77results about How to "Avoid resources" patented technology

The invention discloses a large model reasoning hardware accelerator based on data flow execution, and belongs to the technical field of large model reasoning hardware acceleration. Each node comprises a linear calculation kernel used for sequentially completing query / key / value projection and attention output projection in multi-head attention calculation, dimension raising projection and dimension reduction projection calculation in a feedforward neural network layer, and inverse quantization operation after each projection is output; the multi-head attention calculation kernel is used for completing loading and quantization of query / key / value vectors, multi-head division and cache management, attention score calculation, Softmax normalization, context vector generation and KV cache updating; the RLA kernel is used for completing residual addition, layer normalization, quantization / inverse quantization and nonlinear activation calculation of each layer; the three cores in the same node are located in the same SLR area, and data transmission is carried out through a hardware queue and an HLS blocking mechanism. According to the invention, high hardware utilization rate and low-delay reasoning can be realized.

This invention relates to the field of production management technology and discloses a method for controlling the production process of aluminum decorative locking components. The method includes: analyzing production orders to generate a process chain containing sealing, temporary storage in a humid environment, and threaded locking assembly; configuring production site resources, setting the temporary storage area as a buffer station resource with humidity indicators and capacity limitations; establishing a time mapping mechanism to determine the dynamic response relationship between the actual operation time of the locking assembly process and the waiting time after sealing and the environmental humidity index, reflecting the influence of oxide film hydration product evolution on tribological performance; and performing global production scheduling using waiting time and resource allocation as adjustment factors, outputting execution instructions containing specified temporary storage resources and assembly target cycle parameters. This invention solves the problem of cycle instability caused by tribological performance fluctuations by incorporating the evolution of material surface state into the production scheduling logic, thus achieving precise production instructions.

The application discloses an electric energy-frequency modulation out-clearing method and system considering energy storage energy state elasticity, comprising the following steps: acquiring power grid topology, load prediction data and power grid operation constraint parameters, receiving the frequency modulation market declaration information and the electric energy market declaration information submitted by market subjects; determining the winning frequency modulation capacity and the frequency modulation mileage out-clearing price of each market subject according to a preset frequency modulation out-clearing rule; taking the frequency modulation market out-clearing result as a boundary condition, constructing an electric energy market out-clearing model considering energy storage energy state elasticity, setting the state of energy (SOE) of the energy storage operator at the end time as an optimization variable, and introducing the SOE adjustment cost curve into the objective function; solving the model to determine the day-ahead time node marginal price, the day-ahead time generation power of the generation unit operator, the day-ahead time charging and discharging power of the energy storage operator and the SOE at the end time, and converting the same into a power grid dispatching instruction; and performing frequency modulation market settlement and electric energy market settlement according to the out-clearing result.

The invention discloses a method for preparing a ceramic frit from electronic waste. The ceramic frit is prepared from the following raw materials in parts by mass: 40-55 parts of waste electronic glass, 15-27 parts of polishing powder, 8-10 parts of waste aluminum oxide, 6-8 parts of waste desulfurized gypsum, 3-6 parts of a boron-containing modifier, 5-8 parts of waste fuel and 1-3 parts of a mineralizing agent. The method comprises the following steps: stirring and mixing waste electronic glass, polishing powder, waste aluminum oxide, waste desulfurized gypsum, a boron-containing modifier and a mineralizer, adding waste fuel, further stirring and mixing, and then feeding into a granulator for granulation; feeding the prepared particles into a smelting furnace for calcining, wherein the calcining temperature is 1050-1150 DEG C; and carrying out water quenching molding on the molten material to obtain the ceramic frit. The method for preparing the ceramic frit from the electronic waste solves the technical problems of high energy consumption of ceramic frit calcination and resource waste caused by direct treatment of the electronic waste in the prior art.

The embodiment of the invention provides a workflow allocation method and device based on a multi-cloud environment, equipment and a storage medium. The method comprises the following steps: determining a to-be-allocated workflow, wherein the to-be-allocated workflow comprises a plurality of to-be-allocated tasks with an execution time sequence relationship; obtaining a resource state set corresponding to the plurality of cloud service nodes, wherein the resource state set comprises a plurality of virtual machine states corresponding to the plurality of cloud service nodes; generating a target input text according to the to-be-allocated workflow and the resource state set, and inputting the target input text into the large language model to obtain node allocation information; reading a target virtual machine state set of a target cloud service node corresponding to the node allocation information from the resource state set; and inputting the to-be-allocated workflow and the target virtual machine state set into a pre-trained target decision model to obtain a task allocation relationship between the plurality of to-be-allocated tasks and the virtual machines on the target cloud service node. Therefore, the reasonability of workflow allocation can be improved, and the resource utilization efficiency is improved.

This invention discloses a task scheduling and motion control method for collaborative operation of a robotic arm and an AGV, belonging to the field of intelligent manufacturing and automation control. The method includes: a central scheduling module receiving task requests and decomposing them into a set of subtasks; allocating subtasks to corresponding execution devices using a combination of auction mechanisms and cost function optimization; determining whether task conflicts exist; if so, executing a re-task process; otherwise, executing a collaborative control process. The collaborative control process includes generating a global trajectory using a hybrid path planning method and real-time correction of the local trajectory by a local control module; and achieving synchronous control of the robotic arm and AGV at the handover point using a state machine-based handshake protocol. This invention effectively solves the problem of task conflicts and resource competition among multiple devices through a unified task scheduling and collaborative control mechanism, improving the handover accuracy and operational efficiency of the grasping-delivery process, and is applicable to complex scenarios such as pharmaceutical research and development and chemical production.

This invention discloses a request processing method, apparatus, device, and storage medium, comprising: when the matching result is determined to be a matching failure, inferring the user request using a synchronous direct transmission mode through the target instance to obtain a first key-value cache, and determining the storage method of the first key-value cache based on its cost; when the matching result is determined to be a match with the target key-value cache, inferring the user request using a prefetch incremental mode through the target instance to obtain a second key-value cache, and directly saving the second key-value cache to a storage center. In the synchronous direct transmission mode, the cost of the key-value cache obtained through user request inference is calculated, and key-value caches that meet the cost requirements are saved to the storage center, avoiding invalid space occupation in the storage center and improving the inference efficiency of subsequent user requests. In the prefetch incremental mode, when a key-value cache hit meets the cost requirements, it is directly used to improve inference efficiency and avoid redundant calculations and waste of computing resources.

The present disclosure provides a data processing method based on a distributed system, a data processing device based on a distributed system, an electronic device and a computer readable storage medium, belonging to the field of computer technology. The method comprises: obtaining the number of nodes of the data processing node, obtaining the number of distributed models included in the first model for processing data; the first model is processed in parallel through the distributed system; the distributed model is a part of the first model; determine the target step according to the number of nodes and the number of models; the subtasks of different time steps in the data processing task are allocated to the corresponding data processing nodes; the processing results of the subtasks are interacted according to the target step, so that the data processing node executes the subtasks of the subsequent time steps according to the processing results. The present disclosure can effectively and reasonably process data in a distributed system.

The application discloses a Flash proxy access method and system based on an AMP virtual Ethernet, and the method comprises the following steps: establishing an AMP virtual Ethernet communication link between a master processor and a slave processor; the master processor provides a proxy access service for a Flash storage device, configures a network service end point, and establishes an access mapping relationship between the network service end point and a Flash storage area; the slave processor sends a Flash access request to a corresponding network service end point through the AMP virtual Ethernet communication link; the master processor analyzes and validity checks the Flash access request, determines a target Flash storage area according to the network service end point and the access mapping relationship after the checking, calls a suitable master processor side driving interface to execute a Flash operation, and returns an execution result. The application can improve access consistency, reliability and cross-storage area adaptation capability in a multi-processor environment.

This application relates to a method, apparatus, equipment, and storage medium for optimizing decision-making in post-disaster casualty rescue. The method includes: acquiring multidimensional physiological parameters and individual environmental factors of the casualties; using an improved START classification method and fuzzy comprehensive evaluation for initial injury quantification; and using a Markov state transition model to predict the injury evolution trend, generating a dynamic priority queue. A collaborative optimization model integrating rescue order, rescue route, and rescue method is constructed, aiming to maximize the expected number of survivors. A decomposition-coordination strategy is used to jointly solve the model, generating a globally optimized rescue plan. An event-driven mechanism is used to trigger rescheduling based on collected dynamic environmental data. Rolling time-domain optimization is performed based on the predictive control of the collaborative optimization model, and incremental updates are used to replan only the local plans affected by disturbances, outputting adaptively adjusted rescue instructions. This method can effectively improve the expected number of survivors and overall rescue efficiency.

The invention provides a data processing method for application communication. The method comprises the following steps: monitoring a life cycle event of a native application of a mobile terminal in a communication system through a life cycle manager; based on the life cycle event of the mobile terminal native application, updating the life cycle state of the three-dimensional vehicle control application in the communication system, so that the updated life cycle state of the three-dimensional vehicle control application and the life cycle state of the mobile terminal native application are kept synchronous; wherein the life cycle manager is located in a framework layer between a native application of the mobile terminal and a three-dimensional vehicle control application; the three-dimensional vehicle control application and the mobile terminal native application have different life cycle division modes. According to the scheme, the life state of the native application of the mobile terminal and the life state of the three-dimensional vehicle control application can be kept synchronous, and therefore the stability problem and the resource leakage problem caused by state inconsistency can be avoided.

This invention provides a conflict-aware transactional reconstruction method and apparatus for vehicular networks, belonging to the field of vehicular network configuration and scheduling control technology. It solves problems such as multi-node time slot conflicts, cross-device submission asynchrony, and inability to quickly roll back after update failures in existing vehicular network transactional reconstruction processes. The method includes: upon receiving a reconstruction request, generating a set of scheduling update operations based on the reconstruction request and the latest scheduling status table of each device node in the vehicular network; establishing a conflict graph and a dependency graph based on the scheduling update operation set; determining a set of transaction execution layers and the corresponding hierarchical execution sets for each transaction execution layer based on the conflict graph and dependency graph; and performing transactional reconstruction on each device node based on the set of transaction execution layers and the corresponding hierarchical execution sets, including differential patch pre-writing, atomic pointer switching triggered by a unified timestamp, and distributed consistency monitoring.

The invention discloses an intelligent dynamic coding system based on multi-modal visual identification and a control method, and relates to the technical field of automatic coding, and the system comprises a data synchronous collection module which obtains multi-modal data containing a sensitive area and carries out calibration output; the space-time-confrontation feature fusion module extracts and fuses multi-modal features; the dynamic trajectory prediction module generates a precise motion trajectory; the coding path optimization module generates and optimizes a coding path; the execution and feedback module drives code printing equipment to execute, collects effect data and feeds back the effect data to the preorder module, and an iterative optimization closed loop is formed; an integrated framework is adopted, driving, regulating and monitoring units are deeply fused, the structure is simplified, stability and reliability are improved, multiple scenes are flexibly adapted, meanwhile, a multi-module cooperation mechanism is constructed, signal and energy interaction is optimized, efficiency is improved, energy consumption and cost are reduced, maintenance and upgrading are facilitated through modular design, and practicability and market competitiveness are enhanced.

The application provides a method and system for checking the cross-modal information consistency of a maintenance plan based on a large model, and relates to the technical field of big data analysis. The method comprises the following steps: identifying a target maintenance plan through a power large model, determining key maintenance content, an unstructured proportion, and an attachment type complexity; obtaining a text data complexity and a business risk index based on key maintenance content analysis, as a first checking difficulty coefficient and a second checking difficulty coefficient; determining a third checking difficulty coefficient according to the unstructured proportion and the attachment type complexity evaluation; taking the expected review time limit of the target maintenance plan as a constraint, formulating an adaptive checking depth and an adaptive checking breadth based on the first checking difficulty coefficient, the second checking difficulty coefficient, and the third checking difficulty coefficient; and performing cross-modal information consistency checking on the target maintenance plan according to the adaptive checking depth and the adaptive checking breadth. The dynamic allocation of resources for checking the cross-modal information consistency of a maintenance plan is achieved.

The invention discloses a casting powder slag entrapment resistance evaluation method, device and equipment, a medium and a product. The method comprises the following steps: acquiring a spreading image of to-be-detected casting powder after complete melting on a sample carrier, then determining a spreading area of the to-be-detected casting powder based on the spreading image, finally acquiring a reference spreading area, and comparing the reference spreading area with the spreading area to obtain the slag entrapment resistance of the to-be-detected casting powder, the reference spreading area is obtained by measuring the reference casting powder sample with the viscosity being the preset viscosity value and the slag entrapment defect occurrence rate being lower than the preset threshold value under the same test condition as the to-be-detected casting powder, and the slag entrapment resistance of the casting powder can be judged in advance before production, so that resource waste is reduced, and the productivity is improved.

The invention discloses an intelligent packing decision-making system based on multi-modal data fusion and deep reinforcement learning, which relates to the technical field of intelligent logistics and comprises a cloud decision-making layer deployed in a regional computing center, an edge execution layer arranged in each sorting center and an equipment sensing layer integrated with an IoT (Internet of Things) sensor, the system further comprises an intelligent analysis decision module, a dynamic resource scheduling and elastic capacity expansion module, a digital twinborn verification module and a self-adaptive control closed-loop management module which run on a cloud decision layer. According to the intelligent packing decision-making system provided by the invention, multi-dimensional information such as a time sequence, a spatial topology, an equipment state and package vision is comprehensively considered through a multi-modal data fusion technology, the data limitation of traditional manual decision-making is overcome, and the system can capture dynamic change of data in real time through application of a deep reinforcement learning model; the optimal packing scheme is quickly generated, the business volume fluctuation is effectively dealt with, and the optimal configuration of logistics resources is realized.

The application discloses a kind of multi-radar resource management methods based on risk assessment, belong to radar signal processing technical field.First, the joint state space model of fusing motion state and classification state is constructed, state estimation is carried out using extended Kalman filter, and the misjudgment cost of target classification is modeled by introducing statistical risk function;Subsequently, the expected risk difference before and after observation is calculated, and the expected risk reduction value of the quantitative target current observation value is obtained;Boolean scheduling variable is introduced to construct scheduling optimization model, and the global ERR is maximized as objective function;Finally, local search and conflict resolution strategy are iteratively solved, and the optimal allocation result is quickly converged.The application realizes the scheduling target of key target priority observation, sensor resource efficient utilization and overall risk minimization.

This invention belongs to the interdisciplinary field of machine learning and data center thermal management, specifically disclosing a machine learning-based intelligent control method for data center cooling systems. This method configures each cooling unit as an intelligent agent with perception and decision-making capabilities, achieving distributed collaboration through a digital pheromone communication mechanism. It autonomously adjusts the opening of air and water valves based on pheromone concentration gradients and optimizes energy consumption and temperature deviation using a local reinforcement learning model. In the event of a fault, neighboring agents can automatically identify the anomaly and take over the cooling task. Through the above technical solution, this invention achieves sub-second real-time control, highly reliable self-healing operation, and a significant improvement in cooling energy efficiency.

The portable disassembly and transportation monitoring equipment placing device of the application comprises a power supply box, a continuous monitoring equipment fixing module, an external equipment suspension module, an unmanned aerial vehicle landing / laser radar installation platform and a module interfixing structure; the device can maximize the utilization of functions, save space, reduce costs, play a greater role in extreme conditions and harsh sites in the field, has the advantages of multi-mode power access, short time endurance, external expansion, high expandability of the module, high compactness of the structure, reduction of space occupation, high strength and impact resistance of the structure, high safety of the structure, certain waterproofness of the whole equipment, naked use in rainy and snowy weather, cost reduction, repeated use of each module, high universality, avoidance of cost and resource waste caused by non-universal structural parts and high stability of the system.

The present application relates to the technical field of corpus collection control, and particularly relates to a corpus automatic collection and quality control method and system based on an intelligent algorithm, which constructs state representation by acquiring corpus attribute information and generates an adversarial collection strategy, extracts blind area corpus after execution, selects boundary corpus and collects, constructs feature mapping to determine associated corpus, transmits quality mark value to associated corpus to obtain a corrected quality value, transmits utility measurement value to boundary corpus to obtain a predicted utility value, and finally fuses quality and utility values to execute corpus quality control, thereby improving the comprehensiveness of corpus collection and the accuracy of corpus quality control.

This invention provides a method for recovering silver from waste silver plating solution, belonging to the field of heavy metal industrial wastewater treatment technology. The invention involves adding sodium hypophosphite to waste silver plating solution containing complexed silver. After the sodium hypophosphite dissolves, the pH of the system is adjusted to alkaline. A reduction reaction is then carried out under heating conditions to generate elemental silver particles. This invention utilizes the in-situ generation of hydrogen free radicals from sodium hypophosphite under alkaline heating conditions to achieve direct reduction and complex breaking of stable silver nicotinic acid complexes and elemental silver conversion. This process is completed in one step, simplifying the traditional complex multi-step process of "oxidation-complex breaking-ion release-chemical precipitation" into a single reduction step, significantly shortening the treatment process and improving the reaction rate and treatment efficiency.

The invention provides a computing equipment control method and device, electronic equipment and a medium, and belongs to the technical field of computing equipment control. The method comprises the steps that a task real-time coefficient is determined based on a response requirement field of a target task; determining an average response time and a thread wakeup frequency based on the task behavior data of the target task, and calculating a task real-time score based on the average response time and the thread wakeup frequency; determining a dynamic task type of the target task based on the task real-time coefficient and the task real-time score; determining a target resource allocation strategy based on the dynamic task type and the task priority of the target task; selecting a target resource from the device hardware resource data of the computing device based on the resource demand information of the target task and a target resource allocation strategy; and scheduling a target resource of the computing device to enable the computing device to execute the target task through the target resource. The problems of unreasonable resource allocation and low resource utilization rate in the prior art can be solved.

The application discloses a portable charging pile and belongs to the technical field of charging piles. The portable charging pile comprises a box body, a wall-penetrating terminal and an output connector are arranged on the box body. The wall-penetrating terminal and the output connector provide interfaces for external power supply access and power output, so that the portable charging pile can be freely moved to the vicinity of the external power supply for connection and charging operation. Further, a circuit breaker is arranged in the box body and is electrically connected with the wall-penetrating terminal. A charging assembly is also arranged in the box body, and the charging assembly comprises a charging module, the charging module has an alternating-current terminal and a direct-current terminal, the alternating-current terminal is electrically connected with the circuit breaker, can receive external three-phase alternating current and convert the three-phase alternating current into direct current suitable for charging of an automobile battery. The direct-current terminal is connected with a direct-current side assembly, and the direct-current side assembly is electrically connected with the output connector. The direct-current side assembly is used for conducting the direct current converted by the charging module, and finally transmits the power to a user load through the output connector, so that the charging function is realized.

The application discloses a model calling request routing method and device, equipment, storage medium and product, and relates to the computer technical field. The method comprises the following steps: obtaining a request text of a model calling request; calculating the similarity between the request vector of the request text and the embedding vector of each task domain, and determining the task domain with the highest similarity as a target task domain; if the similarity between the request vector and the embedding vector of the target task domain is less than a similarity threshold value, determining the fusion quality score of each candidate model in the target task domain, determining the utility value of each candidate model based on the fusion quality score, the cost score and the delay score of each candidate model, determining the candidate model with the highest utility value as a target model, and routing the model calling request to the target model. The method can improve the accuracy of model calling request routing.

The application provides a cockpit implementation method and system based on emotion enhancement and scenario service flow intelligent agent, and belongs to the technical field of automobile intelligent cockpits. The method comprises the following steps: acquiring multi-modal data of a target user in real time, fusing the multi-modal data to generate comprehensive user states, planning a service flow based on scenario semantics and the user states, and generating and outputting action execution instructions of internal and external resources according to the service flow. Through multi-modal deep fusion and scenario semantic understanding, the application can accurately capture user states and demands, provide the most suitable service at the most appropriate time, and change passive response into active care. The application sets a user memory vector library to continuously learn and adapt to user preferences, provides a truly 'thousand faces for a thousand people' service, and as the use time increases, the service accuracy and user satisfaction present a continuously rising curve, thereby solving the problem of static experience of traditional systems.

This invention discloses a multifunctional electric firing device for unmanned equipment, characterized by reliable operation, convenient assembly, the ability to switch between firing and firing states, and a certain degree of anti-theft function. It includes a servo motor mounting base and a cover plate, which are located on both sides of the trigger guard of a firearm. The servo motor mounting base and cover plate are connected by clamping screws and clamped and fixed to the trigger guard. Servo motors A and B are mounted on the servo motor mounting base. A selector lever cover is mounted on the output shaft of servo motor A, and the selector lever cover is fitted onto the firearm's selector knob. The selector lever cover is used to rotate the firearm's selector knob. A servo motor arm is mounted on the output shaft of servo motor B, and the servo motor arm cooperates with the front part of the firearm trigger for pulling the trigger.

The application discloses a GPU resource pool dynamic reconstruction method for artificial intelligence training tasks, relates to the technical field of computers, and comprises the following steps: obtaining the hardware computing capacity, the size of the video memory capacity, the memory access efficiency and the multi-dimensional attribute information of the architecture generation difference of each graphics processing unit in the GPU resource pool through a pre-established hardware feature database, and forming unified hardware performance portrait data; the GPU resource pool dynamic reconstruction method for the artificial intelligence training tasks improves the overall utilization efficiency and stability of the GPU resource pool, and improves the execution efficiency of the artificial intelligence training tasks and the long-term optimization effect of resource scheduling.

This application relates to the field of medical risk prediction technology, providing a method and decision support system for preoperative risk prediction and decision support for hemostasis during cesarean section. The method includes: training a multi-factor logistic regression risk prediction model based on a training dataset to obtain the regression coefficients and odds ratios of variables; generating a preoperative risk scoring table by stratifying and scoring risk factors based on the regression coefficients and the mapping relationship between variables and risk factors; calculating the total risk score based on the risk factors to which the values ​​of each variable in the data to be predicted belong, combined with the scores corresponding to each risk factor in the preoperative risk scoring table; and converting the total risk score of each data point to be predicted into a predicted probability and risk level for invasive hemostasis with uterine preservation. This transforms the complex multi-factor logistic regression risk prediction model into a simple preoperative risk scoring table, allowing for easy calculation of the total risk score for the data to be predicted, facilitating rapid prediction of probabilities and enabling personalized preoperative risk probability assessment.

The application discloses a kind of based on multi-core heterogeneous vehicle gateway monitoring method, device, equipment and medium, belong to vehicle gateway monitoring technical field.The method comprises: control application processing core continues to read the current reception total number of Ethernet heartbeat packet, compare with the reception total number of adjacent period, whether new Ethernet heartbeat packet cannot be sent to Ethernet switch chip is monitored, if yes, the accumulation update of exception count is completed and threshold discrimination is carried out, when determining that Ethernet switch chip exists exception, first hardware reset request is sent to real-time processing core, and reset control of Ethernet switch chip is independently completed by real-time processing core.The application can let application processing core focus on the reading and processing of service data, let real-time processing core focus on hardware control response, clearly divide the working boundary of dual-core, realize the rapid identification and autonomous repair of Ethernet switch chip exception, meet the practical application needs of the rapid fault repair of vehicle gateway.

The application discloses a post-disaster rescue task cooperative scheduling method considering performance differences of unmanned aerial vehicles, comprising: constructing an unmanned aerial vehicle capability vector and a task demand vector; calculating the matching degree of the unmanned aerial vehicle and the task by using a weighted cosine similarity; establishing a bidirectional preference sequence on the task side and the unmanned aerial vehicle side; designing and introducing an improved Gale-Shapley stable matching algorithm to realize the stable matching between the unmanned aerial vehicle and the task in a many-to-many mode, so that an initial scheduling scheme is obtained; introducing an environment perception factor, when a re-scheduling condition is triggered due to dynamic environment changes, re-scheduling the initial scheduling scheme, and judging immediate switching or delayed switching based on a smooth switching execution strategy; and distributing the unmanned aerial vehicles to execute all rescue tasks according to the final scheduling scheme. The application can realize the automatic matching of the performance differences of the unmanned aerial vehicles and the task demands, greatly improve the accuracy, real-time performance and stability of the rescue task scheduling, and is suitable for emergency rescue scenes with multiple unmanned aerial vehicle cooperation.