstacle avoidance strategy matching the key obstacles through the decision module; control the vehicle to execute the obstacle avoidance strategy to effectively avoid the key obstacles, wherein the avoidance threshold value is determined by collecting human driving behavior data; The threshold avoidance sub-unit is specifically used to obtain the threshold value according to the formula: The avoidance threshold value is obtained by comparison. Where F(.) is the probability distribution function, AD=1 indicates that the vehicle driver's obstacle avoidance decision is to avoid the obstacle, AD=0 indicates that the vehicle driver's obstacle avoidance decision is not to avoid the obstacle, and output is the avoidance weight value corresponding to each reference obstacle.

6. An unmanned vehicle, characterized in that, The unmanned vehicles include: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for identifying critical obstacles as described in any one of claims 1-4.

7. The driverless vehicle according to claim 6, characterized in that, The driverless vehicles also include: The perception module is used to collect environmental perception data in real time, and to identify obstacles in the environment around the autonomous vehicle and their description information based on the collected environmental perception data.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the method for identifying critical obstacles as described in any one of claims 1-4.

Images