168results about How to "Increased complexity" patented technology

The invention relates to the technical field of artificial intelligence, in particular to a load balancing method and device, electronic equipment, a storage medium and a program product, and the load balancing method comprises the steps: obtaining traffic characteristics of communication links between different nodes in a data block migration process of a distributed storage system; determining a load distribution mode of each communication link based on the flow characteristics of each communication link; based on the flow characteristics and the load distribution mode of each communication link, determining hotspot nodes and corresponding service priorities thereof; determining a target hotspot node based on the access frequencies, the data transmission quantities, the node loads and the corresponding service priorities of all the hotspot nodes; and based on the processing performance, the network state and the storage capacity of the node, screening out a target low-load node meeting a preset condition from the distributed storage system, and migrating data of the target hotspot node to the target low-load node. According to the invention, the network flow distribution and node load balancing state can be dynamically optimized.

The application discloses a feature placement method and system for multi-GPU sampling graph neural network training, uniformly models a sampling graph neural network training process, and obtains hardware topology information; in a pre-sampling process, the number of times of accessing node features is counted as node heat, the node features are sorted according to the node heat, the node features are divided into a plurality of feature blocks according to a single-block capacity, and the feature blocks are divided into hot blocks or cold blocks according to block heat; based on the feature block set obtained by the division and the obtained hardware topology information, a mixed integer programming model is constructed, with the minimum bottleneck link time in a feature extraction stage and the memory load imbalance degree as targets, the mixed integer programming model is solved, and the placement position, access path and link flow of the globally jointly optimized hot block are obtained. The application can obtain a high-quality feature placement scheme under a given hardware constraint, minimize the extraction time, and improve the resource utilization rate of a multi-GPU system.

The invention relates to the technical field of optical anti-counterfeiting, and discloses anti-counterfeiting application of a circular polarization luminescent material. The preparation method of the circularly polarized luminescent material comprises the following steps: providing a compound CS5 shown in the specification, and the compound CS5 shows nematic phase liquid crystal properties in a temperature range of 36-66 DEG C; and adding a chiral agent into the compound CS5 with the nematic liquid crystal property, and reacting to obtain the circularly polarized luminescent material with the chiral nematic phase, . According to the chiral nematic phase circularly polarized luminescent material provided by the invention, the polarization state dimension is successfully introduced into optical anti-counterfeiting, the spanning from single luminescence verification to double verification of luminescence and polarization state switching is realized, and the technical problems of single anti-counterfeiting dimension and easy counterfeiting in the background technology are effectively solved; the method has the remarkable beneficial effects of improving the identification reliability and enhancing the anti-counterfeiting safety.

The invention provides an image sensor and a manufacturing method thereof, and belongs to the technical field of semiconductors. The image sensor comprises a pixel region and a grounding region or comprises the pixel region, the grounding region and a silicon through hole region, and the image sensor at least comprises an epitaxial layer, and a trench isolation structure and a photodiode are arranged in the epitaxial layer of the pixel region; the conductive isolation ring is arranged in the epitaxial layer, surrounds the trench isolation structure and the photodiode, and is connected to a silicon through hole contact of the silicon through hole region or a grounding layer of the grounding region; negative voltage is provided for the conductive isolating ring through contact of the silicon through hole or the grounding layer; the grating is arranged at the top of the groove isolation structure; and the light filtering structure is arranged at the top of the photodiode. According to the image sensor and the manufacturing method thereof provided by the invention, effective isolation of the pixel region can be realized.

The application belongs to the technical field of video encryption transmission, and particularly relates to a video transmission method and system for multimedia video, which comprises the following steps: identity authentication of a receiving end and a sending end; a gray image and an information sequence are randomly generated by the to-be-authenticated end, the information sequence is a binary sequence, and the number of digital 0 and 1 is equal; according to the MAC address of the to-be-authenticated end, the information sequence is hidden in the gray image through a steganography algorithm to obtain an authentication image and send the authentication image to the authentication end; the authentication end extracts an authentication sequence from the received authentication image through the steganography algorithm according to the stored MAC address of the trusted terminal, and the to-be-authenticated end passes the authentication when the authentication sequence satisfies that the number of digital 0 and 1 is equal; after the receiving end and the sending end both pass the authentication, an information key is created through a DH key exchange protocol, and the information key is used for encrypted transmission of the video. The application improves the security of video encryption transmission.

The invention discloses a fault root cause positioning method, apparatus and device, and a readable storage medium. The method comprises the steps of determining a multi-source abnormal feature; obtaining a domain knowledge graph; retrieving an analysis basis from the domain knowledge graph; performing root cause tracing by using the preliminary screening layer in combination with an analysis basis, and generating a root cause reasoning process and a reasoning complex score; when the reasoning complex score is lower than a complex threshold value, generating a root cause analysis report by using a conclusion generation layer; when the reasoning complex score is not lower than a complex threshold value, a depth reasoning layer is called to carry out depth causal reasoning, and a reasoning thought and a root cause propagation path are generated; and generating a root cause analysis report based on the reasoning thought and the root cause propagation path by using a conclusion generation layer. Visibly, according to the invention, through a hierarchical reasoning architecture and a domain knowledge graph, reasoning ideas of faults with different difficulty can be clearly presented while the reasoning efficiency and the fault root cause positioning rate are improved, and the credibility and verifiability of various fault root cause positioning conclusions are improved.

This application provides a text data processing method, apparatus, device, and medium, applicable to the field of sensitive data identification technology. The method includes: performing M character mutations on each text segment in an initial text dataset to obtain M mutated text segments corresponding to each text segment; M being a positive integer; obtaining the mutation evaluation value of the M mutated text segments corresponding to each text segment; in each of the M mutated text segments corresponding to a text segment, if the mutation evaluation value of the i-th mutated text segment is greater than or equal to an evaluation threshold, and the mutation evaluation value of the (i+1)-th mutated text segment is less than the evaluation threshold, then adding the i-th mutated text segment to the result dataset; i being a positive integer less than M; filtering the result dataset to obtain a text-enhanced dataset; the text-enhanced dataset is used to train a sensitive data identification model. Implementing this application embodiment can improve the quality and diversity of text-enhanced data.

This invention provides a Pinyin error correction method that simplifies the noise channel model used in most current Pinyin error correction algorithms by employing a real-time frequency counting method, effectively improving the efficiency and lightweighting of the algorithm. Furthermore, this invention uses a direct character substitution method instead of the traditional edit distance calculation method, avoiding frequent edit distance calculations in the Pinyin error correction algorithm. The dictionary database itself is localized to the individual user, resulting in a highly personalized, targeted, and small-scale dictionary that provides accurate candidate words. Compared to existing algorithms, this invention offers significant improvements in error checking rate, candidate word accuracy, and execution efficiency.

The utility model provides a stir-frying device for tea processing, and relates to the technical field of tea processing, the stir-frying device comprises a rack, the rack is fixedly connected with a shell, the shell is internally and rotatably connected with a frying cylinder, the shell is fixedly communicated with a heating cavity, the shell is fixedly communicated with an exhaust pipe, the frying cylinder is provided with a sieving structure, and the sieving structure is arranged on the exhaust pipe. A sieving structure is arranged in the shell and mainly comprises a plurality of sieving holes, the sieving holes are all formed in the frying cylinder, a fixed cylinder is fixedly connected in the shell, the fixed cylinder is arranged on the outer side of the frying cylinder in a sleeving manner, a discharging pipe is fixedly inserted into the heating cavity, a material guide groove is formed in the inner wall of the fixed cylinder and communicates with the discharging pipe, and a material outlet is formed in the lower end of the fixed cylinder. According to the utility model, the problem that the complexity, time cost and labor cost of the processing flow are increased due to the additional screening process since the screening process is added to remove the chippings after the stir-frying is finished at present is solved.

This invention belongs to the field of multi-objective evolutionary neural architecture search, proposing a multi-objective evolutionary neural architecture search method that integrates convolution and self-attention. By constructing an improved multi-objective evolutionary algorithm, the efficiency of neural architecture search is increased, allowing for faster discovery and preservation of individual neural architectures during the search process, while balancing convergence and diversity. By constructing a search space that integrates convolution and attention operations, the processing capability of global and local information in image processing tasks is improved, enhancing the plasticity of the search space and the generalization performance of individuals across different task sets. Decision variables such as convolution kernel size and number of layers in the search space are encoded as individuals in the population of the improved multi-objective evolutionary algorithm. The optimal individual is searched through the operation of the improved multi-objective evolutionary algorithm and decoded into the optimal network architecture. This invention effectively improves search efficiency and obtains the best-performing neural architecture.

This application provides a submersible gate mold and injection molding equipment. The submersible gate mold includes a mold core, a sprue, and at least one protruding block. The mold core has a communicating cavity and at least one gate. The sprue has at least one inclined runner, which communicates with the gate. The protruding block blocks one end of the inclined runner near the gate, and at least partially protrudes from the sprue and is located at the gate. The protruding block has an auxiliary cut-off channel, through which the gate communicates with the inclined runner. The submersible gate mold reduces mold adjustments and improves the complexity of product molding operations while ensuring less mixing debris and reducing mold wear.

The invention discloses a three-dimensional indoor positioning method based on visual sense and 5G sensing fusion, and relates to the technical field of indoor positioning and space perception. The method comprises the following steps: modeling an indoor reflecting surface to obtain a virtual anchor point; a user azimuth vector detected, identified and tracked by a camera target is projected to a scene reflection surface to construct a candidate reflection path and identify a direct incidence path, and the candidate reflection path and the direct incidence path are mapped to a beam codebook of a base station, so that a beam search range is narrowed, and a higher angle resolution is obtained; the user terminal establishes a communication link through the reference signal, and the base station acquires channel state information; in the channel parameter processing stage, angle and time delay estimation is carried out on channel parameters; an estimation result is matched with a direct or virtual anchor point path, and pseudo paths which do not conform to scene constraints are eliminated, so that a direct path and an effective reflection path are distinguished; the position of the user is jointly solved by using multi-angle constraints formed by the direct path and the reflection path, and continuous estimation of the three-dimensional position of the user is realized by combining a filter.

The embodiment of the invention discloses a cracking risk processing method and device, electronic equipment and a computer program product. Wherein the client calls a communication interface to establish at least two communication connections with the server side, calls a java layer communication interface to establish at least one type of communication connection with the server side, and calls a C / C + + language layer communication interface to establish at least one type of communication connection with the server side; respectively generating at least two messages by using the operation data of the client, and respectively sending the at least two messages to the server through the corresponding communication connections; and the server side detects whether the client side has a cracking risk or not by comparing the consistency of the operation data corresponding to the same request identifier. On one hand, the cracking difficulty of a client communication interface is reinforced through a C / C + + language layer communication interface; and on the other hand, through cross validation of the transmission layer, low-cost and high-reliability cracking risk detection is realized, so that the data authenticity is improved.

The application provides a battery group arrangement equalization strategy optimization method based on a QTA-TLBO double-layer multi-objective algorithm, which comprises the following steps: establishing an upper-layer battery group optimization model comprising an upper-layer optimization objective function and an upper-layer optimization constraint condition, and obtaining an optimal single cell arrangement structure of the battery group; establishing a lower-layer battery group optimization energy scheduling model comprising a lower-layer optimization objective function and a lower-layer optimization constraint condition, and obtaining an optimal equalization strategy of the battery group; establishing a coupling optimization model between the upper-layer battery group single cell arrangement structure and the lower-layer battery group equalization strategy; and using the proposed quantum tunneling annealing-teaching and learning double-layer multi-objective optimization algorithm to optimize and solve the model, so as to obtain an optimal grouping scheme of the battery group and an optimal equalization strategy of each cycle. Through the application, the reasonable arrangement of the single cells of the battery group can be obtained, the long-term reliable operation of the battery group can be realized, and the energy scheduling strategy of the battery group between the single cells can be planned, so that the maximization of battery energy utilization is realized.

This invention relates to the field of substation anomaly identification technology, and discloses a method and system for anomaly identification in substation switchgear. The method includes: acquiring real-time temperature data from multiple monitoring points inside the switchgear and uploading it to a big data analysis platform via wired or wireless communication to achieve continuous monitoring of equipment operating status. Based on the real-time temperature data, the highest temperature of the monitoring area, the highest temperature difference between adjacent monitoring periods, and temperature abrupt changes are calculated. When these parameters exceed preset thresholds, corresponding alarm strategies are triggered in a timely manner. By generating infrared thermal images, extracting areas of temperature distribution difference, and marking high-risk monitoring points, the accuracy and timeliness of anomaly identification are further improved. Simultaneously, a temperature trend prediction model is used to predict future temperature trends, generate early warning signals in advance, and activate a preventative alarm mechanism, providing strong protection for the safe and stable operation of the power system.

The utility model relates to the technical field of wall-mounted voltage stabilizing equipment, in particular to a wall-mounted voltage stabilizer convenient to adjust. According to the technical scheme, the wall-mounted voltage stabilizer convenient to adjust comprises a mounting base, a telescopic assembly, a heat dissipation assembly, a supporting rod, a three-axis spherical adjusting base, a connecting support, a mounting block, an environment detection probe and a voltage stabilizer body, the telescopic assembly is arranged on one side of the mounting base, and the heat dissipation assembly is arranged in the telescopic assembly; a supporting rod is arranged at the top end of the telescopic assembly, a three-axis spherical adjusting base is arranged at the top end of the supporting rod, a connecting support is arranged on one side of the three-axis spherical adjusting base, a mounting block is arranged on one side of the connecting support, and an environment detection probe is arranged on one side of the mounting block; by arranging the multi-angle adjusting structure, the orientation and the installation angle of the voltage stabilizer can be adjusted in real time in a multi-angle mode, the installation support of the voltage stabilizer does not need to be adjusted and installed frequently, and installation convenience of the voltage stabilizer is effectively improved.

The invention discloses an efficient impurity removing device for production and processing of traditional Chinese medicine decoction pieces, and belongs to the technical field of traditional Chinese medicine processing. The efficient impurity removing device is technically characterized by comprising a square mounting frame and further comprising a medicinal material cleaning mechanism, the medicinal material cleaning mechanism is arranged in a cylindrical cleaning sleeve, and the cylindrical cleaning sleeve is fixedly connected with the square mounting frame; the medicinal material cleaning mechanism comprises a first mounting assembly, a medicinal material cleaning assembly and an auxiliary assembly, one end of the first mounting assembly is fixedly connected with the cylindrical cleaning sleeve, the first mounting assembly is movably connected with the medicinal material cleaning assembly, the auxiliary assembly is connected with the first mounting assembly, and one side of the auxiliary assembly is movably connected with the inner wall of the cylindrical cleaning sleeve; the medicinal material draining mechanism is fixedly connected with the square mounting frame, one side of the medicinal material draining mechanism is provided with an auxiliary discharging pipe, the auxiliary discharging pipe is connected with a main discharging pipe, the main discharging pipe is fixedly connected with a cylindrical cleaning sleeve, and the traditional Chinese medicinal material cleaning device has the advantages that the traditional Chinese medicinal materials are fully cleaned and cannot be adhered to the inner wall of the impurity removal barrel.

This application relates to the field of Bluetooth detection technology, and in particular to a Bluetooth detection method, apparatus, electronic device, and storage medium. The method includes: upon receiving a Bluetooth detection information acquisition instruction, retrieving Bluetooth detection information corresponding to the vehicle to be detected from a preset database based on identification information, and sending the Bluetooth detection information to a target cloud to generate a Bluetooth detection result for the vehicle to be detected on the target cloud. This solves the problems in related technologies where Bluetooth detection requires front-end initiation and back-end execution, increasing the complexity of the detection process and thus increasing the probability of detection failure and reducing detection efficiency. By adding Bluetooth query and Bluetooth key loading processes to the vehicle processing module, Bluetooth query and Bluetooth key loading are completed, allowing inspectors to directly query the Bluetooth detection results of the vehicle processing module during cloud-based detection without front-end and back-end interaction, thereby improving Bluetooth detection efficiency.

The invention belongs to the technical field of power systems, and particularly discloses a singular perturbation-reactive decoupling-based order reduction method and device for a network construction type converter. According to the invention, the original state space model is converted; on the basis of a singular perturbation strategy, order reduction processing is carried out according to target variable characteristics; order reduction processing is carried out on a voltage dynamic equation in a reactive power control outer ring, reactive power decoupling is carried out in the mode that a voltage state variable is introduced into an active power control state equation, and a differential equation based on the phase angle relation is established. Through the above mode, aiming at the electromagnetic dynamic process of the fast subsystem, order reduction processing is carried out on the electromagnetic dynamic state variable based on the singular perturbation strategy, order reduction processing is carried out on the voltage dynamic equation in the reactive power control outer loop, high-order dynamic items are eliminated, main dynamic characteristics in the state space are reserved, the model order and the solving complexity are obviously reduced, and the reliability of the system is improved. Therefore, the model order can be systematically reduced on the premise of ensuring key dynamic characteristics and stability analysis precision.

The invention relates to a vertically stacked red, green and blue full-color carrier chip for a microLED display panel. The vertically stacked red, green and blue full-color carrier chip comprises a temporary wafer; and a plurality of LED stacks arranged on the temporary wafer, each of the plurality of LED stacks including a light emitting portion stacked in a vertical direction by a bonding layer, the plurality of LED stacks forming a short channel in a partial region, and emitting only a specific color by current to the light emitting portion where the short channel is not formed, the short channels include a first short channel formed so as to correspond to the width of the light-emitting part, and a second short channel formed so as to penetrate through the light-emitting part.

The invention relates to the technical field of wireless communication, and provides a spread spectrum sequence generation method and device, equipment, a medium and a product, and the method comprises the steps: generating a basic sequence, and generating a password sequence or a chaos sequence; performing field extension mapping on the basic sequence to form a base sequence, and performing field extension mapping on the password sequence or the chaos sequence to form a field extension sequence; performing extension field addition operation on the base sequence and the extension field sequence to obtain an extension field addition sequence; and carrying out extension domain transformation on the extension domain addition sequence to obtain a spread spectrum sequence. According to the method, the spread spectrum sequence meeting the requirement can be generated, and the method has good correlation, long periodicity, good randomness and high safety.

The invention provides a user intention recognition method and system, a medium and a terminal, and the method comprises the steps: obtaining a user query text for a business trip, the types of the user query text comprise train tickets, itinerary planning, flight query, hotel reservation, menu jump, order query, artificial customer service, multi-scene combination, multi-itinerary combination and other intentions; the user query text is preprocessed; performing feature extraction on the preprocessed user query text to obtain query semantic features; and training a user intention recognition model based on the query semantic features so as to obtain the user intention of the user based on the trained user intention recognition model. According to the user intention recognition method and system, the medium and the terminal, data enhancement of complex semantic scenes such as a multi-intention combination scene, a negative-positive scene and a context-dependent scene is supported, the accuracy of user intention recognition is effectively improved, and good field applicability and complex semantic understanding ability are achieved.

This invention discloses a strain of Maggi-Megaki yeast with amine-reducing and flavor-enhancing capabilities and its applications, belonging to the field of microbial fermentation technology. This invention screened a strain of Maggi-Megaki yeast Y21 from fermented soybean paste, which has the ability to degrade biogenic amines. This strain can completely degrade tryptamine, phenylethylamine, putrescine, cadaverine, spermidine, and spermine; simultaneously, it also has a strong degradation effect on histamine and tyramine, with degradation rates reaching 69.58% and 65.32%, respectively. Using the Maggi-Megaki yeast Y21 described in this invention for inoculation and fermentation can significantly reduce the content of various biogenic amines in fermented soybean paste under low-to-medium salt conditions, especially showing outstanding removal effects on histamine, tyramine, phenylethylamine, and polyamine biogenic amines, increasing the total amount of volatile flavor compounds, and is suitable for the safe production of low-salt fermented foods.

The application discloses a kind of low-speed fixed-wing unmanned aerial vehicle longitudinal controllable field and stability analysis method, comprising: determining the desired working range of angle of attack and track inclination state;Angle of attack and track inclination expected value are substituted into trim calculation, and the trim elevator deflection, trim airspeed and trim thrust are obtained;With angle of attack as horizontal axis, track inclination as vertical axis, the trim elevator deflection, trim airspeed and trim thrust are expressed in the form of contour distribution, and the full-state trim contour distribution diagram is obtained;Introduce aircraft performance constraints and control constraints in the full-state trim contour distribution diagram, and obtain the controllable field;Select the desired working point in the controllable field, calculate the speed, angle of attack trend in the neighborhood of the working point, obtain the speed angle of attack phase diagram, and judge the stability of the working point through the speed angle of attack phase diagram.The application is applied to the technical field of unmanned aerial vehicle, has the advantages of low complexity of trim solution, strong result display intuitiveness and high application analysis efficiency.

This invention relates to a dry ice spray cleaning and sterilizing device, aiming to solve the technical problem of the current dry ice spray cleaning and sterilizing devices having only a single spraying operation function. The device includes a cabinet; an embedded sterilization and cleaning tank is provided on the upper surface of the cabinet; the gap between the inner wall of the cabinet and the outer wall of the embedded sterilization and cleaning tank forms a multi-stage operating chamber; the embedded sterilization and cleaning tank is an electrically heated structure; and multiple cleaning components are provided inside the multi-stage operating chambers; and a dual-mode sterilization and cleaning system is provided at the output end of the multi-stage cleaning components; a transparent splash guard is hinged to the outer wall of the cabinet relative to the embedded sterilization and cleaning tank. This invention improves the cleaning effect of high-pressure dry ice spraying by adapting two operating modes to the continuous, efficient, and fully covered contact of the foam cleaner, and adapting the intermittent contact of the dry ice spraying operation to the changes in spray distance caused by angle changes during actual operation.

The application provides a power distribution network peak regulation random optimization method considering three-phase imbalance, and belongs to the technical field of power distribution network control. The method comprises the following steps: establishing a new energy power prediction error probability distribution model, generating a power scene for obtaining possible error values of new energy power, and reducing the power scene; establishing a three-phase power distribution network peak regulation optimization model based on feeder load power control; and simplifying and solving the peak regulation optimization model. The application utilizes feeder load power control technology to coordinate various voltage regulation devices to complete feeder load power control, and comprehensively solves the system peak regulation and imbalance treatment problems. The application focuses on the scene of single-phase access of high-proportion new energy, establishes a multi-objective optimization model based on reactive power voltage control, utilizes feeder load power control technology to reduce the system peak-valley difference in peak-valley time periods, utilizes load regulation characteristics and split-phase voltage optimization to relieve system imbalance in non-peak-valley time periods, and reduces the system peak-valley difference and negative sequence voltage.

The application provides a CHF prediction model determination method, a medium and an electronic device, and relates to the technical field of data processing. The method comprises the following steps: introducing a first norm penalty term and a second norm penalty term into a composite loss function simultaneously, the first norm penalty term being associated with the sum of the absolute values of each coefficient included in a coefficient set in a CHF initial prediction model, and the second norm penalty term being associated with the sum of squares of each coefficient in the coefficient set. In the process of optimizing the CHF initial prediction model with the optimization target of minimizing the composite loss function, the first norm penalty term is used to induce the sparsity of the corresponding polynomial function of the model, and tends to accurately compress the coefficients corresponding to the working condition parameters with weak contribution to the predicted CHF to zero, which is equivalent to automatically removing the redundant terms in the polynomial function. The second norm penalty is used to control the complexity of the model, prevent the value of a certain coefficient in the coefficient set from becoming too large, and effectively improve the stability and anti-overfitting ability of the model.

This specification provides a training method, text processing method, and system for a large language model, comprising: obtaining a sample set, the sample set including multiple sample texts; obtaining a first large language model to be trained, the first large language model employing a linear attention mechanism; reusing at least some model parameters of the first large language model from pre-trained model parameters of a second large language model; the second large language model employing a non-linear attention mechanism; training the first large language model using the sample set; and obtaining a trained target large language model. This method can accelerate the training process, reduce computational and time costs, and improve the performance of the target large language model. Furthermore, it can optimize the computational complexity of attention to linear levels and optimize storage space.

The invention relates to a mask, an overlay error measuring method and device and overlay equipment. The mask plate comprises a mask plate body; the plurality of overlay marks are arranged on the mask plate body, each overlay mark comprises a plurality of overlay sub-marks, and the plurality of overlay sub-marks comprise the overlay sub-marks of which the central points are positioned on the same straight line, the overlay sub-marks of which the central points are the same and the overlay sub-marks arranged in different directions. According to the mask, the overlay error measuring method and device and the overlay equipment, the diversity and complexity of overlay marks of the mask are improved, and then the precision of an overlay process based on the mask is improved.

This invention relates to an online method for generating critical boundary test scenarios for autonomous driving acceleration testing, comprising: generalizing and generating natural driving test scenarios; constructing a scenario complexity model, including static scenario complexity and dynamic scenario complexity; evaluating and testing autonomous vehicles based on the generated natural driving test scenarios and the constructed scenario complexity model; and performing scenario adaptive adjustments based on the test results, including static scenario adaptive adjustments and dynamic scenario adaptive adjustments, to output critical boundary test scenarios. Compared with existing technologies, this invention can generate online critical boundary test scenarios that satisfy the distribution and diversity of natural traffic behavior and have multiple interactions, in order to test the performance boundaries of autonomous vehicles.