27results about How to "Improve execution performance" patented technology

The invention relates to the technical field of control, in particular to a double-arm robot operation skill learning method and system based on big language model reasoning. The method comprises the following steps: firstly, carrying out context semantic modeling on an input natural language task instruction based on a large language model to generate a task semantic graph; in combination with a semantic entity in the task semantic graph and visual perception data collected by a robot, determining a three-dimensional space position of a target object through a multi-modal matching model, and constructing an environment semantic graph containing object nodes and spatial relation edges; generating an action sequence by using a language model according to the task semantic map and the environment semantic map, and generating a collaborative operation strategy based on the two-arm tail end state and an obstacle map; and finally, collecting feedback data of the sensor in real time when the action sequence is executed. According to the method provided by the invention, the understanding and execution capability of the two-arm robot on the unstructured natural language instruction is remarkably improved.

The application relates to a session instruction distribution method and device, equipment, a storage medium and a program product. The method comprises the following steps: receiving a current instruction in a current session, the current instruction being an instruction for calling a target object, and the current instruction carrying a current session identifier corresponding to the current session; for each instance identifier in an instance identifier set, determining a target weight corresponding to the instance identifier based on the current session identifier and the instance identifier, the instance identifier representing an instance for running the target object; selecting a target instance identifier from the instance identifier set based on the target weight, and distributing the current instruction to a target instance represented by the target instance identifier for execution. The method can improve the execution effect.

The invention provides a task execution method and device and related equipment, and relates to the technical field of artificial intelligence and wireless communication, the task execution method comprises the steps that a target request is acquired, and the target request is used for requesting to execute a target task; splitting the target task to obtain a plurality of sub-tasks; obtaining target information matched with the subtask in a preset database, and packaging the subtask and the target information into a target information block; generating execution path information according to the target information block, wherein the execution path information is used for representing an operation step sequence for executing the subtask; and executing the operation step sequence based on the execution path information. Therefore, the execution accuracy of the target task can be improved, namely, the execution effect of the target task is enhanced.

The application discloses a method and system for constructing a load sharing chain of a generator group unit, and relates to the field of data processing specially used for administrative, commercial, financial, management, supervision or prediction purposes. The method comprises the following steps: calculating a comprehensive response capability value of each power plant according to operation parameters and transaction parameters, arranging the comprehensive response capability values in descending order according to the numerical values to form a response sequence, calculating the regulation capability matching degree between adjacent power plants, the difference between the regulation capability matching degrees of each adjacent response section being less than a preset matching degree threshold, calculating the output regulation range and regulation rate of each generator unit, and determining the load distribution coefficient of each response section according to the output regulation range and regulation rate; determining the regulation sequence of the units in all response sections based on the load distribution coefficient, and calculating the maximum regulation amount that can be borne by each unit, and constructing the load sharing chain according to the regulation sequence and the maximum regulation amount. By implementing the method, the efficiency of task regulation of the units of the power plant can be improved.

The application relates to the technical field of large models, and provides a prompt vector generation method, device, equipment and medium, which extracts and embeds the multi-modal information of a target user, generates a personalized query vector carrying the personalized features of the target user, and fuses relevant information of the personalized query vector and a pre-constructed industry knowledge prompt library to generate a target prompt vector of the target user. The target prompt vector is determined in combination with the personalized features of the target user and global user common features, and the personalized and customized processing of the prompt vector is realized. The application can be applied to a financial scene, and when a downstream task such as insurance recommendation generates a personalized execution result for a user according to the target prompt vector, the personalized features of the target user and the global user common features can be simultaneously referred to, the target user's personalization is considered, and the rules or trends gradually presented in the insurance business scene are not deviated from, so that the execution effect of the downstream task is improved.

The application relates to the technical field of unmanned planes, and discloses an unmanned plane adaptive task planning method and system fusing real-time environment perception, which has the technical scheme as follows: inputting unmanned plane cluster physical parameters and task target parameters, completing system hardware calibration, software environment and key module parameter initialization setting; based on a full-closed loop technical architecture of "perception-cognition-modeling-decision-cooperation-feedback", the following sub-steps are executed until the task termination condition is met; through the full-closed loop technical architecture, deep understanding of environment essential laws, dynamic self-evolution of planning schemes and autonomous emergence of cluster global intelligence are realized, and core technical support is provided for efficient, safe and intelligent task execution of unmanned planes in a complex dynamic environment.

This application belongs to the field of radar networking technology, specifically disclosing a method, apparatus, equipment, and medium for multi-task collaborative planning of airborne radar networking. The method includes: calling a pre-constructed multi-task collaborative planning model for airborne radar networking; the model is established with the goal of maximizing the collaborative execution efficiency of airborne radar tasks, constrained by the joint interception probability of the networked radars, the signal-to-noise ratio of the task echo, and the system detection performance; the joint interception probability of the networked radars is determined based on the interception probability when the window functions of the radiated signals of each radar simultaneously overlap in the power domain, time domain, spatial domain, and frequency domain; and solving the multi-task collaborative planning model using the detection range parameters of the current airborne radar network to determine the current task allocation results for each radar. This application can effectively improve the execution efficiency of multi-task collaborative planning of radar networking while realizing multi-task collaborative planning for multiple radars.

The invention relates to the technical field of unmanned collaboration and resource virtualization, and provides a task collaboration-oriented unmanned system virtual resource representation and organization method and device and a storage medium. The method comprises the following steps: firstly, determining physical resource virtualization granularity (platform level, equipment level and function unit level) according to unmanned collaboration styles (task level, resource level and element level); secondly, establishing a resource tree model comprising a resource base class, an attribute resource, a data table resource and a resource sub-tree based on granularity; resource storage is carried out by utilizing an organization structure constructed by a hash table, a list and a set; and finally, forming a resource pool through standardized packaging, providing management interfaces (resource registration, cancellation and state maintenance) and scheduling interfaces (demand response, resource allocation and release), and realizing resource addressing by adopting URI coding. According to the method, the problems of low task execution efficiency, insufficient resource utilization rate and complex management caused by poor coordination task and resource adaptability, non-uniform heterogeneous resource description and non-standard packaging are solved.

The invention belongs to the technical field of unmanned aerial vehicle flight control, and discloses an unmanned aerial vehicle rapid composite control method and system based on fuzzy asymptotic. In the aspect of steady-state precision, the method breaks through the inherent limitation that traditional finite time control can only achieve bounded neighborhood convergence, and the reliability and execution capacity of the system in a high-precision task are remarkably improved. In the aspect of transient response, a nonlinear feedback mechanism in a fractional power form is introduced, so that the system can obtain a super-linear convergence rate in an initial stage or when encountering large-deviation disturbance. In terms of robustness and realizability, a structure combining low-pass filtering and dynamic error compensation is adopted in the method, and implementation difficulty caused by high-order differential calculation is effectively avoided; meanwhile, parameter perturbation, unmodeled dynamics, external wind disturbance and the like are estimated and compensated in combination with an adaptive fuzzy system, and the adaptive capacity and anti-interference performance of the system in a strong-coupling and nonlinear environment are remarkably enhanced.

The application discloses a personalized recommendation method for serving cultural and creative products, and relates to the technical field of personalized recommendation, which comprises the following steps: based on a cultural and creative exhibition area map connection structure and a tourist visit transfer sequence, a node visit continuation probability is counted and a continuation probability peak interval is determined, an access continuation span is calculated, and a first recommendation trigger buffer time limit is obtained; the time limit is subjected to scale reconstruction and interval disturbance to generate a multi-granularity second buffer time limit; an engineering mapping is established based on a time interval and a control action corresponding relationship, the second buffer time limit is converted into a recommendation candidate reservation span and a detection period as a control target; a buffer time limit score is output; the score is scanned along the buffer time limit distribution and an optimal interval is identified, and the optimal second buffer time limit is screened from the optimal interval and applied to cultural and creative product recommendation trigger and candidate management, so that the problem that a low recommendation effect is caused by an unreasonable location sensing cultural and creative product recommendation trigger time is solved.

The invention discloses a multi-unmanned aerial vehicle adaptive target surrounding control method, and relates to the technical field of multi-unmanned aerial vehicle control, and the method comprises the steps: obtaining a kinetic model and a communication topology of an unmanned aerial vehicle; according to the dynamic model, based on a formation transformation operator, determining a formation pattern of the unmanned aerial vehicle; keeping the formation pattern of the unmanned aerial vehicles unchanged, adjusting the motion direction of the virtual unmanned aerial vehicles according to the motion state of the target until the distance between the target and the virtual unmanned aerial vehicles is smaller than a preset distance threshold, and obtaining a surrounding path; constructing a dual-channel event trigger function based on a dynamic event trigger mechanism according to the kinetic model; according to the triggering time, the communication topology and the surrounding path, constructing an unmanned aerial vehicle formation distributed dynamic event triggering target surrounding control law; according to the unmanned aerial vehicle formation distributed dynamic event, triggering a target surrounding control law to surround the target; according to the invention, the trigger threshold can be adjusted in the dynamic change process, and the optimal balance between the control performance and the communication efficiency is realized.

This invention discloses a model-based method for learning robot operation skill parameters, including a single-skill policy learning module and a task-skill parameter learning module. The single-skill policy learning module includes an object-oriented representation model and a policy model conditioned on semantic goals. The task-skill parameter learning module includes an implicit state transition model learning module and an online skill parameter planning module. The single semantic skill policy learning module of this invention combines representation learning and reinforcement learning, integrating the reasoning transformation from visual input to the logical representation of the current system state with task-oriented robot policy learning to generate action parameters that satisfy the target logical state. Using reinforcement learning avoids extensive model design and processing, as well as the need for expert teaching data. Furthermore, by predefining the action sequences of skills based on an operation knowledge base, the search space for actions is reduced, lowering the exploration burden on the agent.

The application relates to the technical field of data reasoning, and discloses a mine equipment intelligent manufacturing production scheduling optimization system and method. The system comprises a topological graph generation module, an initial scheduling Gantt chart generation module, a demand intensity coefficient acquisition module, a flexible scheduling Gantt chart acquisition module and a final scheduling instruction set generation module. A topological graph is constructed based on the process connection of mine equipment. An initial scheduling Gantt chart is generated by taking a component task set as a vertical axis and a corresponding final assembly line task sequence as a horizontal axis. When a disturbance event occurs, an affected process dependency tree is extracted, a task delay cost is evaluated, and a demand intensity coefficient for manufacturing process rescheduling is output. A flexible scheduling Gantt chart is generated based on process beat deviation and equipment utilization in synchronous adjustment of a manufacturing process. A final scheduling instruction set is confirmed through process path dynamic correction and process optimization results. The application can improve the efficiency of a mine equipment manufacturing production process scheduling scheme.

The application discloses a mixed expert model word grouping scheduling method and electronic equipment, and relates to the technical field of artificial intelligence, which comprises the following steps: in response to the arrival of word units in the form of a stream to a preset buffer, obtaining the embedding vector features, routing history and arrival time corresponding to each word unit to splice into the target feature vector corresponding to each word unit, and the preset buffer is configured with a preset waiting time threshold and a capacity upper limit value; clustering and grouping each word unit according to the target feature vector to generate at least one task group; determining the priority of each task group based on the average waiting time of each word unit in each task group and the historical hot index; under the condition of meeting the limitation of the preset buffer, scheduling each task group according to the priority to enable the mixed expert model to perform operation on each word unit. The method of the application improves the utilization rate of experts and the system throughput while ensuring the upper bound of delay through short window buffering and dynamic clustering.

The embodiment of the invention discloses a command processing method and device, a storage medium and a program product, which are used for realizing a cross-platform compatible effect and improving the execution performance of a command. The method comprises the steps that a to-be-processed request is acquired, the to-be-processed request comprises a target command and indication information, the indication information is used for indicating the position of a to-be-processed object in a target system, and the target command is a command needing to be used for executing processing operation corresponding to the target command on the to-be-processed object; a first interface library is called, the first interface library comprises implementation functions of a plurality of programming interfaces, the implementation function of each programming interface is used for supporting access to candidate commands in a programming language script file of a target system, and the candidate commands comprise a target command; and calling an implementation function of a target programming interface corresponding to the target command from the first interface library so as to execute the target command on the to-be-processed object indicated by the indication information.

This invention discloses a GPU code policy-driven generation method and system based on parallelism inference constraints. It collects cross-iteration memory access address features of variables using GDB, constructs a spatiotemporal program graph, and applies a spatiotemporal graph convolutional network for modeling. This yields prediction results for loop parallelism types and a five-dimensional optimization policy prediction vector. For serial loops, it directly returns the original C code. For parallelizable loops, the prediction results are converted into natural language parallelism labels and optimization instruction sets, which are then concatenated in a fixed order to construct structured hints. This drives a large language model as a syntax constraint executor to generate target CUDA code. This invention explicitly decouples parallelism inference from code generation, eliminating erroneous parallelization of the large language model and misjudgment of optimization policies, significantly improving the accuracy and execution performance of the generated code. It can be widely applied in the GPU migration of high-performance computing programs and heterogeneous computing fields.

The application belongs to the technical field of unmanned aerial vehicle flight control, and discloses a kind of unmanned aerial vehicle fast compound control method and system based on fuzzy asymptotic, in steady accuracy aspect, the method of the application breaks through the inherent limitation that traditional finite time control can only realize the convergence of bounded neighborhood, significantly improves the reliability and execution ability of system in high-precision task.In transient response aspect, the method of the application introduces the nonlinear feedback mechanism of fractional power form, and the system can obtain super-linear convergence rate in the initial stage or when encountering large deviation disturbance.In robustness and realizability, the method of the application adopts the structure of low-pass filtering combined with dynamic error compensation, effectively avoids the implementation difficulty brought by high-order differential calculation;At the same time, combined with adaptive fuzzy system, parameter perturbation, unmodeled dynamics and external wind disturbance are estimated and compensated, which significantly enhances the adaptive ability and anti-interference performance of the system in strong coupling and nonlinear environment.

The application provides a medical follow-up management method and system based on large model prompt word engineering, and relates to the technical field of medical management. In view of the poor pertinence of the follow-up strategy in the prior art, the application obtains prompt words in a prompt word library according to a follow-up scene, and optimizes the prompt words to obtain final prompt words; health status data and living habit data of a user are collected; the health status data and the living habit data are input into a follow-up dialogue model, the follow-up dialogue model generates a preset follow-up plan according to the input data; the follow-up dialogue model adjusts the preset follow-up plan according to the final prompt words, and outputs a final follow-up plan; behavior data related to execution of the final follow-up plan by the user is obtained; and the final prompt words are updated in real time according to the health status data, the living habit data and the behavior data. The follow-up strategy generated by the medical follow-up management method has high pertinence.

The application discloses a method for performing hash connection of a distributed database, and relates to the technical field of distributed databases, which comprises the following steps: setting a relative inclination rate, calculating the product of the relative inclination rate and table data volume to obtain an inclination threshold; obtaining a heat map of table statistical information, screening elements in the table which exceed the inclination threshold to obtain an inclination value; expanding the table statistical information by using the inclination value; executing an SQL statement to obtain two input data tables, generating a hash connection physical plan by using the new table statistical information, performing hash distribution, average distribution or mirror distribution on tuples in the input data tables according to the plan, and after each node receives the tuples of the input data tables, establishing a hash table by using input data table data with small data volume, performing detection by using input data table data with large data volume, and finally performing a set union on the hash connection results of each node, so that the set union is the final hash connection result. The application can realize balanced distribution of tasks and shorten the total SQL execution time.

Embodiments of the present specification disclose a method, device and equipment for anti-join processing in a database. The method comprises: obtaining an anti-join to be executed; determining, in a right table of the anti-join, a data column related to a join condition of the anti-join as a target data column; determining a plurality of data blocks into which the target data column is divided, each of the data blocks corresponding to one or more data rows in the target data column; determining NULL statistical information recorded for the data blocks; if the NULL statistical information of the data blocks reflects that the number of NULLs contained in the data blocks is not 0, adjusting the position of the data block in the target data column upwards; and executing the anti-join according to the adjusted target data column, and ending the execution process in advance when NULL in the adjusted target data column is scanned.

The application provides a power distribution network post-disaster fault repair scheduling method and system considering voltage support capability, and belongs to the technical field of post-disaster scheduling of power distribution networks; the method comprises the following steps: obtaining initial data of the power distribution network, estimating and dynamically correcting the fault removal time based on three-point evaluation, and generating a task sequence and team configuration; then, a branch reactive power-voltage sensitivity index is calculated, which is aggregated into node equivalent voltage support capability through series-parallel equivalence, and a load dynamic recoverability discrimination criterion is constructed accordingly; then, the criterion is taken as a recovery boundary, and a repair model aiming at maximizing the load recovery amount is constructed under three constraints; finally, the repair is solved and executed by using a beam search algorithm; the application can accurately depict the dynamic characteristics of voltage support, explicitly process the variation of fault removal time, and give a dynamic optimization scheduling strategy closely coupled with the on-site repair progress, thereby significantly improving the safety and feasibility of post-disaster recovery decisions.

Embodiments of this disclosure provide a method and apparatus for dynamically allocating AI inference tasks. The method includes: real-time monitoring of the workload, memory availability, network latency, bandwidth, and energy consumption of each computing node in a cloud-edge collaborative system, wherein the computing nodes include local edge devices, edge servers, and cloud platforms; dynamically allocating tasks to different computing nodes based on task latency requirements, computing needs, data privacy requirements, and the real-time status of each computing node; adjusting the complexity of the AI ​​model used to execute the task based on the resource usage of the computing nodes; and optimizing the task allocation strategy based on historical performance data and real-time network conditions during task execution.

This invention discloses a method for generating and scheduling management agents based on a large model and dynamic workflow. The method includes: analyzing business requirements through a large model, automatically generating and instantiating specialized management agents according to capability matching; dynamically generating workflows based on a rule engine and node selection probability, and adjusting the process structure in real time according to the execution status; employing a centralized multi-agent scheduling algorithm to decompose tasks into atomic tasks, allocating and adapting agents according to comprehensive scheduling to achieve collaborative scheduling and load balancing; and establishing an execution feedback and optimization mechanism, collecting logs and analyzing them using a large model to update rule weights and scheduling parameters. This invention can automatically generate management agents, shorten deployment cycles, reduce costs, improve process flexibility and system stability, form a self-optimizing closed loop, and significantly improve the efficiency of multi-agent collaboration and the level of system intelligence.

The application relates to the technical field of fire safety, in particular to an automatic water-spraying fire extinguishing system and method based on multi-modal perception and intelligent decision-making, which comprises a perception module, a data processing and control module and an execution module; the data processing and control module is embedded with a hybrid deep learning model; the hybrid deep learning model fuses a convolutional neural network for analyzing smoke images and a long short-term memory network for analyzing temperature sequences; multi-modal sensor data can be deeply fused and intelligently understood, so that high-accuracy fire identification can be realized within 3 seconds.The beneficial effects are that, through deep fusion and interaction of hardware sensors and intelligent algorithms, the problem of slow response and high false alarm rate of traditional systems is effectively solved, and the system is particularly suitable for prevention and control scenes of new fire risks such as electric bicycles, new energy vehicles and the like.

This invention belongs to the field of artificial intelligence technology and relates to a multi-agent recursive task decomposition and adaptive routing method and system. The main steps of this method include task reception and semantic parsing, recursive task decomposition and task dependency graph construction, agent dynamic matching and adaptive routing, knowledge gap identification and "illusion" suppression mechanism, fault-tolerant scheduling and task rerouting, subtask result aggregation and closed-loop evaluation feedback. This invention comprehensively improves the automation, accuracy, and reliability of complex task processing by integrating recursive automatic task decomposition, agent dynamic routing, adaptive knowledge completion, a robust fault-tolerant mechanism, and a result feedback optimization mechanism. It is particularly suitable for complex knowledge processing scenarios such as in-depth interpretation of policy documents, multi-dimensional scientific research information evaluation, and review of large-scale project materials, and has significant technical advantages and broad application prospects.

The invention relates to the technical field of network decision making, in particular to a collaborative decision making method and system in a crowd sensing network based on artificial intelligence, and the method comprises the steps: obtaining the task characteristic information of each sensing terminal in the crowd sensing network, and obtaining the resource environment information of the crowd sensing network; constructing a task-resource matching evaluation model based on the task characteristic information and the resource environment information; based on the task-resource matching evaluation model, initial matching value evaluation values of all tasks in the current to-be-allocated task set for different sensing terminals are determined, and a target task allocation strategy is determined according to the current strategy value matrix and the strategy values corresponding to the multiple task allocation strategies respectively. According to the method, the matching degree between the task and the sensing terminal can be accurately evaluated by constructing the task-resource matching evaluation model, so that the task allocation is optimized, the resource waste can be reduced, and the task completion efficiency and success rate can be improved.

This invention discloses a method, apparatus, device, medium, and product for allocating physical machine resources. The method includes grouping all physical machines in a cloud delivery environment into several physical machine groups; selecting several candidate physical machine groups that meet the needs of a target virtual machine cluster based on the resource information of each physical machine group using a minimum resource utilization priority strategy; determining a target physical machine group based on all candidate physical machine groups using a random selection strategy; and allocating the physical machines in the target physical machine group to the target virtual machine cluster. This method combines a random selection strategy with a minimum resource utilization priority strategy, using the minimum resource utilization strategy to select multiple physical machine groups and the random selection strategy to select a specific physical machine group. This solves the impracticality of the random selection strategy and the resource contention problem of minimum resource utilization, avoiding excessively long waiting times for certain tasks and affecting overall task time and performance.