Unmanned driving control method and system based on brain-like perception

Abstract

The invention discloses an unmanned driving control method and system based on brain-like perception, and the method comprises the steps: carrying out the visual neuron analysis of external driving environment data, judging whether the external driving environment is in a strange driving environment or not, and adjusting the driving memory data call of a memory neuron; fusing the space construction information of the external driving environment and the called driving memory data to generate a short-term driving plan, and combining the driving experience data to generate a driving operation task; and performing navigation neuron analysis on the driving motion data during the execution period of the driving operation task to obtain driving deviation information so as to generate a driving change decision, thereby changing the driving state on line. Driving environment data is analyzed and processed by using brain-like perception, proper driving memory data is called to complete driving planning, driving motion data is analyzed and processed to obtain driving deviation information, online change of a driving state is completed, change continuity and implementation stability of driving operation are ensured, and driving control robustness and accuracy are improved.

Description

A brain-like perception-based autonomous driving control method and system Technical Field

[0001] This invention relates to the field of autonomous driving, and in particular to an autonomous driving control method and system based on brain-like perception. Background Technology

[0002] Autonomous driving, as a key development direction of intelligent driving, currently utilizes a large number of different types of sensors to detect surrounding environmental data and analyze it to adjust the vehicle's driving state. Sensors, acting as the "sensory organs" of autonomous driving scenarios, provide a reliable and sufficient data foundation for driving control. However, autonomous driving typically focuses on judging and predicting complex driving environments and unexpected events. This requires sufficiently powerful computing power to support the processing of surrounding environmental data, inevitably leading to sudden changes and disruptions in vehicle control. It cannot effectively simulate the driving operations of real humans, reducing the smoothness of autonomous driving implementation and the passenger experience. Furthermore, current autonomous driving systems are highly dependent on the environment; when external environmental conditions change complexly, they cannot guarantee corresponding control robustness and accuracy, and cannot maintain good autonomous control performance. Summary of the Invention

[0003] Considering that existing autonomous driving systems typically focus on processing massive amounts of sensor data to cope with complex driving scenarios and unexpected events, they are prone to discontinuous changes in driving operations, reducing the smoothness of autonomous driving implementation and passenger experience. Furthermore, they are severely affected by changes in external environmental conditions, reducing the robustness and accuracy of driving control. In view of the above problems, this invention proposes a brain-like perception-based autonomous driving control method to overcome or at least partially solve the above problems. The method includes: sensing external driving environment data; performing visual neuron analysis on the external driving environment data to determine whether the driving environment is unfamiliar; adjusting the retrieval of driving memory data by memory neurons based on the determination of the external driving environment; fusing spatial structure information of the external driving environment and the retrieved driving memory data to generate a short-term driving plan; generating a driving operation task based on the short-term driving plan and driving experience data; sensing vehicle motion data during the execution of the driving operation task; performing navigation neuron analysis on the vehicle motion data to obtain driving deviation information; and generating a driving change decision based on the driving deviation information to change the driving state online.

[0004] Optionally, external driving environment data is perceived, and visual neural network analysis is performed on the external driving environment data to determine whether the driving environment is unfamiliar. Based on the determination of the external driving environment, the driving memory data retrieval of memory neurons is adjusted, including: perceiving external driving environment image data and external driving environment depth data; performing a first analysis of the external driving environment image data using visual neurons to obtain the object contours and color features of the external driving environment; performing a second analysis of the external driving environment depth data using visual neurons to obtain the object spatial depth features of the external driving environment; and fusing the object contour and color features with the object spatial depth features. The system obtains a set of visual perception features of objects; compares the set of visual perception features with historical driving environment memory data to obtain the object structure similarity between the external driving environment and the historical driving environment, thereby determining whether the driving environment is unfamiliar; if the driving environment is unfamiliar, the system adjusts the memory neurons to call driving operation memory data according to the actual available driving space structure of the external driving environment; wherein, the actual available driving space structure refers to the space structure of the external driving environment that allows the vehicle to drive through in terms of internal dimensions; if the driving environment is not unfamiliar, the system adjusts the memory neurons to call historical driving path data that matches the actual location according to the actual location of the external driving environment.

[0005] Optionally, sensing external driving environment image data and external driving environment depth data includes: real-time acquisition of the current UAV's real-time flight speed; comparing the current UAV's real-time flight speed with a preset flight speed threshold; if the current UAV's real-time flight speed is lower than the preset flight speed threshold, then the data sensing frequency of the current sensing external driving environment image data and external driving environment depth data is not adjusted; if the current UAV's real-time flight speed is not lower than the preset flight speed threshold, then retrieving the number of effective obstacles in the current UAV's field of vision and the average flight speed in the current flight path; adjusting the data sensing frequency of the external driving environment image data using the current UAV's real-time flight speed combined with the number of effective obstacles in the current UAV's field of vision, wherein the adjusted sensing frequency of the external driving environment image data is obtained by the following formula: , where f ing This indicates the perception frequency of the adjusted external driving environment image data; f 0，ing α represents the perception frequency of the external driving environment image data before adjustment; α represents the vehicle speed sensitivity coefficient of the image data, ranging from 1.3 to 1.8; v represents the current flight speed of the drone; v0 represents the preset flight speed threshold value of the image data; β represents the obstacle sensitivity coefficient of the image data, ranging from 0.3 to 1.1; C m Indicates the number of effective obstacles within the current drone's field of view; C obsk represents the maximum number of identifiable obstacles in the image. ing The image environment complexity correction coefficient is represented by the following formula: , where k ing，base represents the basic complexity coefficient, with a value of 0.8; w1, w2, and w3 represent the weight values ​​corresponding to obstacle density, illumination variation rate, and lane line sharpness in the image; k ing，gain This represents the complexity gain coefficient, with a value of 0.17; ρ obs ΔL represents the obstacle density in the normalized image; ΔL represents the rate of change of illumination after normalization; S lane The normalized lane line ambiguity is represented by the following formula: The data perception frequency of the external driving environment depth data is adjusted using the real-time flight speed of the UAV combined with the average flight speed along the current flight path. , where f depth This indicates the perception frequency of the adjusted external driving environment depth data; f 0，depth The sensor frequency of the external driving environment depth data before adjustment; v represents the current flight speed of the drone; v min represents the minimum effective flight speed for depth data; r represents the vehicle speed exponent coefficient for depth data, ranging from 0.4 to 0.8; D ef Indicates the effective depth of field range of the current scene; D ref Indicates the reference depth of field range; v avg This represents the average flight speed corresponding to the current flight path; v avg，max k represents the maximum permissible average flight speed corresponding to the current flight path. depth This represents the depth-of-field environment complexity correction coefficient, which is obtained using the following formula: , where k depth，base This represents the basic coefficient for depth of field complexity, with a value of 0.7; k depth，gain This represents the depth-of-field complexity gain coefficient, with a value of 0.15; ρ h The distance abrupt change density is represented by the normalized value (core factor: the number of points with "distance value abrupt changes" per unit distance in the depth map, reflecting obstacle edges, road surface steps, etc.); h represents the normalized road surface undulation rate; Z represents the normalized depth confidence inverse value; x represents the nonlinear coupling index, with a value range of 0.9-1.3; the external driving environment image data and external driving environment depth data are perceptually manipulated according to the perception frequency of the adjusted external driving environment image data and external driving environment depth data.

[0006] Optionally, a short-term driving plan is generated by integrating spatial structure information of the external driving environment and retrieved driving memory data. Based on the short-term driving plan and driving experience data, a driving operation task is generated, including: when in an unfamiliar driving environment, cognitively filtering the spatial structure information of the external driving environment and identifying a sub-space layout that meets preset driving space conditions; wherein, the spatial structure information refers to the structural dimensions of the barrier-free space in the external driving environment; generating a short-term driving plan based on the sub-space layout and retrieved driving operation memory data; wherein, the short-term driving plan includes adjustments to the driving direction and speed corresponding to a preset distance traveled from the current position within the sub-space layout; when not in an unfamiliar driving environment, a short-term driving plan is generated by integrating spatial structure information of the external driving environment and retrieved historical driving path data; wherein, the short-term driving plan includes adjustments to the driving direction and speed corresponding to a preset distance traveled along a historical driving path in the external driving environment; and generating a driving operation task for the vehicle based on the short-term driving plan and driving experience data for the vehicle; wherein, the driving operation task includes time-related operations on at least one of the vehicle's steering wheel, accelerator, and brake.

[0007] Optionally, the driving motion data during the execution of the driving operation task is sensed, and navigation neuron analysis is performed on the driving motion data to obtain driving deviation information; based on the driving deviation information, a driving change decision is generated to change the driving state online, including: sensing driving direction data and driving speed data during the execution of the driving operation task; inputting the driving direction data and driving speed data into the grid cells and head orientation cells of the navigation neuron for analysis, respectively, to obtain a driving direction deviation sequence and a driving speed deviation sequence during driving; wherein, the driving direction deviation sequence includes the deviation angle between the actual driving direction and the planned driving direction in several consecutive unit time intervals; the driving speed deviation sequence includes the deviation speed between the actual driving speed and the planned driving speed in several consecutive unit time intervals; a driving change decision is generated based on the driving direction deviation sequence and the driving speed deviation sequence; wherein, the driving change decision includes the timing of the driving change and the content of the driving change action; based on the driving change decision, the action of at least one of the steering wheel, accelerator, and brake is changed online according to the driving change timing.

[0008] As one aspect of the present invention, embodiments of the present invention also provide an unmanned driving control system based on brain-like perception, comprising: a driving environment recognition module, used to perceive external driving environment data, perform visual neuron analysis on the external driving environment data, and determine whether the driving environment is unfamiliar; a driving memory data retrieval module, used to adjust the driving memory data retrieval of memory neurons according to the judgment result of the external driving environment; a driving plan generation module, used to fuse the spatial structure information of the external driving environment and the retrieved driving memory data to generate a short-term driving plan; a driving task generation module, used to generate a driving operation task according to the short-term driving plan and driving experience data; a driving deviation determination module, used to perceive driving motion data during the execution of the driving operation task, perform navigation neuron analysis on the driving motion data, and obtain driving deviation information; and a driving change module, used to generate a driving change decision according to the driving deviation information, thereby changing the driving state online.

[0009] Optionally, the driving environment recognition module is used to perceive external driving environment data, perform visual neuron analysis on the external driving environment data, and determine whether it is in an unfamiliar driving environment, including: perceiving external driving environment image data and external driving environment depth data; performing a first analysis of the external driving environment image data using visual neurons to obtain the object contours and color features of the external driving environment; performing a second analysis of the external driving environment depth data using visual neurons to obtain the object spatial depth features of the external driving environment; fusing the object contour and color features and the object spatial depth features to obtain an object visual perception feature set; and comparing the object visual perception feature set with historical driving environment records. The system retrieves the similarity of object structures between the external driving environment and the historical driving environment to determine whether the driver is in an unfamiliar driving environment. The driving memory data retrieval module adjusts the retrieval of driving memory data by the memory neurons based on the determination of the external driving environment. This includes: when in an unfamiliar driving environment, adjusting the memory neurons to retrieve driving operation memory data based on the actual available driving space structure of the external driving environment; wherein, the actual available driving space structure refers to the spatial structure of the external driving environment that allows the vehicle to travel through in terms of its internal dimensions; when not in an unfamiliar driving environment, adjusting the memory neurons to retrieve historical driving path data matching the actual location of the external driving environment.

[0010] Optionally, sensing external driving environment image data and external driving environment depth data includes: real-time acquisition of the current UAV's real-time flight speed; comparing the current UAV's real-time flight speed with a preset flight speed threshold; if the current UAV's real-time flight speed is lower than the preset flight speed threshold, then the data sensing frequency of the current sensing external driving environment image data and external driving environment depth data is not adjusted; if the current UAV's real-time flight speed is not lower than the preset flight speed threshold, then retrieving the number of effective obstacles in the current UAV's field of vision and the average flight speed in the current flight path; adjusting the data sensing frequency of the external driving environment image data using the current UAV's real-time flight speed combined with the number of effective obstacles in the current UAV's field of vision, wherein the adjusted sensing frequency of the external driving environment image data is obtained by the following formula: , where f ing This indicates the perception frequency of the adjusted external driving environment image data; f 0，ing α represents the perception frequency of the external driving environment image data before adjustment; α represents the vehicle speed sensitivity coefficient of the image data, ranging from 1.3 to 1.8; v represents the current flight speed of the drone; v0 represents the preset flight speed threshold value of the image data; β represents the obstacle sensitivity coefficient of the image data, ranging from 0.3 to 1.1; C m Indicates the number of effective obstacles within the current drone's field of view; C obs k represents the maximum number of identifiable obstacles in the image. ing The image environment complexity correction coefficient is represented by the following formula: , where k ing，base represents the basic complexity coefficient, with a value of 0.8; w1, w2, and w3 represent the weight values ​​corresponding to obstacle density, illumination variation rate, and lane line sharpness in the image; k ing，gain This represents the complexity gain coefficient, with a value of 0.17; ρ obs ΔL represents the obstacle density in the normalized image; ΔL represents the rate of change of illumination after normalization; S lane The normalized lane line ambiguity is represented by the following formula: The data perception frequency of the external driving environment depth data is adjusted using the real-time flight speed of the UAV combined with the average flight speed along the current flight path. , where f depth This indicates the perception frequency of the adjusted external driving environment depth data; f 0，depth The sensor frequency of the external driving environment depth data before adjustment; v represents the current flight speed of the drone; v minrepresents the minimum effective flight speed for depth data; r represents the vehicle speed exponent coefficient for depth data, ranging from 0.4 to 0.8; D ef Indicates the effective depth of field range of the current scene; D ref Indicates the reference depth of field range; v avg This represents the average flight speed corresponding to the current flight path; v avg，max k represents the maximum permissible average flight speed corresponding to the current flight path. depth This represents the depth-of-field environment complexity correction coefficient, which is obtained using the following formula: , where k depth，base This represents the basic coefficient for depth of field complexity, with a value of 0.7; k depth，gain This represents the depth-of-field complexity gain coefficient, with a value of 0.15; ρ h The distance abrupt change density is represented by the normalized value (core factor: the number of points with "distance value abrupt changes" per unit distance in the depth map, reflecting obstacle edges, road surface steps, etc.); h represents the normalized road surface undulation rate; Z represents the normalized depth confidence inverse value; x represents the nonlinear coupling index, with a value range of 0.9-1.3; the external driving environment image data and external driving environment depth data are perceptually manipulated according to the perception frequency of the adjusted external driving environment image data and external driving environment depth data.

[0011] Optionally, the driving plan generation module is used to integrate the spatial structure information of the external driving environment and the driving memory data to generate a short-term driving plan, including: when in an unfamiliar driving environment, cognitively filtering the spatial structure information of the external driving environment and identifying a sub-space layout that meets preset driving space conditions; wherein, the spatial structure information refers to the structural dimensions of the barrier-free space in the external driving environment; generating a short-term driving plan based on the sub-space layout and the driving operation memory data; wherein, the short-term driving plan includes driving direction and driving speed adjustment operations corresponding to a preset distance traveled from the current position within the sub-space layout; When not in an unfamiliar driving environment, the system integrates spatial structure information of the external driving environment with historical driving path data to generate a short-term driving plan. This short-term driving plan includes adjustments to driving direction and speed corresponding to a preset distance traveled along the historical driving path in the external driving environment. The driving task generation module generates driving operation tasks based on the short-term driving plan and driving experience data, including: generating driving operation tasks for the vehicle based on the short-term driving plan and driving experience data for the vehicle itself. These driving operation tasks include time-related operations on at least one of the vehicle's steering wheel, accelerator, and brake.

[0012] Optionally, the vehicle deviation determination module is used to sense vehicle motion data during the execution of the driving operation task, and perform navigation neuron analysis on the vehicle motion data to obtain vehicle deviation information, including: sensing vehicle direction data and vehicle speed data during the execution of the driving operation task; inputting the vehicle direction data and the vehicle speed data into the grid cells and head orientation cells of the navigation neuron for analysis, respectively, to obtain a vehicle direction deviation sequence and a vehicle speed deviation sequence during driving; wherein, the vehicle direction deviation sequence includes the deviation angle between the actual driving direction and the planned driving direction in several consecutive unit time intervals; the vehicle speed deviation sequence includes the deviation speed between the actual driving speed and the planned driving speed in several consecutive unit time intervals; the driving change module is used to generate a driving change decision based on the vehicle deviation information, thereby changing the driving state online, including: generating a driving change decision based on the vehicle direction deviation sequence and the vehicle speed deviation sequence; wherein, the driving change decision includes the timing of the driving change and the content of the driving change action; according to the driving change decision, changing the action of at least one of the steering wheel, accelerator, and brake online according to the timing of the driving change.

[0013] The beneficial effects of the above-mentioned technical solutions provided in the embodiments of the present invention include at least the following: The embodiments of the present invention provide an unmanned driving control method and system based on brain-like perception, which performs visual neuron analysis on external driving environment data to determine whether it is in an unfamiliar driving environment, thereby adjusting the recall of driving memory data by memory neurons; it integrates the spatial structure information of the external driving environment and the recalled driving memory data to generate a short-term driving plan, and combines driving experience data to generate driving operation tasks; it performs navigation neuron analysis on the driving motion data during the execution of the driving operation tasks to obtain driving deviation information, thereby generating driving change decisions and changing the driving state online. By using brain-like perception to analyze and process driving environment data, recalling appropriate driving memory data to complete driving planning, and analyzing and processing driving motion data to obtain driving deviation information, the driving state is changed online, ensuring the continuity and smoothness of driving operation changes, and improving the robustness and accuracy of driving control.

[0014] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0015] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 is a flowchart illustrating the autonomous driving control method based on brain-like perception provided in an embodiment of the invention; Figure 2 is a structural schematic diagram illustrating the autonomous driving control system based on brain-like perception provided in an embodiment of the invention. Detailed Implementation

[0017] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0018] Please refer to Figure 1. An embodiment of this application provides a brain-like perception-based autonomous driving control method. This brain-like perception-based autonomous driving control method includes: sensing external driving environment data; performing visual neuron analysis on the external driving environment data to determine whether the driving environment is unfamiliar; adjusting the retrieval of driving memory data by memory neurons based on the determination result of the external driving environment; fusing spatial structure information of the external driving environment and the retrieved driving memory data to generate a short-term driving plan; generating a driving operation task based on the short-term driving plan and driving experience data; sensing vehicle motion data during the execution of the driving operation task; performing navigation neuron analysis on the vehicle motion data to obtain driving deviation information; and generating a driving change decision based on the driving deviation information to change the driving state online.

[0019] The beneficial effects of the above embodiments are as follows: This brain-like perception-based autonomous driving control method utilizes brain-like perception to analyze and process driving environment data, determines the type of external driving environment, and calls appropriate driving memory data to assist in meeting the needs of different driving environments; it then integrates the spatial structure information of the external driving environment and driving memory data to generate a driving plan, and combines driving experience data to generate driving operation tasks, providing an effective reference for autonomous driving; it also analyzes the vehicle motion data during the execution of driving operation tasks to obtain driving deviation information, generates driving change decisions, meets the online feedback change of driving status, ensures the continuity and smoothness of driving operation changes, and improves the robustness and accuracy of driving control.

[0020] In another embodiment, external driving environment data is perceived, and visual neural network analysis is performed on the external driving environment data to determine whether the driving environment is unfamiliar. Based on the determination of the external driving environment, the driving memory data retrieval of the memory neurons is adjusted, including: perceiving external driving environment image data and external driving environment depth data; performing a first analysis of the external driving environment image data using visual neural networks to obtain the object contours and color features of the external driving environment; performing a second analysis of the external driving environment depth data using visual neural networks to obtain the object spatial depth features of the external driving environment; fusing the object contour and color features and the object spatial depth features to obtain an object visual perception feature set; comparing the object visual perception feature set with historical driving environment memory data to obtain the object structure similarity between the external driving environment and the historical driving environment, thereby determining whether the driving environment is unfamiliar; if the driving environment is unfamiliar, the memory neurons are adjusted to retrieve driving operation memory data based on the actual available driving space structure of the external driving environment; wherein, the actual available driving space structure refers to the space structure of the external driving environment that allows the vehicle to drive through in terms of its internal dimensions; if the driving environment is not unfamiliar, the memory neurons are adjusted to retrieve historical driving path data that matches the actual location of the external driving environment.

[0021] The beneficial effects of the above embodiments are that the external driving environment is the primary problem faced by autonomous driving. Considering the complex spatial and temporal relationships between people and objects in the external driving environment, accurately identifying the dynamic states of people and objects in the external driving environment plays a crucial role in the timeliness and accuracy of autonomous driving responses. Furthermore, autonomous driving generates multimodal historical data on environments it has previously traversed, which not only aids in self-learning but also forms effective driving memory experience, providing an accurate basis for autonomous control and adaptation to the driving environment. The actual driving situations faced by autonomous driving differ between unfamiliar and familiar driving environments (i.e., not unfamiliar driving environments). Generally, autonomous driving faces fewer problems in familiar driving environments and has more historical driving memory data for reference. To improve the control effectiveness of autonomous driving in different types of driving environments, it is necessary to distinguish in advance whether the current driving environment is unfamiliar or familiar.

[0022] Considering that autonomous vehicles collect spatial visual data of the external driving environment, such as roads and buildings, during operation, this data is not only used in the current autonomous driving control process but also stored for self-learning. By comparing the current external driving environment with its stored spatial visual data, the type of external driving environment can be determined. Specifically, camera sensors and depth vision sensors (such as sensors based on binocular vision for depth detection) can be used to perceive external driving environment image data and external driving environment depth data, respectively. The external driving environment image data can be, but is not limited to, RGB images of the external driving environment, and the external driving environment depth data can be, but is not limited to, spatial depth values ​​of all objects in the external driving environment viewed from the vehicle's perspective. The system analyzes external driving environment image data to obtain object contour and color features, and analyzes external driving environment depth data to obtain object spatial depth features (i.e., spatial depth value of each object). Then, using each object as a reference, it fuses the object contour and color features and object spatial depth features corresponding to each object to obtain the object visual perception feature set corresponding to all objects in the external driving environment. It can be understood that the system extracts the contour, color, spatial depth and other multimodal features of each object from the above object visual perception feature set to meet the needs of subsequent judgment of the type of external driving environment.

[0023] Furthermore, historical driving environment memory data is extracted from the vehicle's database. This historical driving environment memory data may include, but is not limited to, records of the visual and spatial states of objects in the driving environment during the vehicle's historical driving process. By comparing the object visual perception feature set with the historical driving environment memory data, the object structure similarity between the external driving environment and the historical driving environment (i.e., the similarity in outline, color, and depth of field between different objects) is obtained. If the total number of objects whose object structure similarity between the external driving environment and the historical driving environment is greater than a preset similarity threshold exceeds a preset number threshold, the external driving environment is determined to be a familiar driving environment (i.e., not an unfamiliar driving environment); otherwise, the external driving environment is determined to be a not unfamiliar driving environment.

[0024] When in an unfamiliar driving environment, the memory neurons retrieve driving operation memory data based on the spatial structure of the external driving environment, which allows the vehicle to traverse the area. This driving operation memory data may include, but is not limited to, historical driving operation memory data corresponding to other spatial structures that are the same as or similar to the aforementioned spatial structures during previous driving experiences. When not in an unfamiliar driving environment, the memory neurons retrieve historical driving path data matching the actual location of the external driving environment. This historical driving path data refers to the historical driving path data matched when the vehicle traversed the aforementioned actual location during previous driving experiences in a familiar driving environment. By distinguishing whether the external driving environment is unfamiliar and retrieving the appropriate type of driving memory data, a reliable basis for forming short-term driving plans can be provided.

[0025] In another embodiment, sensing external driving environment image data and external driving environment depth data includes: real-time acquisition of the current drone's real-time flight speed; comparing the current drone's real-time flight speed with a preset flight speed threshold; if the current drone's real-time flight speed is lower than the preset flight speed threshold, then no adjustment is made to the data sensing frequency of the current external driving environment image data and external driving environment depth data; if the current drone's real-time flight speed is not lower than the preset flight speed threshold, then the number of effective obstacles within the drone's field of vision and the average flight speed along the current flight path are retrieved; the data sensing frequency of the external driving environment image data is adjusted using the current drone's real-time flight speed combined with the number of effective obstacles within the drone's field of vision, wherein the adjusted external driving environment image data sensing frequency is obtained using the following formula: , where f ing This indicates the perception frequency of the adjusted external driving environment image data; f 0，ing α represents the perception frequency of the external driving environment image data before adjustment; α represents the vehicle speed sensitivity coefficient of the image data, ranging from 1.3 to 1.8; v represents the current flight speed of the drone; v0 represents the preset flight speed threshold value of the image data; β represents the obstacle sensitivity coefficient of the image data, ranging from 0.3 to 1.1; C m Indicates the number of effective obstacles within the current drone's field of view; C obs k represents the maximum number of identifiable obstacles in the image. ing The image environment complexity correction coefficient is represented by the following formula: , where k ing，base represents the basic complexity coefficient, with a value of 0.8; w1, w2, and w3 represent the weight values ​​corresponding to obstacle density, illumination variation rate, and lane line sharpness in the image; k ing，gainThis represents the complexity gain coefficient, with a value of 0.17; ρ obs ΔL represents the obstacle density in the normalized image; ΔL represents the rate of change of illumination after normalization; S lane The normalized lane line ambiguity is represented by the following formula: The data perception frequency of the external driving environment depth data is adjusted using the real-time flight speed of the UAV combined with the average flight speed along the current flight path. , where f depth This indicates the perception frequency of the adjusted external driving environment depth data; f 0，depth The sensor frequency of the external driving environment depth data before adjustment; v represents the current flight speed of the drone; v min represents the minimum effective flight speed for depth data; r represents the vehicle speed exponent coefficient for depth data, ranging from 0.4 to 0.8; D ef Indicates the effective depth of field range of the current scene; D ref Indicates the reference depth of field range; v avg This represents the average flight speed corresponding to the current flight path; v avg，max k represents the maximum permissible average flight speed corresponding to the current flight path. depth This represents the depth-of-field environment complexity correction coefficient, which is obtained using the following formula: , where k depth，base This represents the basic coefficient for depth of field complexity, with a value of 0.7; k depth，gain This represents the depth-of-field complexity gain coefficient, with a value of 0.15; ρ h The distance abrupt change density is represented by the normalized value (core factor: the number of points with "distance value abrupt changes" per unit distance in the depth map, reflecting obstacle edges, road surface steps, etc.); h represents the normalized road surface undulation rate; Z represents the normalized depth confidence inverse value; x represents the nonlinear coupling index, with a value range of 0.9-1.3; the external driving environment image data and external driving environment depth data are perceptually manipulated according to the perception frequency of the adjusted external driving environment image data and external driving environment depth data.

[0026] The beneficial effects of the above embodiments are as follows: Traditional systems typically collect data at a fixed frequency, resulting in significant waste of computing power and energy. The core effect of this solution is that the system no longer simply collects data at high speed continuously, but uses the real-time flight speed as the core state variable to trigger frequency adjustment. When the speed is below a threshold, the base frequency is maintained to save resources. During high-speed flight, the system can intelligently allocate limited computing resources to the most needed moments, increasing the sensing frequency to acquire denser environmental information, thereby ensuring safety at critical moments and maximizing computing efficiency. The above solution does not rely solely on speed but deeply integrates environmental characteristics, enabling the sensing system to "understand the complexity of the scene."

[0027] On the one hand, for image data, frequency adjustment is not only related to speed, but also strongly coupled with the number of effective obstacles and the complexity of the image environment. In complex scenes with dense obstacles, abrupt changes in lighting, and blurred lane lines, the system automatically improves its "visual alertness" to capture details at a higher frequency, effectively addressing long-tail problems such as "ghosting," glare from strong light, and worn lane lines. For depth data, frequency adjustment is related to the effective depth range, average path speed, and the complexity of the depth environment. This allows the system to prioritize the modeling accuracy within the current flight space. For example, in complex terrain with numerous obstacle edges (high density of abrupt changes in distance) or drastic road surface undulations, the system proactively increases the scanning frequency to build a more real-time and accurate 3D environment map, providing high-quality data for obstacle avoidance and path planning. During high-speed flight, any perception delay can lead to catastrophic consequences. This solution specifically enhances the system's safety margin. By introducing exponential coefficients (such as α and r), the perception frequency increases superlinearly with speed. This ensures that the system can obtain a far greater flow of perception data than usual at extremely high speeds, significantly shortening the latency of the perception-decision-control closed loop and providing a valuable time window for avoiding dynamic obstacles. Introducing the average flight speed along the current flight path as one of the adjustment factors for the depth-of-field frequency gives the system "path pre-cognition" capabilities. When the system anticipates entering a path that typically requires high-speed flight, it will increase the depth-of-field perception frequency in advance or simultaneously. For drones, energy consumption is life. This solution, through intelligent adjustment, brings direct benefits to battery life.

[0028] On the other hand, sensors and computing units are major power consumers on drones. By automatically reducing the sensing frequency in low-risk scenarios (low speed, simple environment), the system can significantly reduce unnecessary power consumption, thereby effectively extending the flight time per flight. Reducing the computational load also means reducing processor heat generation, which helps maintain the system operating at a better temperature and improves the reliability and lifespan of electronic components.

[0029] This adaptive mechanism enables the same hardware system to flexibly adapt to diverse task requirements. Whether performing high-speed logistics delivery, low-speed fine mapping, or navigating complex urban canyons, the system can automatically adjust its perception strategy, demonstrating stronger task robustness. The solution designs independent and differentiated adjustment formulas for image and depth data, acknowledging the different roles of the two types of data in environmental perception (semantic information vs. geometric information), achieving more professional and efficient collaborative perception.

[0030] In another embodiment, a short-term driving plan is generated by integrating spatial structure information of the external driving environment and invoked driving memory data. Based on the short-term driving plan and driving experience data, a driving operation task is generated, including: when in an unfamiliar driving environment, cognitively filtering the spatial structure information of the external driving environment to identify a sub-space layout that meets preset driving space conditions; wherein, spatial structure information refers to the structural dimensions of the barrier-free space in the external driving environment; generating a short-term driving plan based on the sub-space layout and invoked driving operation memory data; wherein, the short-term driving plan includes driving direction and speed adjustment operations corresponding to a preset distance traveled from the current position within the sub-space layout; when not in an unfamiliar driving environment, a short-term driving plan is generated by integrating spatial structure information of the external driving environment and invoked historical driving path data; wherein, the short-term driving plan includes driving direction and speed adjustment operations corresponding to a preset distance traveled along a historical driving path in the external driving environment; generating a driving operation task for the vehicle based on the short-term driving plan and driving experience data for the vehicle; wherein, the driving operation task includes time-related operations on at least one of the vehicle's steering wheel, accelerator, and brake.

[0031] The beneficial effect of the above embodiments is that when in an unfamiliar driving environment, it is necessary to identify and locate the available driving space from scratch for the unfamiliar driving environment. First, the structural size information of the barrier-free space in the external driving environment is cognitively filtered to obtain a subspace layout that meets the preset driving space conditions. The preset driving space conditions refer to the subspace's size being greater than the vehicle's width and / or length in a certain dimension. Then, based on the subspace layout and the driving operation memory data mentioned above, a short-term driving plan is generated, including the driving direction and driving speed adjustment operations corresponding to the vehicle's travel of a preset distance from the current position within the subspace layout. This provides guidance information for safe and collision-free driving of the vehicle in unfamiliar driving environments.

[0032] When not in an unfamiliar driving environment, the spatial structure information of the external driving environment and the historical driving path data mentioned above are mapped together to the world coordinate system where the external driving environment is located, generating a short-term driving plan that includes driving direction and driving speed adjustment operations corresponding to the preset distance traveled along the historical driving path in the external driving environment, providing guidance information for the vehicle to drive efficiently and stably in a familiar driving environment.

[0033] To enable vehicles to match their driving performance with different short-term driving plans, based on the short-term driving plans and driving experience data of the vehicle, time-related operation tasks are generated for at least one of the vehicle's steering wheel, accelerator, and brake pedals. These operation tasks address changes in the operational state of at least one of these pedals at different points in time or between different time intervals during the execution of the short-term driving plan, such as the steering wheel's direction / angle of rotation, the accelerator pedal's depressor position, and the brake pedal's depressor position. By generating these driving operation tasks for the vehicle, a reference basis is provided for subsequent adjustments to the autonomous driving control, achieving feedback-based closed-loop adjustment control for autonomous driving.

[0034] In another embodiment, vehicle motion data during the execution of a driving operation task is sensed, and navigation neurons are used to analyze the vehicle motion data to obtain vehicle deviation information. Based on the vehicle deviation information, a driving change decision is generated to change the driving state online. This includes: sensing vehicle direction data and vehicle speed data during the execution of the driving operation task; inputting the vehicle direction data and vehicle speed data into the grid cells and head orientation cells of the navigation neurons for analysis to obtain a vehicle direction deviation sequence and a vehicle speed deviation sequence during the driving process; wherein, the vehicle direction deviation sequence includes the deviation angle between the actual vehicle direction and the planned driving direction in several consecutive unit time intervals; the vehicle speed deviation sequence includes the deviation speed between the actual vehicle speed and the planned driving speed in several consecutive unit time intervals; a driving change decision is generated based on the vehicle direction deviation sequence and the vehicle speed deviation sequence; wherein, the driving change decision includes the timing of the driving change and the content of the driving change action; based on the driving change decision, the action of at least one of the steering wheel, accelerator, and brake is changed online according to the driving change timing.

[0035] The beneficial effects of the above embodiments are that, during autonomous driving, the vehicle's actual driving direction and speed will inevitably deviate due to its own mechanical performance and / or airflow disturbances in the external driving environment, making it impossible to obtain the expected driving direction and speed during the execution of the driving operation task. To accurately adjust the autonomous driving feedback, the (actual) driving direction data and (actual) driving speed data during the execution of the driving operation task are first sensed. These data are then input into the grid cells and head-oriented cells of the navigation neurons for analysis, respectively, to obtain the driving direction deviation sequence and driving speed deviation sequence during driving, thereby characterizing the vehicle's driving direction and speed deviation at the time level during autonomous driving. Furthermore, based on the driving direction and speed deviation sequences, a driving change decision, including the timing and action of driving change, is generated. This provides navigation for online feedback and adjustment of autonomous driving, thereby changing the action of at least one of the steering wheel, accelerator, and brake online at the corresponding driving change timing, ensuring the robustness and accuracy of autonomous driving control and maintaining good autonomous control performance.

[0036] Please refer to Figure 2. One embodiment of this application provides a brain-like perception-based autonomous driving control system. This brain-like perception-based autonomous driving control system includes: a driving environment recognition module, used to sense external driving environment data, perform visual neuron analysis on the external driving environment data, and determine whether the driving environment is unfamiliar; a driving memory data retrieval module, used to adjust the retrieval of driving memory data by memory neurons based on the judgment result of the external driving environment; a driving plan generation module, used to fuse the spatial structure information of the external driving environment and the retrieved driving memory data to generate a short-term driving plan; a driving task generation module, used to generate driving operation tasks based on the short-term driving plan and driving experience data; a vehicle deviation determination module, used to sense driving motion data during the execution of the driving operation task, perform navigation neuron analysis on the driving motion data, and obtain driving deviation information; and a driving change module, used to generate driving change decisions based on the driving deviation information, thereby changing the driving state online.

[0037] The beneficial effects of the above embodiments are that the brain-like perception-based unmanned driving control system utilizes brain-like perception to analyze and process driving environment data, determine the type of external driving environment, and call appropriate driving memory data to assist in meeting the needs of different driving environments; then, it integrates the spatial structure information of the external driving environment and driving memory data to generate driving plans, and combines driving experience data to generate driving operation tasks, providing effective unmanned driving references; it also analyzes the vehicle motion data during the execution of driving operation tasks to obtain driving deviation information, generate driving change decisions, meet the online feedback changes of driving status, ensure the continuity and smoothness of driving operation changes, and improve the robustness and accuracy of driving control.

[0038] In another embodiment, the driving environment recognition module is used to perceive external driving environment data, perform visual neuron analysis on the external driving environment data, and determine whether the driving environment is unfamiliar. This includes: perceiving external driving environment image data and external driving environment depth data; performing a first analysis of the external driving environment image data using visual neurons to obtain the object contours and color features of the external driving environment; performing a second analysis of the external driving environment depth data using visual neurons to obtain the object spatial depth features of the external driving environment; fusing the object contour and color features and the object spatial depth features to obtain an object visual perception feature set; and comparing the object visual perception feature set with historical driving environment memory data. The system obtains the object structure similarity between the external driving environment and the historical driving environment to determine whether the driver is in an unfamiliar driving environment. The driving memory data retrieval module is used to adjust the driving memory data retrieval of the memory neurons based on the judgment result of the external driving environment. This includes: when the driver is in an unfamiliar driving environment, adjusting the memory neurons to retrieve driving operation memory data based on the actual available driving space structure of the external driving environment; where the actual available driving space structure refers to the space structure of the external driving environment that allows the vehicle to drive through in terms of its internal dimensions; when the driver is not in an unfamiliar driving environment, adjusting the memory neurons to retrieve historical driving path data that matches the actual location of the external driving environment.

[0039] The beneficial effects of the above embodiments are that the external driving environment is the primary problem faced by autonomous driving. Considering the complex spatial and temporal relationships between people and objects in the external driving environment, accurately identifying the dynamic states of people and objects in the external driving environment plays a crucial role in the timeliness and accuracy of autonomous driving responses. Furthermore, autonomous driving generates multimodal historical data on environments it has previously traversed, which not only aids in self-learning but also forms effective driving memory experience, providing an accurate basis for autonomous control and adaptation to the driving environment. The actual driving situations faced by autonomous driving differ between unfamiliar and familiar driving environments (i.e., not unfamiliar driving environments). Generally, autonomous driving faces fewer problems in familiar driving environments and has more historical driving memory data for reference. To improve the control effectiveness of autonomous driving in different types of driving environments, it is necessary to distinguish in advance whether the current driving environment is unfamiliar or familiar.

[0040] Considering that autonomous vehicles collect spatial visual data of the external driving environment, such as roads and buildings, during operation, this data is not only used in the current autonomous driving control process but also stored for self-learning. By comparing the current external driving environment with its stored spatial visual data, the type of external driving environment can be determined. Specifically, camera sensors and depth vision sensors (such as sensors based on binocular vision for depth detection) can be used to perceive external driving environment image data and external driving environment depth data, respectively. The external driving environment image data can be, but is not limited to, RGB images of the external driving environment, and the external driving environment depth data can be, but is not limited to, spatial depth values ​​of all objects in the external driving environment viewed from the vehicle's perspective. The system analyzes external driving environment image data to obtain object contour and color features, and analyzes external driving environment depth data to obtain object spatial depth features (i.e., spatial depth value of each object). Then, using each object as a reference, it fuses the object contour and color features and object spatial depth features corresponding to each object to obtain the object visual perception feature set corresponding to all objects in the external driving environment. It can be understood that the system extracts the contour, color, spatial depth and other multimodal features of each object from the above object visual perception feature set to meet the needs of subsequent judgment of the type of external driving environment.

[0041] Furthermore, historical driving environment memory data is extracted from the vehicle's database. This historical driving environment memory data may include, but is not limited to, records of the visual and spatial states of objects in the driving environment during the vehicle's historical driving process. By comparing the object visual perception feature set with the historical driving environment memory data, the object structure similarity between the external driving environment and the historical driving environment (i.e., the similarity in outline, color, and depth of field between different objects) is obtained. If the total number of objects whose object structure similarity between the external driving environment and the historical driving environment is greater than a preset similarity threshold exceeds a preset number threshold, the external driving environment is determined to be a familiar driving environment (i.e., not an unfamiliar driving environment); otherwise, the external driving environment is determined to be a not unfamiliar driving environment.

[0042] When in an unfamiliar driving environment, the memory neurons retrieve driving operation memory data based on the spatial structure of the external driving environment, which allows the vehicle to traverse the area. This driving operation memory data may include, but is not limited to, historical driving operation memory data corresponding to other spatial structures that are the same as or similar to the aforementioned spatial structures during previous driving experiences. When not in an unfamiliar driving environment, the memory neurons retrieve historical driving path data matching the actual location of the external driving environment. This historical driving path data refers to the historical driving path data matched when the vehicle traversed the aforementioned actual location during previous driving experiences in a familiar driving environment. By distinguishing whether the external driving environment is unfamiliar and retrieving the appropriate type of driving memory data, a reliable basis for forming short-term driving plans can be provided.

[0043] In another embodiment, sensing external driving environment image data and external driving environment depth data includes: real-time acquisition of the current drone's real-time flight speed; comparing the current drone's real-time flight speed with a preset flight speed threshold; if the current drone's real-time flight speed is lower than the preset flight speed threshold, then no adjustment is made to the data sensing frequency of the current external driving environment image data and external driving environment depth data; if the current drone's real-time flight speed is not lower than the preset flight speed threshold, then the number of effective obstacles within the drone's field of vision and the average flight speed along the current flight path are retrieved; the data sensing frequency of the external driving environment image data is adjusted using the current drone's real-time flight speed combined with the number of effective obstacles within the drone's field of vision, wherein the adjusted external driving environment image data sensing frequency is obtained using the following formula: , where f ing This indicates the perception frequency of the adjusted external driving environment image data; f 0，ing α represents the perception frequency of the external driving environment image data before adjustment; α represents the vehicle speed sensitivity coefficient of the image data, ranging from 1.3 to 1.8; v represents the current flight speed of the drone; v0 represents the preset flight speed threshold value of the image data; β represents the obstacle sensitivity coefficient of the image data, ranging from 0.3 to 1.1; C m Indicates the number of effective obstacles within the current drone's field of view; C obs k represents the maximum number of identifiable obstacles in the image. ing The image environment complexity correction coefficient is represented by the following formula: , where k ing，base represents the basic complexity coefficient, with a value of 0.8; w1, w2, and w3 represent the weight values ​​corresponding to obstacle density, illumination variation rate, and lane line sharpness in the image; k ing，gainThis represents the complexity gain coefficient, with a value of 0.17; ρ obs ΔL represents the obstacle density in the normalized image; ΔL represents the rate of change of illumination after normalization; S lane The normalized lane line ambiguity is represented by the following formula: The data perception frequency of the external driving environment depth data is adjusted using the real-time flight speed of the UAV combined with the average flight speed along the current flight path. , where f depth This indicates the perception frequency of the adjusted external driving environment depth data; f 0，depth The sensor frequency of the external driving environment depth data before adjustment; v represents the current flight speed of the drone; v min represents the minimum effective flight speed for depth data; r represents the vehicle speed exponent coefficient for depth data, ranging from 0.4 to 0.8; D ef Indicates the effective depth of field range of the current scene; D ref Indicates the reference depth of field range; v avg This represents the average flight speed corresponding to the current flight path; v avg，max k represents the maximum permissible average flight speed corresponding to the current flight path. depth This represents the depth-of-field environment complexity correction coefficient, which is obtained using the following formula: , where k depth，base This represents the basic coefficient for depth of field complexity, with a value of 0.7; k depth，gain This represents the depth-of-field complexity gain coefficient, with a value of 0.15; ρ h The distance abrupt change density is represented by the normalized value (core factor: the number of points with "distance value abrupt changes" per unit distance in the depth map, reflecting obstacle edges, road surface steps, etc.); h represents the normalized road surface undulation rate; Z represents the normalized depth confidence inverse value; x represents the nonlinear coupling index, with a value range of 0.9-1.3; the external driving environment image data and external driving environment depth data are perceptually manipulated according to the perception frequency of the adjusted external driving environment image data and external driving environment depth data.

[0044] The beneficial effects of the above embodiments are as follows: Traditional systems typically collect data at a fixed frequency, resulting in significant waste of computing power and energy. The core effect of this solution is that the system no longer simply collects data at high speed continuously, but uses the real-time flight speed as the core state variable to trigger frequency adjustment. When the speed is below a threshold, the base frequency is maintained to save resources. During high-speed flight, the system can intelligently allocate limited computing resources to the most needed moments, increasing the sensing frequency to acquire denser environmental information, thereby ensuring safety at critical moments and maximizing computing efficiency. The above solution does not rely solely on speed but deeply integrates environmental characteristics, enabling the sensing system to "understand the complexity of the scene."

[0045] On the one hand, for image data, frequency adjustment is not only related to speed, but also strongly coupled with the number of effective obstacles and the complexity of the image environment. In complex scenes with dense obstacles, abrupt changes in lighting, and blurred lane lines, the system automatically improves its "visual alertness" to capture details at a higher frequency, effectively addressing long-tail problems such as "ghosting," glare from strong light, and worn lane lines. For depth data, frequency adjustment is related to the effective depth range, average path speed, and the complexity of the depth environment. This allows the system to prioritize the modeling accuracy within the current flight space. For example, in complex terrain with numerous obstacle edges (high density of abrupt changes in distance) or drastic road surface undulations, the system proactively increases the scanning frequency to build a more real-time and accurate 3D environment map, providing high-quality data for obstacle avoidance and path planning. During high-speed flight, any perception delay can lead to catastrophic consequences. This solution specifically enhances the system's safety margin. By introducing exponential coefficients (such as α and r), the perception frequency increases superlinearly with speed. This ensures that the system can obtain a far greater flow of perception data than usual at extremely high speeds, significantly shortening the latency of the perception-decision-control closed loop and providing a valuable time window for avoiding dynamic obstacles. Introducing the average flight speed along the current flight path as one of the adjustment factors for the depth-of-field frequency gives the system "path pre-cognition" capabilities. When the system anticipates entering a path that typically requires high-speed flight, it will increase the depth-of-field perception frequency in advance or simultaneously. For drones, energy consumption is life. This solution, through intelligent adjustment, brings direct benefits to battery life.

[0046] On the other hand, sensors and computing units are major power consumers on drones. By automatically reducing the sensing frequency in low-risk scenarios (low speed, simple environment), the system can significantly reduce unnecessary power consumption, thereby effectively extending the flight time per flight. Reducing the computational load also means reducing processor heat generation, which helps maintain the system operating at a better temperature and improves the reliability and lifespan of electronic components.

[0047] This adaptive mechanism enables the same hardware system to flexibly adapt to diverse task requirements. Whether performing high-speed logistics delivery, low-speed fine mapping, or navigating complex urban canyons, the system can automatically adjust its perception strategy, demonstrating stronger task robustness. The solution designs independent and differentiated adjustment formulas for image and depth data, acknowledging the different roles of the two types of data in environmental perception (semantic information vs. geometric information), achieving more professional and efficient collaborative perception.

[0048] In another embodiment, the driving plan generation module is used to integrate the spatial structure information of the external driving environment and the driving memory data to generate a short-term driving plan, including: when in an unfamiliar driving environment, cognitively filtering the spatial structure information of the external driving environment and identifying a sub-space layout that meets preset driving space conditions; wherein, the spatial structure information refers to the structural dimensions of the barrier-free space in the external driving environment; generating a short-term driving plan based on the sub-space layout and the driving operation memory data; wherein, the short-term driving plan includes driving direction and driving speed adjustment operations corresponding to a preset distance traveled from the current position within the sub-space layout; when not in an unfamiliar driving environment, integrating the spatial structure information of the external driving environment and the historical driving path data to generate a short-term driving plan; wherein, the short-term driving plan includes driving direction and driving speed adjustment operations corresponding to a preset distance traveled along the historical driving path in the external driving environment; the driving task generation module is used to generate driving operation tasks based on the short-term driving plan and driving experience data, including: generating driving operation tasks for the vehicle based on the short-term driving plan and driving experience data of the vehicle; wherein, the driving operation tasks include time-related operation tasks of at least one of the steering wheel, accelerator, and brake of the vehicle.

[0049] The beneficial effect of the above embodiments is that when in an unfamiliar driving environment, it is necessary to identify and locate the available driving space from scratch for the unfamiliar driving environment. First, the structural size information of the barrier-free space in the external driving environment is cognitively filtered to obtain a subspace layout that meets the preset driving space conditions. The preset driving space conditions refer to the subspace's size being greater than the vehicle's width and / or length in a certain dimension. Then, based on the subspace layout and the driving operation memory data mentioned above, a short-term driving plan is generated, including the driving direction and driving speed adjustment operations corresponding to the vehicle's travel of a preset distance from the current position within the subspace layout. This provides guidance information for safe and collision-free driving of the vehicle in unfamiliar driving environments.

[0050] When not in an unfamiliar driving environment, the spatial structure information of the external driving environment and the historical driving path data mentioned above are mapped together to the world coordinate system where the external driving environment is located, generating a short-term driving plan that includes driving direction and driving speed adjustment operations corresponding to the preset distance traveled along the historical driving path in the external driving environment, providing guidance information for the vehicle to drive efficiently and stably in a familiar driving environment.

[0051] To enable vehicles to match their driving performance with different short-term driving plans, based on the short-term driving plans and driving experience data of the vehicle, time-related operation tasks are generated for at least one of the vehicle's steering wheel, accelerator, and brake pedals. These operation tasks address changes in the operational state of at least one of these pedals at different points in time or between different time intervals during the execution of the short-term driving plan, such as the steering wheel's direction / angle of rotation, the accelerator pedal's depressor position, and the brake pedal's depressor position. By generating these driving operation tasks for the vehicle, a reference basis is provided for subsequent adjustments to the autonomous driving control, achieving feedback-based closed-loop adjustment control for autonomous driving.

[0052] In another embodiment, the vehicle deviation determination module is used to sense vehicle motion data during the execution of the driving operation task, and perform navigation neuron analysis on the vehicle motion data to obtain vehicle deviation information, including: sensing vehicle direction data and vehicle speed data during the execution of the driving operation task; inputting the vehicle direction data and vehicle speed data into the grid cells and head orientation cells of the navigation neuron for analysis, respectively, to obtain the vehicle direction deviation sequence and vehicle speed deviation sequence during the driving process; wherein, the vehicle direction deviation sequence includes the deviation angle between the actual driving direction and the planned driving direction in several consecutive unit time intervals; the vehicle speed deviation sequence includes the deviation speed between the actual driving speed and the planned driving speed in several consecutive unit time intervals; the driving change module is used to generate driving change decisions based on the vehicle deviation information, thereby changing the driving state online, including: generating driving change decisions based on the vehicle direction deviation sequence and the vehicle speed deviation sequence; wherein, the driving change decision includes the timing of the driving change and the content of the driving change action; according to the driving change decision, changing the action of at least one of the steering wheel, accelerator, and brake online according to the driving change timing.

[0053] The beneficial effects of the above embodiments are that, during autonomous driving, the vehicle's actual driving direction and speed will inevitably deviate due to its own mechanical performance and / or airflow disturbances in the external driving environment, making it impossible to obtain the expected driving direction and speed during the execution of the driving operation task. To accurately adjust the autonomous driving feedback, the (actual) driving direction data and (actual) driving speed data during the execution of the driving operation task are first sensed. These data are then input into the grid cells and head-oriented cells of the navigation neurons for analysis, respectively, to obtain the driving direction deviation sequence and driving speed deviation sequence during driving, thereby characterizing the vehicle's driving direction and speed deviation at the time level during autonomous driving. Furthermore, based on the driving direction and speed deviation sequences, a driving change decision, including the timing and action of driving change, is generated. This provides navigation for online feedback and adjustment of autonomous driving, thereby changing the action of at least one of the steering wheel, accelerator, and brake online at the corresponding driving change timing, ensuring the robustness and accuracy of autonomous driving control and maintaining good autonomous control performance.

[0054] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. This disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. An autonomous driving control method based on brain-like perception, characterized in that, include: The system senses external driving environment data and performs visual neuron analysis on the external driving environment data to determine whether the driver is in an unfamiliar driving environment. Based on the assessment of the external driving environment, the system adjusts the retrieval of driving memory data by memory neurons; it integrates the spatial structure information of the external driving environment with the retrieved driving memory data to generate a short-term driving plan; based on the short-term driving plan and driving experience data, it generates a driving operation task; it senses the vehicle motion data during the execution of the driving operation task, performs navigation neuron analysis on the vehicle motion data to obtain driving deviation information; and based on the driving deviation information, it generates a driving change decision to change the driving state online.

2. The autonomous driving control method based on brain-like perception as described in claim 1, characterized in that: The system perceives external driving environment data and performs visual neuron analysis on this data to determine if the driving environment is unfamiliar. Based on the determination, it adjusts the memory neurons' retrieval of driving memory data, including: perceived external driving environment image data and external driving environment depth data. The system performs a first analysis of the external driving environment image data using visual neurons to obtain the object contours and color features of the external driving environment. It performs a second analysis of the external driving environment depth data using visual neurons to obtain the object spatial depth features of the external driving environment. It fuses the object contour and color features with the object spatial depth features to obtain an object visual perception feature set. It compares the object visual perception feature set with historical driving environment memory data to obtain the object structure similarity between the external driving environment and the historical driving environment, thereby determining if the driving environment is unfamiliar. If the driving environment is unfamiliar, the system adjusts the memory neurons' retrieval of driving operation memory data based on the actual usable driving space structure of the external driving environment. The actual usable driving space structure refers to the spatial structure of the external driving environment that allows the vehicle to travel through in terms of its internal dimensions. If the driving environment is not unfamiliar, the system adjusts the memory neurons' retrieval of historical driving path data that matches the actual location of the external driving environment.

3. The autonomous driving control method based on brain-like perception as described in claim 1, characterized in that: The perception of external driving environment image data and external driving environment depth data includes: real-time acquisition of the current UAV's real-time flight speed; comparison of the current UAV's real-time flight speed with a preset flight speed threshold; if the current UAV's real-time flight speed is lower than the preset flight speed threshold, no adjustment is made to the data perception frequency of the current perception of external driving environment image data and external driving environment depth data; if the current UAV's real-time flight speed is not lower than the preset flight speed threshold, the number of effective obstacles within the current UAV's field of vision and the average flight speed along the current flight path are retrieved; the data perception frequency of the external driving environment image data is adjusted using the current UAV's real-time flight speed combined with the number of effective obstacles within the current UAV's field of vision, wherein the adjusted perception frequency of the external driving environment image data is obtained using the following formula: , where f ing This indicates the perception frequency of the adjusted external driving environment image data; f 0，ing α represents the perception frequency of the external driving environment image data before adjustment; α represents the vehicle speed sensitivity coefficient of the image data, ranging from 1.3 to 1.8; v represents the current flight speed of the drone; v0 represents the preset flight speed threshold value of the image data; β represents the obstacle sensitivity coefficient of the image data, ranging from 0.3 to 1.1; C m Indicates the number of effective obstacles within the current drone's field of view; C obs k represents the maximum number of identifiable obstacles in the image. ing The image environment complexity correction coefficient is represented by the following formula: , where k ing，base represents the basic complexity coefficient, with a value of 0.8; w1, w2, and w3 represent the weight values ​​corresponding to obstacle density, illumination variation rate, and lane line sharpness in the image; k ing，gain This represents the complexity gain coefficient, with a value of 0.17; ρ obs ΔL represents the obstacle density in the normalized image; ΔL represents the rate of change of illumination after normalization; S lane The normalized lane line ambiguity is represented by the following formula: The data perception frequency of the external driving environment depth data is adjusted using the real-time flight speed of the UAV combined with the average flight speed along the current flight path. , where f depth This indicates the perception frequency of the adjusted external driving environment depth data; f 0，depth The sensor frequency of the external driving environment depth data before adjustment; v represents the current flight speed of the drone; v min represents the minimum effective flight speed for depth data; r represents the vehicle speed exponent coefficient for depth data, ranging from 0.4 to 0.8; D ef Indicates the effective depth of field range of the current scene; D ref Indicates the reference depth of field range; v avg This represents the average flight speed corresponding to the current flight path; v avg，max k represents the maximum permissible average flight speed corresponding to the current flight path. depth This represents the depth-of-field environment complexity correction coefficient, which is obtained using the following formula: , where k depth，base This represents the basic coefficient for depth of field complexity, with a value of 0.7; k depth，gain This represents the depth-of-field complexity gain coefficient, with a value of 0.15; ρ h denoted by , h represents the distance abrupt change point density after normalization; denoted by h represents the road surface undulation rate after normalization; denoted by Z represents the depth confidence inverse value after normalization; denoted by x represents the nonlinear coupling index, with a value range of 0.9-1.3; the external driving environment image data and external driving environment depth data are perceptually manipulated according to the perception frequency of the adjusted external driving environment image data and external driving environment depth data.

4. The autonomous driving control method based on brain-like perception as described in claim 2, characterized in that: By integrating spatial structure information of the external driving environment and driving memory data, a short-term driving plan is generated; Based on the short-term driving plan and driving experience data, a driving operation task is generated, including: when in an unfamiliar driving environment, cognitively filtering the spatial structure information of the external driving environment and identifying a sub-space layout that meets preset driving space conditions; wherein, the spatial structure information refers to the structural dimensions of the barrier-free space in the external driving environment; generating a short-term driving plan based on the sub-space layout and the retrieved driving operation memory data; wherein, the short-term driving plan includes driving direction and speed adjustment operations corresponding to a preset distance traveled from the current position within the sub-space layout; when not in an unfamiliar driving environment, integrating the spatial structure information of the external driving environment and the retrieved historical driving path data to generate a short-term driving plan; wherein, the short-term driving plan includes driving direction and speed adjustment operations corresponding to a preset distance traveled along the historical driving path in the external driving environment; and generating a driving operation task for the vehicle based on the short-term driving plan and driving experience data of the vehicle; wherein, the driving operation task includes time-related operations on at least one of the vehicle's steering wheel, accelerator, and brake.

5. The autonomous driving control method based on brain-like perception as described in claim 4, characterized in that: The system senses vehicle motion data during the execution of the driving operation task, performs navigation neuron analysis on the vehicle motion data, and obtains vehicle deviation information. Based on the vehicle deviation information, a driving change decision is generated to change the driving state online. This includes: sensing the driving direction data and driving speed data during the execution of the driving operation task; inputting the driving direction data and driving speed data into the grid cells and head orientation cells of the navigation neuron for analysis, respectively, to obtain the driving direction deviation sequence and driving speed deviation sequence during the driving process; wherein, the driving direction deviation sequence includes the deviation angle between the actual driving direction and the planned driving direction in several consecutive unit time intervals; the driving speed deviation sequence includes the deviation speed between the actual driving speed and the planned driving speed in several consecutive unit time intervals; generating a driving change decision based on the driving direction deviation sequence and the driving speed deviation sequence; wherein, the driving change decision includes the timing of the driving change and the content of the driving change action; and, based on the driving change decision, changing the action of at least one of the steering wheel, accelerator, and brake online according to the driving change timing.

6. An unmanned driving control system based on brain-like perception, characterized in that, include: The driving environment recognition module is used to sense external driving environment data, perform visual neuron analysis on the external driving environment data, and determine whether the driver is in an unfamiliar driving environment. The driving memory data retrieval module is used to adjust the driving memory data retrieval of memory neurons based on the judgment results of the external driving environment. The driving plan generation module is used to integrate the spatial structure information of the external driving environment and the driving memory data to generate a short-term driving plan; the driving task generation module is used to generate driving operation tasks based on the short-term driving plan and driving experience data. The vehicle deviation determination module is used to sense the vehicle motion data during the execution of the driving operation task, perform navigation neuron analysis on the vehicle motion data, and obtain vehicle deviation information; the vehicle deviation modification module is used to generate driving change decisions based on the vehicle deviation information, thereby changing the driving state online.

7. The unmanned driving control system based on brain-like perception as described in claim 6, characterized in that: The driving environment recognition module is used to perceive external driving environment data, perform visual neuron analysis on the external driving environment data, and determine whether the driving environment is unfamiliar. This includes: perceiving external driving environment image data and external driving environment depth data; performing a first analysis of the external driving environment image data using visual neurons to obtain the object contours and color features of the external driving environment; performing a second analysis of the external driving environment depth data using visual neurons to obtain the object spatial depth features of the external driving environment; fusing the object contour and color features and the object spatial depth features to obtain an object visual perception feature set; and comparing the object visual perception feature set with the historical driving environment memory. The system obtains the object structure similarity between the external driving environment and the historical driving environment to determine whether the driver is in an unfamiliar driving environment. The driving memory data retrieval module is used to adjust the driving memory data retrieval of the memory neurons based on the judgment result of the external driving environment. This includes: when in an unfamiliar driving environment, adjusting the memory neurons to retrieve driving operation memory data based on the actual available driving space structure of the external driving environment; wherein, the actual available driving space structure refers to the space structure of the external driving environment that allows the vehicle to drive through in terms of its internal dimensions; when not in an unfamiliar driving environment, adjusting the memory neurons to retrieve historical driving path data that matches the actual location of the external driving environment.

8. The unmanned driving control system based on brain-like perception as described in claim 7, characterized in that: The perception of external driving environment image data and external driving environment depth data includes: real-time acquisition of the current UAV's real-time flight speed; comparison of the current UAV's real-time flight speed with a preset flight speed threshold; if the current UAV's real-time flight speed is lower than the preset flight speed threshold, no adjustment is made to the data perception frequency of the current perception of external driving environment image data and external driving environment depth data; if the current UAV's real-time flight speed is not lower than the preset flight speed threshold, the number of effective obstacles within the current UAV's field of vision and the average flight speed along the current flight path are retrieved; the data perception frequency of the external driving environment image data is adjusted using the current UAV's real-time flight speed combined with the number of effective obstacles within the current UAV's field of vision, wherein the adjusted perception frequency of the external driving environment image data is obtained using the following formula: , where f ing This indicates the perception frequency of the adjusted external driving environment image data; f 0，ing α represents the perception frequency of the external driving environment image data before adjustment; α represents the vehicle speed sensitivity coefficient of the image data, ranging from 1.3 to 1.8; v represents the current flight speed of the drone; v0 represents the preset flight speed threshold value of the image data; β represents the obstacle sensitivity coefficient of the image data, ranging from 0.3 to 1.1; C m Indicates the number of effective obstacles within the current drone's field of view; C obs k represents the maximum number of identifiable obstacles in the image. ing The image environment complexity correction coefficient is represented by the following formula: , where k ing，base represents the basic complexity coefficient, with a value of 0.8; w1, w2, and w3 represent the weight values ​​corresponding to obstacle density, illumination variation rate, and lane line sharpness in the image; k ing，gain This represents the complexity gain coefficient, with a value of 0.17; ρ obs ΔL represents the obstacle density in the normalized image; ΔL represents the rate of change of illumination after normalization; S lane The normalized lane line ambiguity is represented by the following formula: The data perception frequency of the external driving environment depth data is adjusted using the real-time flight speed of the UAV combined with the average flight speed along the current flight path. , where f depth This indicates the perception frequency of the adjusted external driving environment depth data; f 0，depth The sensor frequency of the external driving environment depth data before adjustment; v represents the current flight speed of the drone; v min represents the minimum effective flight speed for depth data; r represents the vehicle speed exponent coefficient for depth data, ranging from 0.4 to 0.8; D ef Indicates the effective depth of field range of the current scene; D ref Indicates the reference depth of field range; v avg This represents the average flight speed corresponding to the current flight path; v avg，max k represents the maximum permissible average flight speed corresponding to the current flight path. depth This represents the depth-of-field environment complexity correction coefficient, which is obtained using the following formula: , where k depth，base This represents the basic coefficient for depth of field complexity, with a value of 0.7; k depth，gain This represents the depth-of-field complexity gain coefficient, with a value of 0.15; ρ h denoted by , h represents the distance abrupt change point density after normalization; denoted by h represents the road surface undulation rate after normalization; denoted by Z represents the depth confidence inverse value after normalization; denoted by x represents the nonlinear coupling index, with a value range of 0.9-1.3; the external driving environment image data and external driving environment depth data are perceptually manipulated according to the perception frequency of the adjusted external driving environment image data and external driving environment depth data.

9. The unmanned driving control system based on brain-like perception as described in claim 6, characterized in that: The driving planning generation module is used to integrate the spatial structure information of the external driving environment and the retrieved driving memory data to generate a short-term driving plan, including: when in an unfamiliar driving environment, cognitively filtering the spatial structure information of the external driving environment and identifying a sub-space layout that meets preset driving space conditions; wherein, the spatial structure information refers to the structural dimensions of the barrier-free space in the external driving environment; generating a short-term driving plan based on the sub-space layout and the retrieved driving operation memory data; wherein, the short-term driving plan includes driving direction and driving speed adjustments corresponding to a preset distance traveled from the current position within the sub-space layout; when not in an unfamiliar driving environment... When in an unfamiliar driving environment, a short-term driving plan is generated by integrating the spatial structure information of the external driving environment and the historical driving path data. The short-term driving plan includes adjustments to the driving direction and speed corresponding to a preset distance traveled along the historical driving path in the external driving environment. The driving task generation module generates driving operation tasks based on the short-term driving plan and driving experience data, including: generating driving operation tasks for the vehicle based on the short-term driving plan and driving experience data for the vehicle itself; wherein the driving operation tasks include time-related operations on at least one of the vehicle's steering wheel, accelerator, and brake.

10. The unmanned driving control system based on brain-like perception as described in claim 6, characterized in that: The vehicle deviation determination module is used to sense vehicle motion data during the execution of the driving operation task, and perform navigation neuron analysis on the vehicle motion data to obtain vehicle deviation information, including: sensing vehicle direction data and vehicle speed data during the execution of the driving operation task; inputting the vehicle direction data and vehicle speed data into the grid cells and head orientation cells of the navigation neuron for analysis, respectively, to obtain a vehicle direction deviation sequence and a vehicle speed deviation sequence during driving; wherein, the vehicle direction deviation sequence includes the deviation angle between the actual vehicle direction and the planned driving direction in several consecutive unit time intervals; the vehicle speed deviation sequence includes the deviation speed between the actual vehicle speed and the planned driving speed in several consecutive unit time intervals; the driving change module is used to generate driving change decisions based on the vehicle deviation information, thereby changing the driving state online, including: generating driving change decisions based on the vehicle direction deviation sequence and the vehicle speed deviation sequence; wherein, the driving change decision includes the timing of the driving change and the content of the driving change action; according to the driving change decision, changing the action of at least one of the steering wheel, accelerator, and brake online according to the driving change timing.

Images