A distributed driving decision-making generation method, system and storage medium

Abstract

A driving decision generation method, system and storage medium based on distributed management includes collecting multi-source environmental data around the vehicle; pre-processing the multi-source environmental data and performing data fusion to obtain environmental perception data; identifying the current driving situation based on environmental perception data through a deep learning algorithm; classifying the current driving situation according to a preset classification method to determine the type of driving situation the vehicle is in; each intelligent agent in the multi-intelligent system shares environmental perception data and driving situation type through a distributed protocol to generate local decisions corresponding to each intelligent agent; and fusion of the local decisions corresponding to each intelligent agent to form a global driving decision. The driving strategy and decision priority are flexibly adjusted according to the situation classification results, which improves the flexibility of intelligent driving, integrates the local decisions of each intelligent agent, and ensures the advancement and safety of the intelligent driving method.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to a distributed driving decision-making generation method, system and storage medium. Background Art

[0002] Existing intelligent driving systems mainly rely on a highly integrated single-agent architecture, which integrates multiple sensor data such as radar, LiDAR, camera, and global positioning system (GPS), combined with advanced machine vision and artificial intelligence algorithms, to achieve comprehensive perception and decision-making of the vehicle's surrounding environment. However, although this architecture has demonstrated high autonomous driving capabilities in specific scenarios, its limitations are gradually revealed when faced with complex and changing real driving environments.

[0003] First, the single-agent architecture faces challenges in allocating computing resources and processing power when handling multiple tasks, such as performing environmental perception, decision-making, and vehicle control simultaneously. This "one-shoulder" model makes it difficult for the system to ensure the comprehensiveness of perception and the accuracy of decision-making in complex or unexpected situations, thereby increasing the risk of traffic accidents. Especially in high-risk scenarios such as emergency obstacle avoidance on highways and pedestrian detection and avoidance in urban road conditions, the reaction speed and decision-making accuracy of a single agent become key factors restricting its performance.

[0004] Secondly, different driving scenarios have significantly different requirements for intelligent driving systems. From urban roads to highways, from sunny days to bad weather, each scenario has its own unique challenges. It is difficult for a single intelligent agent to be deeply optimized for each specific scenario, resulting in unsatisfactory performance in specific environments. For example, the frequent appearance of pedestrians, non-motor vehicles and changing traffic signals in urban environments place higher demands on the system's real-time response and decision-making capabilities; on highways, more emphasis is placed on vehicle distance control, lane keeping and stability at high speeds.

[0005] Furthermore, existing intelligent driving systems generally lack sufficient flexibility and scalability. When the driving environment changes significantly or the system needs to cope with new driving scenarios, the adjustment ability of a single intelligent agent is limited, making it difficult to quickly adapt to the new environment and optimize its perception and decision-making strategies. This limitation restricts the breadth and reliability of intelligent driving systems in practical applications and hinders the further promotion of unmanned driving technology.

[0006] In summary, the existing intelligent driving methods have technical problems of poor advancement and flexibility. Summary of the invention

[0007] The main purpose of the present invention is to provide a distributed driving decision-making method, system and storage medium, aiming to solve the technical problem that the existing distributed driving decision-making method has poor user experience.

[0008] To achieve the above-mentioned purpose, the present invention provides a driving decision generation method based on distribution, which includes the following steps: collecting multi-source environmental data around a vehicle, the multi-source environmental data including at least two types of environmental image data, radar data, satellite data, and IMU data; preprocessing the multi-source environmental data, and fusing the preprocessed multi-source environmental data to obtain environmental perception data; identifying the current driving situation based on the environmental perception data through a deep learning algorithm; classifying the current driving situation according to a preset classification method to determine the type of driving situation in which the vehicle is located, the driving situation type being urban roads, highways, complex intersections, or bad weather; sharing environmental perception data and driving situation types between the intelligent agents in the multi-intelligent system through a distributed protocol to generate local decisions corresponding to each intelligent agent; fusing the local decisions corresponding to each intelligent agent to form a global driving decision, the global driving decision including three items: path planning, speed control, and obstacle avoidance strategy.

[0009] Optionally, preprocessing of multi-source environmental data includes at least the following steps: performing noise filtering on the multi-source environmental data respectively to obtain effective data of environmental images, effective data of radar, effective data of GPS, and effective data of IMU; taking the timestamp and position information of satellite data as a reference, performing time synchronization and spatial alignment on the multi-source environmental data; performing feature extraction on the effective data of environmental images to obtain the position data of physical objects; performing feature extraction on the effective radar data to obtain one or more of the distance, speed, radial speed, and angle of the physical objects; performing feature extraction on the effective GPS data to obtain the vehicle positioning data; performing feature extraction on the effective IMU data to obtain one or more of the posture data, acceleration, and angular velocity of the vehicle.

[0010] Optionally, the physical objects include one or more of other vehicles, pedestrians, animals, obstacles, traffic signs, and lane lines.

[0011] Optionally, each intelligent agent is divided according to driving function requirements, and the driving function requirements include: environmental perception, path planning, behavior decision-making and control execution.

[0012] Optionally, the local decisions corresponding to each agent are fused, which includes at least the following steps:

[0013] Verify local decisions; set weight values ​​for each local decision based on the type of driving scenario to determine the priority of each local decision; merge local decisions using preset merging rules or algorithms.

[0014] Optionally, it also includes: iteratively optimizing multi-agent parameters based on environmental perception data and global driving decisions to adjust the decision-making strategy of each agent; optimizing each agent based on the comparison results of the local decision corresponding to each agent and the global decision.

[0015] Optionally, it also includes: performing periodic detection on all intelligent agents in the multi-intelligent system according to the abnormality detection cycle, and if an abnormality exists in the intelligent agent, handling the faulty intelligent agent through a preset intelligent agent abnormality response method.

[0016] Optionally, the method for coping with agent abnormalities includes at least one of restarting the faulty agent, switching the faulty agent to a backup agent, and reconfiguring multi-agent parameters.

[0017] Corresponding to the distributed driving decision generation method, the present invention provides a distributed driving decision generation system, which includes: a data acquisition module, which is used to collect multi-source environmental data around the vehicle, and the multi-source environmental data includes at least two or more data among environmental image data, radar data, satellite data, and IMU data; a data processing module, which is used to pre-process the multi-source environmental data and perform data fusion on the pre-processed multi-source environmental data to obtain environmental perception data; a driving situation recognition module, which is used to identify the current driving situation based on the environmental perception data through a deep learning algorithm; a driving situation classification module, which is used to classify the current driving situation according to a preset classification method to determine the type of driving situation in which the vehicle is located, and the driving situation type is urban road, highway, complex intersection or bad weather; a local decision generation module, which is used for each intelligent agent in the multi-intelligent system to share environmental perception data and driving situation type through a distributed protocol to generate local decisions corresponding to each intelligent agent; a global driving decision generation module, which is used to perform decision fusion on the local decisions corresponding to each intelligent agent to form a global driving decision, and the global driving decision includes three contents: path planning, speed control and obstacle avoidance strategy.

[0018] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which a distributed driving decision generation program is stored. When the distributed driving decision generation program is executed by a processor, the steps of the distributed driving decision generation method as described above are implemented.

[0019] The beneficial effects of the present invention are:

[0020] (1) Compared with the prior art, the present invention realizes accurate classification of complex driving situation types, providing a reliable basis for the formulation of intelligent driving strategies; flexibly adjusts driving strategies and decision priorities according to driving situation types, improving the flexibility of intelligent driving; realizes information sharing and collaboration among intelligent agents through distributed protocols, improving the intelligence and robustness of overall decision-making; further integrates the local decisions of each intelligent agent to form a globally optimal intelligent driving decision, ensuring the advancement and safety of the intelligent driving method; the present invention can be widely used in various scenarios that require automatic generation of driving decisions, providing a wider range of technical application possibilities;

[0021] (2) Compared with the prior art, the present invention ensures the consistency of multi-source environmental data in the time dimension through preprocessing, provides a basis for subsequent data fusion, improves the purity of data, and reduces the risk of misjudgment and misoperation; and can extract key information from valid data, providing strong support for subsequent situation recognition and decision-making;

[0022] (3) Compared with the prior art, the present invention not only focuses on vehicles and pedestrians, but also considers multiple objects such as animals, obstacles, traffic signs, lane lines, etc., thus improving the comprehensiveness and accuracy of environmental perception;

[0023] (4) Compared with the prior art, the present invention divides the intelligent agents according to the driving function requirements, realizes the decoupling and modularization of functions, and facilitates system maintenance and upgrade;

[0024] (5) Compared with the prior art, the present invention ensures driving safety and efficiency by integrating the local decisions corresponding to each intelligent agent;

[0025] (6) Compared with the prior art, the present invention continuously improves the decision-making ability of each agent and the overall performance of the system by iteratively optimizing the parameters of the multi-agent. Moreover, based on the comparison results between local and global decisions, the agent is optimized in a targeted manner, thereby improving the accuracy and stability of the system.

[0026] (7) Compared with the prior art, the present invention can timely detect and handle agent anomalies through periodic detection, thus ensuring the continuous and stable operation of the system;

[0027] (8) Compared with the prior art, the present invention can quickly and effectively handle intelligent agent anomalies through preset response methods, reducing the impact of faults on the intelligent driving system. It also provides a variety of abnormality response solutions, and the most appropriate processing method can be selected according to actual conditions, thereby improving the flexibility and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0029] Figure 1 A simplified flowchart of an embodiment of a driving decision-making generation method based on distributed management according to the present invention;

[0030] Figure 2 It is a framework diagram of an embodiment of a distributed driving decision generation system according to the present invention. DETAILED DESCRIPTION

[0031] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] like Figure 1 As shown, a driving decision generation method based on distribution of the present invention includes the following steps: collecting multi-source environmental data around the vehicle, the multi-source environmental data including at least two types of environmental image data, radar data, satellite data, and IMU data; preprocessing the multi-source environmental data, and fusing the preprocessed multi-source environmental data to obtain environmental perception data; identifying the current driving situation based on the environmental perception data through a deep learning algorithm; classifying the current driving situation according to a preset classification method to determine the type of driving situation in which the vehicle is located, the driving situation type being urban roads, highways, complex intersections, or bad weather; sharing environmental perception data and driving situation types between the agents in the multi-intelligent system through a distributed protocol to generate local decisions corresponding to each agent; fusing the local decisions corresponding to each agent to form a global driving decision, the global driving decision including path planning, speed control, and obstacle avoidance strategy.

[0033] In this embodiment, data fusion of the pre-processed multi-source environmental data specifically includes the following steps:

[0034] Correlate physical objects from different sensors to achieve target matching; for example, by comparing the location of a physical object in environmental image data with the distance and angle of a physical object in radar data to confirm whether they are the same physical object;

[0035] For multi-source data that have been confirmed to include the same physical object, their feature values ​​(such as position, speed, etc.) are merged to achieve feature fusion.

[0036] In this embodiment, the current driving situation is mainly identified through the following aspects:

[0037] Traffic conditions: Is it congested? Are there any accidents?

[0038] Weather conditions: sunny, rainy, snowy, foggy, etc.

[0039] Time of day: Is it day or night? Is it peak hour?

[0040] Road type: highway, city street, country road, etc.

[0041] Regulatory restrictions: Legal requirements for speed limits, prohibited lane changes, etc.

[0042] In this embodiment, the deep learning algorithm used to identify the current driving situation is pre-trained based on the situation model. The training process specifically includes: selecting a suitable neural network structure and using a large number of labeled environmental perception data sets to train the model. The training labels include traffic status, weather conditions, time factors, road types, regulatory restrictions, etc.; using the back propagation algorithm to adjust the weights so that the inference results are as close as possible to the actual labels. In the actual application stage, the environmental perception data can be input to obtain the current traffic status, weather conditions, time factors, road types, regulatory restrictions, etc.

[0043] In this embodiment, the preset classification method is to classify according to road type first, with urban roads as one category and highways as another category, and then further subdivided according to weather conditions. For adverse driving conditions such as rain, snow, fog, etc. that cause reduced visibility or slippery roads, they are classified as bad weather categories. Finally, for situations that are neither urban roads nor highways (such as rural roads) and situations with traffic congestion, they are all classified as complex intersections. It can be understood that the above categories are only simple classifications in the embodiment of the present invention. In actual application, more categories can be considered as needed.

[0044] It should be made clear that local decisions also include one or two of the following: path planning, speed control, and obstacle avoidance strategies, which are consistent with global decisions. The difference in local decisions lies in the sensitivity of each agent to environmental perception data. For example, in the distributed system, agent 1 is more sensitive to giving way to pedestrians, while agent 2 is more sensitive to traffic efficiency, resulting in decision differences.

[0045] Preferably, the decision priorities are from high to low: safety, legality, comfort, and efficiency. That is, safety decisions have the highest priority, such as obstacle avoidance, emergency stop, and distance keeping; compliance with regulations is second, such as traffic lights and signs, speed control, lane use, etc.; the third is comfort, including smooth acceleration and deceleration, smooth steering, and the fourth is efficiency optimization, such as path planning, lane change and overtaking, etc.

[0046] It is understandable that before collecting multi-source environmental data around the vehicle, it is necessary to ensure that the models of each agent are successfully loaded and a reliable communication link is established. Through distributed model loading, each agent can be quickly started and assigned appropriate functions. At the same time, the communication verification step ensures the accuracy and timeliness of information transmission between agents, laying a solid foundation for the subsequent collaborative work of the system. Specifically, it includes:

[0047] a) Model distributed loading: When the intelligent driving system starts, the various agent models required by the intelligent driving system are loaded first. Each model is distributed on different computing nodes to achieve parallel computing and distributed processing.

[0048] b) Function allocation: The system allocates functions such as perception, decision-making and control to different agents according to actual driving needs. Each agent prepares the corresponding algorithm module according to its assigned function.

[0049] c) Communication verification: After the agent is initialized, the system performs communication verification to ensure that the communication links between all agents are unobstructed and can exchange information in real time.

[0050] The present invention realizes accurate classification of complex driving situation types, providing a reliable basis for the formulation of intelligent driving strategies; flexibly adjusts driving strategies and decision priorities according to driving situation types, thereby improving the flexibility of intelligent driving; realizes information sharing and collaboration among intelligent agents through distributed protocols, thereby improving the intelligence and robustness of overall decision-making; further integrates the local decisions of each intelligent agent to form a globally optimal intelligent driving decision, thereby ensuring the advancement and safety of the intelligent driving method; the present invention can be widely used in various scenarios that require automatic generation of driving decisions, providing a wider range of technical application possibilities.

[0051] The present invention realizes the data sharing and information transmission mechanism among various intelligent agents through the data sharing part, ensuring that all intelligent agents can obtain the necessary environmental information and decision-making data, which is crucial to maintaining the overall consistency and coordination of the system.

[0052] In this embodiment, preprocessing of multi-source environmental data includes at least the following steps: performing noise filtering on the multi-source environmental data to obtain effective environmental image data, radar effective data, GPS effective data, and IMU effective data; taking the timestamp and position information of the satellite data as a reference, performing time synchronization and spatial alignment on the multi-source environmental data; performing feature extraction on the effective environmental image data to obtain the position data of the physical object; performing feature extraction on the radar effective data to obtain one or more of the distance, speed, radial speed, and angle of the physical object; performing feature extraction on the GPS effective data to obtain the vehicle positioning data; performing feature extraction on the IMU effective data to obtain one or more of the vehicle's posture data, acceleration, and angular velocity.

[0053] In this embodiment, the timestamp and position information of the satellite data are used as a reference to perform time synchronization and spatial alignment on the multi-source environmental data, specifically including: timestamp standardization processing, coordinate system conversion processing and unit consistency processing.

[0054] In this embodiment, the coordinate system conversion process is specifically: converting the vehicle coordinate system, the world coordinate system, and the sensor coordinate system into a unified coordinate system, which can be specifically achieved through existing technologies and will not be described in detail here. Among them, the vehicle coordinate system refers to a local coordinate system centered on the vehicle, the world coordinate system is a global coordinate system, usually established based on geographic coordinate information provided by satellite data; the sensor coordinate system specifically refers to the local coordinate system of each sensor, for example, the camera may be based on the image plane, and the radar has a polar coordinate system, etc.

[0055] In this embodiment, the unit consistency processing refers to the unification of units such as distance and speed.

[0056] The present invention ensures the consistency of multi-source environmental data in the time dimension through preprocessing, provides a basis for subsequent data fusion, reduces redundant information and noise interference in the data, thereby improving the overall perception accuracy, improving the purity of the data, and reducing the risks of misjudgment and misoperation; and can extract key information from valid data, providing strong support for subsequent situation recognition and decision-making.

[0057] In this embodiment, the physical objects include one or more of other vehicles, pedestrians, animals, obstacles, traffic signs, and lane lines.

[0058] The present invention not only focuses on vehicles and pedestrians, but also takes into account multiple objects such as animals, obstacles, traffic signs, lane lines, etc., thereby improving the comprehensiveness and accuracy of environmental perception.

[0059] In this embodiment, each agent is divided according to the driving function requirements, which include: environmental perception, path planning, behavioral decision-making and control execution. Through such functional division, it can be ensured that each agent can maximize its performance in its field of expertise, thereby improving the response speed and decision-making accuracy of the entire system. Specifically, the local decisions generated by different agents are different. Agents focusing on environmental perception are very good at obstacle avoidance, agents focusing on path planning are more concerned about traffic efficiency, agents focusing on behavioral decision-making are good at real-time acceleration, deceleration, lane change and other operations, and agents focusing on control execution focus on precise control of the vehicle itself.

[0060] The present invention divides the intelligent agents according to the driving function requirements, realizes the decoupling and modularization of functions, and facilitates system maintenance and upgrade. Dividing the intelligent agents by driving function requirements can also ensure that each intelligent agent can maximize its performance in its field of expertise, thereby improving the response speed and decision-making accuracy of the entire system.

[0061] In this embodiment, the local decisions corresponding to each agent are fused, which includes at least the following steps:

[0062] Verify local decisions to ensure data consistency and accuracy, which helps identify potential data errors or malicious attacks;

[0063] According to the type of driving scenario, the weight value of each local decision is set to determine the priority of each local decision. For example, speed control may be more important than obstacle avoidance on a highway, while obstacle avoidance becomes the primary consideration at a complex intersection. In addition, different weight values ​​are assigned to each agent based on factors such as safety and efficiency to adjust its importance in the overall decision.

[0064] Local decisions are merged using preset merging rules or algorithms, which can adopt voting mechanisms or methods based on cost function optimization. If there are major conflicts in local decisions, human drivers or other external interventions are required as the final merging means.

[0065] It should be noted that since each agent is divided according to the driving function requirements, there are at least four agents in this example. Each agent has the ability to make decisions, so it will generate local decisions for path planning, speed control, and obstacle avoidance strategies. The weight value of each content in the local decision is determined by the functional focus of each agent. For example: an agent that focuses on environmental perception will set a higher weight value for path planning and obstacle avoidance strategies than other contents; an agent that focuses on control execution will set a higher weight value for speed control than other contents; an agent that focuses on path planning will set a higher weight value for path planning than other contents; an agent that focuses on behavioral decision-making will set a higher weight value for speed control and obstacle avoidance strategies than other contents.

[0066] However, when the local decisions corresponding to each agent are subsequently fused, the driving scenario type needs to be considered. For example, when the driving scenario type is a highway, during decision fusion, the local decision generated by the agent focusing on behavioral decision making will be set with a higher weight value than other local decisions; when the driving scenario type is an urban highway, the local decision generated by the agent focusing on path planning will be set with a higher weight value than other local decisions; when the driving scenario type is a complex intersection, the local decision generated by the agent focusing on environmental perception will be set with a higher weight value than other local decisions; when the driving scenario type is bad weather, the local decision generated by the agent focusing on control execution will be set with a higher weight value than other local decisions.

[0067] It should be noted that the weight setting bias of the local decision content of each intelligent agent is only an exemplary description, and can be adjusted according to actual needs in practical applications, and does not constitute an improper limitation on the present invention. It is understandable that the specific numerical value of the weight value can be set according to actual needs, as long as it can meet the requirements of setting the weight value of the present invention.

[0068] The present invention ensures driving safety and efficiency by integrating the local decisions corresponding to each intelligent agent.

[0069] In this embodiment, it also includes: iteratively optimizing multi-agent parameters based on environmental perception data and global driving decisions to adjust the decision-making strategy of each agent; optimizing each agent based on the comparison results of the local decision corresponding to each agent and the global decision.

[0070] The present invention continuously improves the decision-making ability of each agent and the overall performance of the system by iteratively optimizing multi-agent parameters; and, based on the comparison results of local decisions and global decisions, targeted optimization of the agents is performed to improve the accuracy and stability of the system.

[0071] In this embodiment, it also includes: periodically detecting all intelligent agents in the multi-intelligent system according to the abnormality detection cycle, and if there is an abnormality in the intelligent agent, processing the faulty intelligent agent through a preset intelligent agent abnormality response method.

[0072] Through periodic detection, the present invention can timely discover and handle intelligent agent anomalies (such as hardware failure, sensor failure, communication interruption, etc.), ensuring the continuous and stable operation of the system.

[0073] In this embodiment, the method for coping with agent abnormalities includes at least one of restarting the faulty agent, switching the faulty agent to a backup agent, and reconfiguring multi-agent parameters.

[0074] Preferably, reconfiguring the multi-agent parameters is specifically resetting to the parameter state (i.e., the agent model state) of the last iteration.

[0075] The present invention can quickly and effectively handle intelligent agent anomalies through preset response methods, reduce the impact of faults on the intelligent driving system, and also provide a variety of abnormality response plans. The most appropriate processing method can be selected according to actual conditions, thereby improving the flexibility and reliability of the system.

[0076] like Figure 2 As shown, the present invention also provides a driving decision generation system based on distributed control, which includes: a data acquisition module 10, which is used to collect multi-source environmental data around the vehicle, and the multi-source environmental data includes at least two or more data among environmental image data, radar data, satellite data, and IMU data; a data processing module 20, which is used to pre-process the multi-source environmental data and fuse the pre-processed multi-source environmental data to obtain environmental perception data; a driving situation recognition module 30, which is used to identify the current driving situation based on the environmental perception data through a deep learning algorithm; a driving situation classification module 40, which is used to classify the current driving situation according to a preset classification method to determine the type of driving situation in which the vehicle is located, and the driving situation type is urban road, highway, complex intersection or bad weather; a local decision generation module 50, which is used for each intelligent agent in the multi-intelligent system to share environmental perception data and driving situation type through a distributed protocol to generate local decisions corresponding to each intelligent agent; a global driving decision generation module 60, which is used to fuse the local decisions corresponding to each intelligent agent to form a global driving decision, and the global driving decision includes three contents: path planning, speed control and obstacle avoidance strategy.

[0077] In this embodiment, the distributed driving decision generation system also includes: an optimization module for iteratively optimizing multi-agent parameters based on environmental perception data and global driving decisions (the parameters that usually need to be iteratively optimized refer to the parameters of the neural network model) to adjust the decision strategy of each agent; and optimizing each agent based on the comparison results of the local decision corresponding to each agent and the global decision.

[0078] In this embodiment, the distributed driving decision-making system also includes: a safety detection module, which is used to perform periodic detection on all intelligent agents in the multi-intelligent system according to the abnormality detection cycle. If there is an abnormality in the intelligent agent, the faulty intelligent agent is handled by a preset intelligent agent abnormality response method.

[0079] The embodiment of the present invention further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the memory in the above embodiment; or a computer-readable storage medium that exists independently and is not installed in a device. The computer-readable storage medium stores at least one instruction, which is loaded and executed by a processor to implement Figure 1 The driving decision generation method based on distribution is shown. The computer readable storage medium can be a read-only memory, a disk or an optical disk, etc.

[0080] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the system embodiment and the storage medium embodiment, since they are basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0081] Furthermore, in this document, the terms "comprises," "comprising," or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device that includes the element.

[0082] The above description shows and describes the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the invention, through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art shall not depart from the spirit and scope of the present invention, and shall be within the scope of protection of the claims attached to the present invention.

Claims

1. A driving decision generation method based on distribution, characterized in that: The following steps are involved: Collect multi-source environmental data around the vehicle, where the multi-source environmental data includes at least two of the following: environmental image data, radar data, satellite data, and IMU data; Preprocess the multi-source environmental data, and perform data fusion on the preprocessed multi-source environmental data to obtain environmental perception data; Preprocessing the multi-source environmental data includes at least the following steps: performing noise filtering on the multi-source environmental data to obtain effective environmental image data, effective radar data, effective GPS data, and effective IMU data; taking the timestamp and location information of the satellite data as a reference, performing time synchronization and spatial alignment on the multi-source environmental data; performing feature extraction on the effective environmental image data to obtain the location data of the physical object; performing feature extraction on the effective radar data to obtain one or more of the distance, speed, radial speed, and angle of the physical object; performing feature extraction on the effective GPS data to obtain the vehicle positioning data; performing feature extraction on the effective IMU data to obtain one or more of the posture data, acceleration, and angular velocity of the vehicle; Through deep learning algorithms, the current driving situation is identified based on environmental perception data; Classify the current driving situation according to a preset classification method to determine the type of driving situation the vehicle is in, where the driving situation type is urban road, highway, complex intersection or bad weather; The agents in the multi-intelligence system share environmental perception data and driving scenario types through a distributed protocol to generate local decisions corresponding to each agent; The local decisions of each agent are integrated to form a global driving decision, which includes path planning, speed control and obstacle avoidance strategy. Each agent is divided according to the driving function requirements, which include: environmental perception, path planning, behavioral decision-making and control execution; the weight value of each content in the local decision is determined by the functional focus ability of each agent; among them, the agent focusing on environmental perception will set a higher weight value for path planning and obstacle avoidance strategy than other contents; the agent focusing on control execution will set a higher weight value for speed control than other contents; the agent focusing on path planning will set a higher weight value for path planning than other contents; the agent focusing on behavioral decision-making will set a higher weight value for speed control and obstacle avoidance strategy than other contents; The decision fusion of the local decisions corresponding to each agent includes at least the following steps: Verify local decisions; According to the driving scenario type, the weight value of each local decision is set to determine the priority of each local decision; Merge local decisions using preset merging rules or algorithms; When fusing the local decisions corresponding to each agent, the driving scenario type needs to be considered; when the driving scenario type is a highway, during decision fusion, the local decision generated by the agent focusing on behavioral decision-making will be set with a higher weight value than other local decisions; when the driving scenario type is an urban highway, the local decision generated by the agent focusing on path planning will be set with a higher weight value than other local decisions; when the driving scenario type is a complex intersection, the local decision generated by the agent focusing on environmental perception will be set with a higher weight value than other local decisions; when the driving scenario type is severe weather, the local decision generated by the agent focusing on control execution will be set with a higher weight value than other local decisions.

2. The distributed driving decision generation method according to claim 1, characterized in that: Physical objects include other vehicles, pedestrians, animals, obstacles, traffic signs, lane lines, or more than one of them.

3. The distributed driving decision generation method according to claim 1, characterized in that: Also includes: Iteratively optimize multi-agent parameters based on environmental perception data and global driving decisions to adjust the decision-making strategies of each agent; Each intelligent agent is optimized based on the comparison results of its corresponding local decision and global decision.

4. The distributed driving decision making method according to claim 1, characterized in that: Also includes: All agents in the multi-intelligent system are periodically checked according to the anomaly detection cycle. If an agent is abnormal, the faulty agent is handled through the preset agent abnormality response method.

5. The distributed driving decision making method according to claim 4, characterized in that: The method for coping with agent abnormalities includes at least one of restarting the faulty agent, switching the faulty agent to a backup agent, and reconfiguring multi-agent parameters.

6. A distributed driving decision-making system, characterized in that: include: A data acquisition module is used to collect multi-source environmental data around the vehicle, where the multi-source environmental data includes at least two or more of environmental image data, radar data, satellite data, and IMU data; A data processing module is used to preprocess multi-source environmental data and perform data fusion on the preprocessed multi-source environmental data to obtain environmental perception data; Preprocessing the multi-source environmental data includes at least the following steps: performing noise filtering on the multi-source environmental data to obtain effective environmental image data, effective radar data, effective GPS data, and effective IMU data; taking the timestamp and location information of the satellite data as a reference, performing time synchronization and spatial alignment on the multi-source environmental data; performing feature extraction on the effective environmental image data to obtain the location data of the physical object; performing feature extraction on the effective radar data to obtain one or more of the distance, speed, radial speed, and angle of the physical object; performing feature extraction on the effective GPS data to obtain the vehicle positioning data; performing feature extraction on the effective IMU data to obtain one or more of the posture data, acceleration, and angular velocity of the vehicle; A driving situation recognition module, which is used to recognize the current driving situation based on environmental perception data through a deep learning algorithm; A driving situation classification module is used to classify the current driving situation according to a preset classification method to determine the type of driving situation the vehicle is in, where the driving situation type is urban road, highway, complex intersection or bad weather; The local decision generation module is used for sharing environmental perception data and driving scenario types between agents in the multi-intelligent system through a distributed protocol to generate local decisions corresponding to each agent; each agent is divided according to driving function requirements, which include: environmental perception, path planning, behavioral decision-making and control execution; the weight value of each content in the local decision is determined by the functional focus of each agent; among them, the agent focusing on environmental perception will set higher weight values ​​for path planning and obstacle avoidance strategies than other contents; the agent focusing on control execution will set higher weight values ​​for speed control than other contents; the agent focusing on path planning will set higher weight values ​​for path planning than other contents; the agent focusing on behavioral decision-making will set higher weight values ​​for speed control and obstacle avoidance strategies than other contents; The global driving decision generation module is used to fuse the local decisions corresponding to each intelligent agent to form a global driving decision, which includes three contents: path planning, speed control and obstacle avoidance strategy; the decision fusion of the local decisions corresponding to each intelligent agent includes at least the following steps: verifying the local decision; setting the weight value of each local decision according to the driving scenario type to determine the priority of each local decision; merging the local decisions using a preset merging rule or algorithm; when fusing the local decisions corresponding to each intelligent agent, the driving scenario type needs to be considered; among which, when the driving scenario type is a highway, during the decision fusion, the weight value of the local decision generated by the intelligent agent focusing on behavioral decision-making will be set higher than that of other local decisions; when the driving scenario type is an urban highway, the weight value of the local decision generated by the intelligent agent focusing on path planning will be set higher than that of other local decisions; when the driving scenario type is a complex intersection, the weight value of the local decision generated by the intelligent agent focusing on environmental perception will be set higher than that of other local decisions; when the driving scenario type is bad weather, the weight value of the local decision generated by the intelligent agent focusing on control execution will be set higher than that of other local decisions.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a distributed driving decision generation program, and when the distributed driving decision generation program is executed by the processor, the steps of the distributed driving decision generation method according to any one of claims 1 to 5 are implemented.

Images