68 results about "Task adaptation" patented technology

The invention provides a multi-agent cooperation method, system and device and a storage medium, and relates to the technical field of multi-agent collaboration.The method comprises the steps that initial role allocation is conducted on multiple agents, one agent is an observer, the other agent is a coordinator, and the other agents are all executors; a strategy network based on deep reinforcement learning is introduced according to the running state of the multiple agents to dynamically adjust role allocation of the multiple agents, and a role allocation strategy is dynamically adjusted according to task completion rewards, role conflict punishment and resource conflict rewards; the coordinator constructs a task priority and a dependency relationship based on the task graph or the task dependency tree, and dynamically allocates tasks according to the state, the capability vector and the task adaptation degree of the executor; conflicts are recognized through resource contention detection, task overlapping detection and behavior conflict detection, and the conflicts are coordinated. According to the invention, multi-agent responsibilities are layered, and the task completion efficiency is improved through task allocation and conflict detection and coordination.

The invention discloses a building robot task planning and skill learning method and system, and relates to the technical field of intelligent construction and building robot control, and the method comprises the steps: receiving construction target description and construction environment information through a multi-modal task-skill collaborative generation model; and generating a structured semantic vector and reasoning a construction planning graph containing a plurality of subtasks. And collecting operation behavior parameters of the building robot in each subtask according to the construction planning graph, constructing an operation sample, and inputting the operation sample into the multi-modal task-skill collaborative generation model for skill learning and action instruction generation. And in the construction execution process, various sensing data are fused for real-time judgment, and if the task execution state is inconsistent with the planning state, the multi-mode task-skill collaborative generation model adjusts an instruction based on the latest state and outputs a reasoning chain. According to the method, the deployment efficiency and task adaptability of the robot system in actual construction are improved, and a foundation is laid for man-machine cooperation and unmanned construction in the field.

The embodiment of the invention provides an intelligent learning path recommendation method and system based on a dynamic state. The method is applied to the technical field of intelligent learning recommendation, and comprises the following steps: calculating a skill adaptation weight, a load weight and an emergency degree weight in real time based on a dynamic state of a target employee, including a current skill improvement condition, a workload and a task emergency degree; combining the skill matching degree, the learning strength and the task association degree of each learning unit in the candidate learning path, and performing comprehensive evaluation by using a multi-weight scoring function; and sorting the paths according to the comprehensive score, generating a personalized recommended learning path set and sending the personalized recommended learning path set to an employee terminal, thereby realizing intelligent learning path recommendation with dynamic adaptation and accurate matching. According to the scheme, personalized learning path pushing aiming at actual post requirements and working states of the employees can be realized, correlation, urgency and acceptability of learning contents are improved, learning efficiency and task adaptability are remarkably enhanced, and the employees are helped to quickly compete with post targets.

The embodiment of the invention provides a humanoid robot industrial task scene generation method and system based on a large language model, and belongs to the technical field of industrial automation and artificial intelligence. The method comprises the steps of obtaining an industrial task database; extracting and converting the task parameters according to an industrial task database to obtain first task prompt information; performing thinking chain task decomposition on the industrial task to obtain subtask sequence information; training the large language model to obtain an initial task execution strategy; performing multi-dimensional evaluation on the initial task execution strategy to obtain an evaluation result; optimizing the first task prompt information according to the evaluation result to obtain second task prompt information; and adjusting the large language model according to the second task prompt information to obtain a target task execution strategy. According to the embodiment of the invention, the task adaptability and strategy reliability of the humanoid robot in a complex industrial scene can be improved, and the flexibility and efficiency of industrial automatic production are remarkably improved.

The invention provides a cue word optimization method and system based on an approximate sub-module function and continuous learning, and relates to the technical field of artificial intelligence multi-mode perception.The method comprises the steps that a candidate cue word set is constructed, a combined objective function based on the property of the approximate sub-module function is designed, and a candidate cue word set is constructed; solving the combined objective function by adopting a greedy selection algorithm combining random disturbance, multi-round iteration and a task self-adaptive mechanism, realizing optimal selection of a candidate cue word set, gradually selecting a cue word with the maximum gain from the candidate cue word set, adding the cue word into an optimal subset, and when a new cue word is selected, selecting the cue word with the maximum gain into the optimal subset. And if the target cue words are selected, carrying out local optimization once, after all the target cue words are selected, carrying out joint optimization on all the cue words in a continuous space by adopting an alternate optimization strategy, and carrying out continuous iteration until the optimized cue words are obtained. According to the method and the device, efficient self-adaptive updating of the cue words of the language model is realized, so that the generalization performance and robustness of the model in zero-sample, few-sample and concept drift scenes are improved.

The invention discloses an adaptive task unloading mechanism driven by meta-reinforcement learning in an edge computing environment. According to the mechanism, a system architecture composed of a user equipment layer and an edge server layer is constructed, and task analysis, state perception and strategy optimization processes are combined to realize unloading scheduling optimization of a multi-task dependent structure. Task unloading is modeled as a Markov decision process, a double-layer training mechanism is adopted, local strategy training is realized by utilizing a near-end strategy optimization algorithm, and the generalization ability of the system is improved in combination with cross-task meta-strategy learning. According to the method, the sequence is fused into the sequence neural network structure and the multi-head attention mechanism, and the accuracy and efficiency of unloading strategy generation are improved while the task dependency relationship is modeled. The mechanism has good task adaptability and delay optimization performance in a dynamic edge computing environment, and is suitable for various mobile computing scenes.

The invention provides a language model training method and device, electronic equipment and a storage medium. The method comprises the following steps: training an initial language model to obtain a weight matrix, and decomposing the weight matrix to obtain a basic feature matrix and an initial task adaptation matrix; updating the initial task adaptation matrix to obtain an updated task adaptation matrix; determining the current sensitivity of each singular vector direction in the update task adaptation matrix of the current training period; determining a sensitive direction from a plurality of singular vector directions according to the current sensitivity, and performing regularization processing on a scaling coefficient associated with the sensitive direction to obtain a regularization loss function; in response to determining that the change sensitivity of the singular vector direction in the continuous preset number of training periods is in an attenuation state, performing parameter resetting on the zoom coefficient to obtain an updated zoom coefficient; and carrying out merging processing on the basic feature matrix and the update task adaptation matrix, and taking the update language model as a target language model.

The invention relates to a time sequence lightweight analysis method which comprises the following steps: 1) constructing a time sequence lightweight adaptive network model which comprises a basic input layer, a core function module, an optimization training module and a task adaptation layer; 2) inputting the time sequence data into a basic input layer, processing the time sequence data, and outputting a patch sequence; 3) in the core function module, performing core noise filtering and frequency domain feature enhancement on the patch sequence to obtain a time domain enhancement feature; 4) performing feature fusion on the time domain enhancement features to obtain interaction enhancement features; and 5) inputting the interaction enhancement feature into a task adaptation layer to obtain a time sequence analysis result. According to the method, accurate capture of nonlinear features and complex structures of time sequence data, fusion of short-term and long-term dependency relationships and universal adaptation of multiple types of time sequence tasks can be effectively achieved, model lightweight and calculation efficiency are considered at the same time, and balance of performance and efficiency is achieved.

The invention relates to a target detection model robustness improvement method based on task adaptation. The method comprises the following steps: S1, constructing different types of adversarial samples of clean samples according to adversarial loss types; s2, calculating the adversarial loss of the constructed adversarial sample; s3, screening out an adversarial sample with the maximum adversarial loss for participating in adversarial training; s4, retraining the target detection model, and updating the model parameters of the target detection model; and S5, performing image target detection by using the target detection model trained in the step S4. The adversarial samples are generated based on different loss types, the sample with the maximum loss is selected for adversarial training, and meanwhile, the robustness of the target detection model is improved in combination with the adversarial training method based on category perception.

The invention discloses a large language model task adaptation method and system based on multi-modal prompt learning, and relates to the technical field of multi-modal learning and prompt-driven task adaptation, and the method comprises the steps: collecting multi-modal data, carrying out the unified coding, fusing the multi-modal data into shared semantic representation, and analyzing a task instruction and a context to construct a task state vector. A structured prompt is generated based on the fused representation and task semantics, and an optimizer is introduced to dynamically adjust the prompt content to adapt to feedback changes. And jointly embedding prompt and input into the large model, scheduling a fine tuning mechanism to execute reasoning, and outputting an evaluation index to be compared with a threshold value for feedback. According to the method, through multi-modal semantic fusion, task state modeling and prompt optimization regulation and control, accurate adaptation and dynamic response of a large language model in a complex task are achieved, the self-adaptability of a prompt structure, the consistency of model reasoning and the stability of task execution are improved, and the method has remarkable generalization ability and engineering practical value.

The invention discloses a heterogeneous large language model multi-agent-oriented multi-scale task collaborative orchestration method and device, and the method comprises the steps: firstly obtaining a to-be-orchestrated task processing demand and the task adaptation capability of each agent, and constructing a heterogeneous interaction graph containing agent nodes and task nodes; thirdly, clustering and grouping nodes in the graph according to the task adaptation capability of the intelligent agent and the similarity degree of task processing requirements; then, performing coarse-grained processing on the grouped graphs by using a pre-trained multi-scale collaborative task arrangement model to obtain a coarse-grained collaborative arrangement scheme; and inputting the task nodes into a task scheduling priority model to obtain priorities, and carrying out fine-grained processing on the coarse-grained scheme by using the priorities to obtain a fine-grained collaborative arrangement scheme between the intelligent agent and the corresponding task. The invention aims to solve the problem of low scheduling efficiency caused by agent isomerism, high communication overhead and task dynamic complexity in task arrangement.

The invention discloses an urban traffic cooperative scheduling method and system based on a large model and multiple agents. The method comprises the steps that natural language task description is converted into a task semantic graph and a structured prompt; generating a plurality of strategy candidates by using a large language model, and performing language scoring and reasoning arbitration; based on the scoring result, selecting an optimal strategy for multi-agent execution; behavior execution data are collected and evaluated and fed back, and strategy closed-loop updating is achieved through parameter optimization and Prompt fine tuning; and a distributed task embedding mechanism and a lightweight migration module are combined, so that the multi-task adaptability and the training efficiency of the system are improved. The cooperative training system constructed by the invention has the capabilities of natural language interaction, strategy interpretability, behavior controllability and task migration, and is suitable for various complex urban tasks such as traffic jam dispersion, emergency response, signal lamp linkage and the like.

The invention discloses an intelligent task scenario closed-loop optimization system and method, and belongs to the field of simulation deduction and intelligent decision making. According to the system, on the basis of natural language interaction, a locally deployed large language model and a vector database are combined, and rapid generation and dynamic optimization of task scenarios are achieved. A user inputs a task target through a natural language, a system retrieves related knowledge and constructs a prompt word context, a structured task scenario is generated, and then simulation deduction is automatically carried out. The system supports a human-in-the-loop decision-making mechanism, task planning can be optimized according to the real-time situation, and key parameters are allowed to be manually adjusted after deduction. By analyzing a simulation result, a cue word template is further corrected, fine adjustment training is conducted on the model through a high-quality sample, and continuous improvement of language understanding and logical reasoning ability is achieved. According to the method, a closed-loop process from generation, deduction, feedback to retraining is constructed, and the task adaptability and decision-making ability of the model in a complex environment are remarkably enhanced.

The invention discloses a recommendation system user behavior sequence feature enhancement method based on a big language model, which comprises the following steps: acquiring a recommendation data set, performing feature structure analysis according to the use of the recommendation data set in a recommendation system and a conventional field to construct a first text, and inputting the big language model to generate a data set description; constructing a second text according to the target task and the sequence feature type, and inputting the second text and the data set description into a large language model to generate feature semantics; constructing a third text based on the feature semantics, and inputting the large language model to generate feature codes; feature codes are embedded into a prediction model training environment, and automatic generation and loading of new features in training data are achieved; and carrying out a new feature experiment in the prediction model framework, and evaluating and generating a new round of sequence features. The method has higher expressive power, context perceptibility and task adaptability in a behavior data modeling-oriented recommendation system and a personalized system, and can effectively solve the technical bottlenecks of an existing method in the aspects of structure, expression, generalization and the like.

The invention relates to the technical field of energy management, in particular to a fourth-generation residence intelligent energy management and energy saving system and method, and the system comprises a response feature recognition module, a response gradient calculation module, a task adaptation sorting module, an instruction time sequence regulation and control module and an inertia correction module. In the method, the response fluctuation equipment is dynamically identified and labeled by identifying the response time offset of the equipment in the continuous task, extracting the starting and finishing offset characteristics, finishing response inertia classification in combination with the equipment type and the operation cycle, calculating the frequency change of the offset change and identifying the equipment with an unstable response state; response scheduling priority weights are distributed, control tasks of equipment are ranked, response lag records are extracted in combination with sending of control signals and response time difference, dynamic correction of equipment control precision and ordered adjustment of execution rhythm are achieved, the phenomena of multi-equipment operation conflict and response out-of-control are effectively avoided, and the service life of the equipment is prolonged. And the coordination of equipment regulation and control and the flexibility of system energy efficiency scheduling are improved.

The present application provides a task adaptation and federated learning method and system for edge side visual analysis, for task adaptation, in the scene of federated learning, by preventing negative transfer to ensure the high precision of the model in visual analysis, so that the model can be trained through local samples, and can also interact with other edge devices to learn the relevant task information of other edge devices, thereby improving the precision of the model in visual analysis of the edge device; at the same time, the overhead in federated learning is reduced, the communication overhead is large when the edge device interacts, the resources of the edge device are limited, by using local task knowledge accumulation, the communication overhead and calculation overhead of the whole training process in response to task adaptation are reduced under the condition of ensuring high learning precision, and the communication size of the edge device will not increase with the increase of the task.

The invention provides a small sample training method and device for an OCR model, electronic equipment and a medium, and the method comprises the steps: carrying out the data enhancement of a small sample training data set, generating a plurality of meta-tasks, and enabling each meta-task to comprise a plurality of image samples and corresponding task domain identifiers and text recognition result labels; training an initial OCR (Optical Character Recognition) model by taking the plurality of image samples of each meta-task and the corresponding task domain identifiers as training samples and taking the text recognition result labels as sample labels; the initial OCR model comprises a task adaptation layer, an initial text detection layer, an initial text recognition layer and an initial correction layer; the task adaptation layer takes the task domain identifier as input to generate task adaptation parameters, and the task adaptation parameters are used for adaptively updating the parameters of the initial text detection layer, the initial text recognition layer and the initial correction layer. According to the invention, model training can be carried out based on small samples, the training efficiency is high, the generalization ability of the model is improved, and the training cost is reduced.

The embodiment of the application provides a kind of based on large language model humanoid robot industrial task scene generation method and system, belong to industrial automation and artificial intelligence technical field.The method comprises: obtaining industrial task database;According to the task parameter extraction and transformation of industrial task database, obtain first task prompt information;Thought chain task decomposition is carried out to industrial task, and subtask sequence information is obtained;Large language model is trained, and initial task execution strategy is obtained;The initial task execution strategy is evaluated in multidimension, and evaluation result is obtained;According to evaluation result, first task prompt information is optimized, and second task prompt information is obtained;According to second task prompt information, large language model is adjusted, and target task execution strategy is obtained.The embodiment of the application can improve the task adaptability and strategy reliability of humanoid robot in complex industrial scene, significantly improve the flexibility and efficiency of industrial automation production.

This invention relates to the field of model adaptation technology, specifically to a multi-task adaptation method based on a hybrid sparse expert network. The method includes: establishing a pre-trained model on a large-scale dataset; establishing a downstream multi-task dataset; constructing a hybrid sparse expert network, including: building a backbone network based on the pre-trained model on the large-scale dataset, and adding an adaptation network and a multi-task head to the backbone network; training the parameters of the adaptation network and the multi-task head in the hybrid sparse expert network based on the downstream multi-task dataset; and applying the trained hybrid sparse expert network to various downstream tasks. This invention can be used for various visual models and different downstream tasks, achieving improved multi-task adaptation performance under low computational cost conditions.

This invention discloses a robot autonomous configuration service system built around a large-model agent framework, including a robot autonomous task system, a large-model agent framework, an application system, and an evaluation system. The large-model agent framework receives user instructions, parses task requirements, and generates task execution strategies. The robot autonomous task system drives the robot to complete physical actions based on the task execution strategies and introduces an agent extension mechanism to adapt the large-model agent framework to different types of robots and task environments. The application system is used for data transmission, storage, and visual interactive operation. The evaluation system collects task execution data in real time and outputs quantitative evaluation results to optimize the execution strategy. This system effectively reduces the threshold of human-machine interaction and the difficulty of understanding two-way intentions, significantly improving the robot's task adaptation flexibility and execution reliability, and is suitable for various robot autonomous operation needs in multiple scenarios such as inspection, delivery, and search.

The invention relates to a pipeline task adaptation method and device for an intelligent liquid outlet machine and computer equipment. The method comprises the following steps: obtaining product formula information of each product type, current pipeline information and current order information of an intelligent liquid outlet machine of each store, and generating pipeline adaptation information of the intelligent liquid outlet machine of each store in each product type; for each store, pipeline task information corresponding to the current order information is generated through pipeline adaptation information of the intelligent liquid outlet machine of the store in each product type, and execution feedback information, new pipeline information and new order information of the intelligent liquid outlet machine of the store during execution of the current order information are collected; and generating new pipeline task information corresponding to the new order information of the store through a pipeline task adjustment strategy, replacing the current order information with the new order information, and returning to execute the above steps according to the new pipeline task information and the replaced pipeline task information. By adopting the method, the product meal delivery efficiency and meal delivery accuracy of the intelligent liquid delivery machine can be improved.

The invention discloses a cross-platform task adaptation method. The method comprises the steps of obtaining a to-be-operated task; based on multiple pieces of preset operation cost gradient information corresponding to each target hardware acceleration platform in the multiple target hardware acceleration platforms, determining an evaluation value of the operation cost of executing the to-be-operated task by each target hardware acceleration platform; each piece of preset operation cost gradient information corresponds to one operation action, and each piece of preset operation cost gradient information represents the operation cost when the corresponding target hardware acceleration platform executes the same operation action of different data scales; the computing cost represents the efficiency and / or computing power utilization rate when the target hardware acceleration platform executes the to-be-computed task; and determining the target hardware acceleration platform for executing the to-be-operated task based on the evaluation value of the operation cost of executing the to-be-operated task by the target hardware acceleration platform.

The application provides a full-process dynamic task planning AI agent scheduling method and system for SEO, relates to the technical field of agent scheduling, and preconfigures rule configuration information after building an agent core architecture, connects the preconfigured rule configuration information to a plurality of LLM models and a full-network information collection interface, and obtains preparation architecture building information; the preparation architecture building information is used for creative task analysis through a retrieval hub, creative task analysis data is obtained, the creative task analysis data is analyzed for task allocation through the retrieval hub, task allocation analysis data is obtained; the keywords, outline and content are sequentially generated and verified and analyzed through the multi-agent architecture building information, and agent scheduling data is obtained, the application effectively solves the contradiction between the plurality of LLM models and task adaptation and the problem of scheduling hub computing power overload under high concurrency, breaks the vicious circle of the two, does not require manual intervention, and greatly improves the full-process automation level of SEO content creation.

The invention belongs to the field of machine learning, and particularly relates to a method and system for supplementing task adaptation of a large model and a readable storage medium. The method comprises the following steps: inputting task data to be processed into a specific layer; the specific layer comprises at least two selected specific small models respectively fitting different distributions; the specific small model is a small model trained by using a training set corresponding to the task to be adapted; judging whether the confidence coefficient of the input task data processed by the corresponding specific small model is greater than or equal to a first set confidence coefficient threshold value or not according to the sequence of the performance of the specific small model in the specific layer from high to low, if so, adopting a processing result of the specific small model, and if not, continuing to judge; and if the confidence coefficient processed by no specific small model in the specific layer is greater than or equal to a first set confidence coefficient threshold value, processing the input data by a large model or an enhancement layer. Tasks difficult to adapt to the large model are supplemented through the small model, and parameters of the large model do not need to be modified.

The invention discloses a space-time big data processing method based on a pre-training-fine tuning architecture, and solves the problems of complex relation and high calculation cost when a traditional method processes space-time multi-attribute data by combining a pre-trained Transform model and a prompt fine tuning mechanism. Specifically, the model framework comprises a time encoder, a space encoder and a head network, and common features among a plurality of time-space attributes are efficiently captured in a parameter sharing mode. In the pre-training stage, the model learns a general spatio-temporal pattern in a large-scale spatio-temporal data set, and then fine tuning is performed on specific spatio-temporal attributes by using a trainable spatio-temporal prompt token through a prompt fine tuning stage, so that efficient task adaptation and low calculation cost are realized. The method not only can improve the space-time prediction precision, but also has good mobility, adapts to unseen space-time attributes, and has wide application prospects.

The invention discloses a cooperative tray scheduling control method based on multiple AGVs (Automatic Guided Vehicles). According to the method, intelligent task release is realized by establishing a digital identifier for a tray and calculating a task priority; each AGV evaluates the ability of itself and obtains a task through distributed bidding; when the single AGV is insufficient in capability, dynamic alliance collaborative operation is formed; the alliance plans a path through a space-time diagram and eliminates conflicts; during execution, a hierarchical control architecture is adopted to keep synchronization; after the task is completed, performance is evaluated, and the alliance is disintegrated; and the system continuously monitors the health state of the AGV and realizes predictive maintenance. Through three innovations of tray intelligence, dynamic alliance and predictive maintenance, the limitation of traditional centralized scheduling is broken through, the system robustness, task adaptability and operation reliability are improved, and an efficient technical scheme is provided for intelligent logistics.

The invention discloses a robot strategy learning method and a related device, and the method comprises the steps: constructing a training data set based on a sharing control framework of teleoperation; the training data set is utilized, a course training method for gradually weakening visual input disturbance is adopted, the robot strategy is trained and learned, the trained and learned robot strategy is obtained, the method and the related device can improve the data collection efficiency and naturalness, and meanwhile the generalization ability and the task adaptability of the robot strategy are improved.

This invention discloses a method, apparatus, device, and storage medium for dynamic allocation of warehousing tasks. The method includes: receiving a warehousing task request; determining a target task to be processed based on the warehousing task request; determining multiple candidate execution devices currently available for executing the target task based on preset resource status data; determining a task adaptation value corresponding to each candidate execution device based on the task attributes of the target task and the device attributes of the candidate execution devices; selecting a candidate execution device that meets preset conditions as the target execution device based on the task adaptation value, and issuing the target task to the target execution device. Because this invention determines the task adaptation value corresponding to each candidate execution device based on task attributes and device attributes, and then selects the target execution device based on the task adaptation value, compared with the prior art, this invention improves the scientific nature of warehousing task allocation, thereby improving the overall warehousing operation efficiency and system resource utilization.

The invention relates to the technical field of artificial intelligence and man-machine interaction training, in particular to a method and system for constructing a cognitive training configuration scheme. The method comprises the steps of completing cognitive state initialization and constructing a task feature model based on initial behavior data, historical task performance and interaction preference parameters of a pilot; based on the cognitive state and the task feature model, constructing a guide process and generating a man-machine interaction behavior sequence; operation strategies and intervention behaviors in the task process are collected, and a training process set is constructed; generating a strategy evolution trajectory according to the strategy distribution of the training process set and the failure label; then path analysis and behavior attribution are carried out, a meta-cognitive behavior characteristic index set is extracted, task configuration and interaction nodes are optimized in combination with the results, and a cognitive training configuration scheme is generated. According to the method, the change process of the cognitive state of the pilot can be dynamically captured, a task adaptation mechanism driven by cognition is constructed, and the element cognitive ability of the pilot and the task strain ability under the complex situation are improved.