187 results about "Action plan" patented technology

Devices and techniques are generally described for security threat mitigation for generative machine learning models. In some examples, first prompt data including first data associated with a first natural language input and a first span may be determined. An LLM may determine first plan data using the first prompt data. The first plan data may include a call to the first API. A first classifier model may determine a first trust score for the first span. A first attention score may be determined for the first span and the first action plan. Second plan data may be generated based on at least one of the first trust score and the first attention score or the second trust score and the second attention score.

Disclosed are a method, system, and apparatus of response plan modification through artificial intelligence applied to ambient data communicated to an incident commander. According to one embodiment, the method includes analyzing a description of a fire in progress, automatically generating an incident action plan through an Incident Action Artificial-Intelligence Model (“IAAIM”) based on the description of the fire in progress, and modifying the incident action plan based on a trusted radio communication to an incident commander.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a multi-level memory collaborative decision-making method, device, equipment and medium, and the method comprises the steps: obtaining environment observation data, and updating a multi-level memory module; retrieving the memory generation candidate communication content; high-level actions are extracted according to the task target, and a to-be-selected action plan is formed; reasoning and analyzing the action plan and the communication content through a language model, and determining whether to send; if so, incorporating into a to-be-executed action set; and generating a final plan based on the action plan and the action set, and decomposing the final plan into basic operation instructions to interact with the environment. According to the method, the environment, history and instructions are processed through the reasoning ability of the language model in combination with multi-level memory, the collaborative plan and the communication strategy of the intelligent agent are optimized in the environment with limited communication and incomplete information, and the decision consistency and the task execution efficiency are improved.

The present disclosure relates to systems, non-transitory computer-readable media, and methods for using a large language model and digital content items (e.g., digital files) stored in a content management system to generate workflow outputs for external computing platforms. For instance, in some embodiments, the disclosed systems receive a workflow request that includes a natural language description of an objective to be accomplished via a workflow. The disclosed systems determine an action plan for completing the workflow and executes the action plan by accessing digital content items (e.g., digital files) stored in a content management system and extracting or generating data from the contents of the digital content items. Further, the disclosed systems use a large language model to generate a workflow output that is provided to an external computing platform for use.

The invention relates to a large model and small model collaborative robot operation action real-time control method and a large model and small model collaborative robot operation action real-time control system. The method comprises the following steps: firstly, collecting and preprocessing multi-modal data; carrying out data coding on the multi-modal data to obtain a feature vector; aligning the multi-modal data by using a cross-modal token alignment technology to obtain a feature representation; then training a neural network model by using the feature vector and the feature representation to obtain a trained large model; pruning, distilling and quantifying the trained large model to generate a small model; the small model is deployed to an edge computing device, the edge computing device obtains current instruction text data in real time and conducts reasoning on the current instruction text data, and an action plan of the robot is generated; and finally, the robot is controlled to operate and act in real time by means of action planning of the robot and feedback signals of the sensors. Compared with the prior art, the method has the advantages of low delay, high efficiency, capability of improving tactile perception, self-adaptive grabbing and real-time action planning and the like.

The invention relates to the technical field of artificial intelligence and natural language processing, in particular to an agent system based on a large language model and an interaction method, and the system comprises a data collection module which is used for receiving text, voice or other forms of data input by a user in real time; the natural language processing module is used for performing semantic understanding and grammatical analysis on the received data to generate an executable user intention; the knowledge reasoning module is used for reasoning an optimal response result or action scheme according to the requirements of the user and historical interaction records; the method has the beneficial effects that the interaction experience of the user is greatly improved through natural language processing, sentiment analysis and personalized response technologies. The system can perform accurate response according to the emotion and demand of the user, and can provide efficient and intimate services in applications such as intelligent customer service, intelligent assistant, personalized recommendation and the like.

The invention discloses a security reinforcement learning-driven large language model security decision agent, and the decision agent comprises a high-level semantic planner which is used for receiving a target and constraint instruction in a text form, receiving a language or visual observation signal of an environment at the same time, and outputting text formatted security risk information and suggested action planning; the low-layer action actuator is used for receiving low-dimensional observation and semantic codes of the environment, and the semantic codes are output by the high-layer semantic planner after text embedding conversion; the strategy network of the low-layer action actuator outputs a final safety action; the training alignment module is used for optimizing the strategy network and the value network; a high-level semantic planner is fed back and prompted through reward and cost signals collected through environment interaction, and parameters of a strategy network and a value network are trained through a security reinforcement learning algorithm. According to the method, the decision cannot violate the given text security constraint while the decision of the given text target is completed.

The invention discloses an action scheme computable modeling method and system, particularly relates to the field of simulated confrontation, is used for solving the problem of multi-platform collaborative path planning in a dynamic environment, takes a multi-constraint hierarchical hypergraph as a core, carries out fusion abstraction on space, time and resources, and compresses a search domain through convex hull relaxation expansion and gradient sampling so as to realize multi-platform collaborative path planning. Meanwhile, a stable factor is dynamically adjusted by means of resource phase mapping and quantum taboo migration, and it is ensured that a global optimal solution is rapidly approached; fine time sequence correction is further achieved through symplectic time window rearrangement and conjugate gradient slippage, the survival potential and coverage confidence coefficient of the path are verified through adversarial prediction simulation, and the reliability and safety of the scheme under the disturbance of a frequently-changing three-dimensional geographic space, a dynamic forbidden area and moving obstacles are enhanced; therefore, strict time sequence and resource requirements of the multi-platform collaborative task are effectively met, and advanced computational modeling support is provided for the multi-platform collaborative task.

The invention relates to the technical field of robots, and discloses a task planning method and system for a robot, and the method comprises the steps: obtaining a user instruction; based on the large language model and the semantic coding model, retrieving in a memory bank to obtain environment perception information related to the user instruction; inputting a first cue word determined based on the user instruction and the environment perception information into the large language model, and generating an overall action sequence corresponding to the user instruction; and for each sub-action in the whole action sequence, determining a second cue word, inputting the second cue word into the large language model, and generating a detailed action plan corresponding to the current sub-action until each sub-action in the whole action sequence is traversed. According to the method, the natural language or unstructured instruction can be received, the intention of the user can be understood, and the executable overall action sequence and the detailed action plan of each sub-action in the sequence are generated, so that the task issued by the user is completed, it is ensured that the robot can execute the task, and the accuracy and efficiency of task execution are improved.

Embodiments of the present disclosure provide an alert analysis method, a server, and a storage medium. In the alert analysis method, once a target question associated with a target security alert of a user resource is obtained, a security model is used to perform action planning for the target question to obtain action planning information. On the basis of tool invocation information in the action planning information, a target alert handling tool is invoked to perform an alert information processing operation associated with the target question to obtain an alert information processing result. On the basis of the alert information processing result, the security model can output an alert analysis result corresponding to the target question. On the one hand, the security knowledge capability of the security model is used to realize automatic replies to security alert related questions; on the other hand, by using the action planning capability of the security model, an alert handling tool is invoked to process the alert information associated with the target question, so that security knowledge beyond the security model can be fully utilized, thereby effectively improving the accuracy of the alert analysis result corresponding to the target question.

Methods, systems, and non-transitory computer readable storage media are disclosed for generating action plans utilizing a large language model with a best-first search model. The disclosed system determines a request to utilize a large language model to generate an action plan via one or more software tools. The disclosed system generates the action plan by traversing a decision tree comprising an action space involving the one or more software tools by iteratively: selecting, utilizing a best-first search model, an action from a set of possible actions in the action space of the decision tree; and expanding, utilizing the best-first search model, the action space of the decision tree to include an additional set of possible actions. The disclosed system also executes the action plan via one or more interactions with the one or more software tools according to the action.

The invention relates to the technical field of robots, and discloses a heterogeneous multi-machine cooperation method, a terminal and a readable storage medium, and the method comprises the steps: obtaining environment perception data, and obtaining a multi-modal fusion scene prompt according to the environment perception data; distributing a role for each agent according to the multi-modal fusion scene prompt, and generating an action plan of each agent according to the multi-modal fusion scene prompt and the role of each agent; and controlling each agent to execute the respective corresponding action plan. According to the method, the multi-modal fusion scene prompt is extracted according to the environment perception data, the parameters of different scenes are matched according to the multi-modal fusion scene prompt, the role of each agent is allocated through the role allocation algorithm, and the method does not depend on predefined roles and the parameters specific to the scenes; and the task execution capability of the system in a dynamic scene is improved.

The invention discloses an unmanned vehicle optimal action sequence generation system and method based on cooperation of a large model and a special model, and relates to the technical field of unmanned vehicle control. In the system, after a task understanding and decision vertical large model generates cue word segments based on instruction text information, scene state information and a task planning knowledge base, the cue word segments are divided into a plurality of cue word segments; decomposing the received task into various sub-tasks for task planning, and generating a plurality of action strategies and specific parameter calling information corresponding to the various sub-tasks, so as to obtain multiple types of action schemes; calling corresponding special model analytic operation based on specific parameter calling information of the special model of each action strategy in the current scheme, and dynamically integrating each analytic operation result to obtain an action sequence of the current scheme; and based on the action sequence of each action scheme, determining a comprehensive evaluation value of each action scheme so as to determine an optimal action sequence. The accuracy, the efficiency, the adaptability and the automation degree of generating the action sequence of the unmanned vehicle are improved.

The invention discloses an action management identification method based on earthquake site disaster investigation, and relates to the technical field of data processing, and the method comprises the steps: determining a to-be-acted earthquake site region, and planning a disaster search task corresponding to a plurality of unmanned aerial vehicles in the to-be-acted earthquake site region; issuing the disaster situation search task to multiple unmanned aerial vehicles, controlling the unmanned aerial vehicles to execute the disaster situation search task, and obtaining an actual disaster situation investigation result corresponding to the disaster situation search task; according to an actual disaster situation investigation result, generating disaster situation points on the electronic map corresponding to the earthquake field area to be moved, and obtaining a disaster situation investigation map; and finally, the disaster situation investigation map is fed back to equipment specified by a commander to guide disaster relief actions of the commander, so that the earthquake site disaster situation investigation efficiency is improved, comprehensive disaster situation data can be formed, the commander can comprehensively understand the disaster situation, the commander is assisted to quickly form a global and optimal action scheme, and the disaster relief efficiency of the commander is improved. And the disaster disposal efficiency is improved.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as intelligent, financial science and technology, medical health and the like, and discloses a multi-modal fusion feature driven action control method, device, equipment and medium, and the method comprises the steps: obtaining first modal input information and second modal input information, the method comprises the steps of generating a corresponding first modal feature vector and a corresponding second modal feature vector, fusing the two modal feature vectors to generate a multi-modal fusion feature, generating an action instruction based on the multi-modal fusion feature, generating an initial action plan in combination with the current state of equipment, the current environment information and a task target, and performing motion planning. And generating a global optimal action sequence based on the initial action plan, and controlling the equipment to execute the global optimal action sequence. According to the method, the global optimal action sequence is generated and equipment execution is controlled by combining the action instruction generated by the multi-modal fusion information with the equipment state, the environment information and the task target, so that the decision-making effect and the dynamic adaptive capacity in a complex environment are improved.

The invention discloses a distributed computing power scheduling method and system based on AIGC, and relates to the field of computing power scheduling, first, GPU index flow and request index flow are obtained in real time, time sequence alignment and feature engineering are carried out on the GPU index flow and the request index flow, and meanwhile semantic feature extraction is carried out on a business intention text; and fusing the processed features into a vectorization system state, inputting the vectorization system state into a generative AI scheduling agent to predict a future model demand, and outputting a model demand prediction JSON (JavaScript Object Notation). Then, the planner module generates an optimized cache action plan according to the prediction and the current cache state, and instructs a scheduling executor to pre-load or intelligently replace the model. And finally, when the request router receives a new request, the updated cache state can be queried, and the request is directly distributed to the GPU node loaded with the target model, so that the cold start delay is remarkably reduced, and the computing power resource utilization efficiency and the user experience are improved.

The invention discloses a mechanical arm control method and system based on imitation learning, and belongs to the field of intelligent robots and automatic control. The data collection layer designs a data collection method based on a VR platform, and is responsible for collecting teaching data to make a complex scene data set, performing coordinate conversion to solve angle information of each joint of the mechanical arm, realizing VR teleoperation teaching and teaching data collection, and transmitting the teaching data to the ACT algorithm layer for multi-modal image fusion; the multi-modal image fusion ACT algorithm layer carries out feature extraction and fusion denoising on complex illumination scene data, then infrared and visible light information fusion complementary features are combined with an ACT algorithm to obtain predicted action parameters, the predicted action parameters are converted into corresponding action plans through a motion control strategy, and a mechanical arm is driven to autonomously complete tasks. The method solves the problems that the performance of a mainstream imitation learning algorithm is limited in a complex illumination environment, the motion of the mechanical arm is prone to deflection and the like, and the perceptual performance and the capacity of adapting to complex scenes of the mechanical arm under the complex illumination condition are remarkably improved.

The present application discloses a distributed computing power scheduling method and system based on AIGC, which relates to the field of computing power scheduling. First, the GPU indicator stream and request indicator stream are obtained in real time, and time series alignment and feature engineering are performed on them. At the same time, semantic feature extraction is performed on the business intent text, and the above-mentioned processed features are fused into vectorized system states and input into the generative AI scheduling agent to predict future model requirements and output the model requirement prediction JSON. Subsequently, the planner module generates an optimized cache action plan based on this prediction and the current cache status to guide the scheduling executor to pre-load or intelligently replace the model. Finally, when the request router receives a new request, it can query the updated cache status and directly distribute the request to the GPU node that has loaded the target model, thereby significantly reducing the cold start delay and improving the computing power resource utilization efficiency and user experience.

The application discloses a safe decision-making agent of a large language model driven by safe reinforcement learning, and the decision-making agent comprises: a high-level semantic planner, which is used for receiving target and constraint instructions in a text form, receiving language or visual observation signals of an environment, and outputting safe risk information and a suggestion action plan in a text format; a low-level action executor, which is used for receiving low-dimensional observation and semantic encoding of the environment, wherein the semantic encoding is output by the high-level semantic planner after text embedding conversion; a strategy network of the low-level action executor, which outputs a final safe action; a training alignment module, which is used for optimizing the strategy network and a value network; reward and cost signals collected through environment interaction are fed back to the high-level semantic planner, and parameters of the strategy network and the value network are trained through a safe reinforcement learning algorithm. The application is convenient for realizing the decision-making of a given text target while ensuring that the decision-making does not violate a given text safety constraint.

To provide a system for providing, for improving productivity of a meeting, means for creating a minute and a summary, means for creating meeting graphic recording, means for creating an action list, and means for creating a meeting productivity score.SOLUTION: A system, that is an AIaaS product for improving productivity of a meeting, includes: means for creating a minute and a summary, means for creating meeting graphic recording, means for creating an action plan, and means for creating a meeting productivity score.SELECTED DRAWING: Figure 1

The invention provides an enterprise safety production risk assessment system, which comprises a task library module used for acquiring and screening internal and external operation tasks of an enterprise area. And the step module is used for carrying out step decomposition on each operation task according to an execution function. And the hazard module is used for identifying hazard information in the enterprise operation process. And the risk analysis module is used for identifying the risk possibly caused by the hazard and analyzing the process sequence of converting the risk into the risk and causing the accident. And the risk assessment and grading module is used for carrying out element valuing, calculation and grading processing on the risk. And the control measure module is used for determining a risk response mode and a control measure. And the action plan module is used for generating an action plan and prompting the implementation condition of the risk control measures to the user. According to the system, the whole-process informatization operation of operation risk assessment is realized, the relevance between an accident sequence and a protective measure is particularly deeply analyzed, and the accuracy of risk assessment and the pertinence of a control measure are improved.

The invention relates to the technical field of atlas construction and recognition, in particular to a teenager psychological question recognition and intervention method fusing a teaching atlas and grading, which comprises the following steps: S1, data acquisition is carried out, a family system atlas is constructed, and the family system atlas represents family members and family rules through nodes, representing the relationship strength between the members and the rule execution consistency probability through edges; s2, jointly inputting the home system atlas and user question description input in real time into a situation-risk joint inference model, and outputting an enhanced question identifier fusing a situation type, a risk level and a most fragile relationship path; and S3, taking the most fragile relationship path as a core retrieval key, retrieving an intervention strategy from the strategy knowledge base, and generating a targeted micro-action scheme. According to the method, the family system atlas is constructed, and the most fragile relation path mechanism is introduced, so that the structural modeling and context awareness inference of the family structure and relation dynamics are realized.

Providing technological means to assist human deductive thinking in considering business ideas. [Solution] The business plan consideration support device 1 of the present invention comprises a syntax input unit 111 that inputs sales syntax related to business plan support into a generation AI, a gap measurement unit 112 that measures the gap between the sales target related to the business plan and current sales, and an AI support unit 113 that causes the generation AI to generate an action plan to fill the gap. The syntax input unit 111 inputs a sales syntax (Y=P×N×F×C) that breaks down the sales target (Y) into elements including average customer spending (P), number of customers (N), purchasing frequency (F), and execution ability (C) into the generation AI, and the AI ​​support unit 113 generates an action plan based on the sales syntax, including goals related to average customer spending, number of customers, purchasing frequency, and execution ability.

The invention provides a transformer area carbon reduction potential evaluation method considering user side interaction response. The method comprises the following steps: obtaining transformer area node carbon potential according to transformer area power data and a sampling time period; obtaining user side carbon emission according to the transformer area node carbon potential, the user side power data and the sampling time period; obtaining a new energy output ratio average value and an energy storage equipment capacity ratio according to the transformer area power data; according to the user-side power data, obtaining the power supply reliability and the user comfort of the transformer area; obtaining a user side interaction response capability according to the user side carbon emission; according to the user side interaction response capability, the new energy output ratio average value, the energy storage equipment capacity ratio, the transformer area power supply reliability and the user comfort, constructing an evaluation matrix of the transformer area to be evaluated; and selecting the zone area with the highest carbon reduction potential based on the evaluation matrix. According to the evaluation method provided by the invention, the carbon reduction potential of the transformer area can be scientifically and comprehensively evaluated, and decision support is provided for making effective carbon emission reduction strategies and action plans.

An information processing apparatus includes a crowd characteristic estimator and a plan controller. The crowd characteristic estimator estimates information regarding a characteristic of a crowd on the basis of a sensing result of an external environment. The crowd is a collection of individuals present in the external environment. The plan controller controls a creation mode of an action plan for a movable body in the external environment on the basis of at least the information regarding the characteristic.

The invention discloses a business building updating evaluation and treatment system, and the system comprises a data collection module which is used for obtaining the position information, building attribute data and operation attribute data of a target business building; the zone bit attribute data generation module is used for generating zone bit attribute data; the location evaluation module is used for analyzing the location attribute data and generating a location score; the building evaluation module analyzes the building attribute data and generates a building score; the operation evaluation module is used for analyzing the operation attribute data and generating an operation score; the updating demand calculation module is used for calculating a building updating demand level according to a formula; the visual report generation module is used for generating a digital report containing a comprehensive evaluation matrix graph, a diagnosis conclusion and an updating strategy suggestion; and the governance scheme recommendation module is used for outputting a customized action scheme. According to the invention, through integration of position, building and operation data, a standardized evaluation framework is established, and multi-dimensional quantitative analysis of building values is realized.

A large language model (LLM) system includes a personal AI agent, an enterprise AI agent, and a public AI agent. A method for operating the LLM includes: executing a first action plan to obtain a first integrated answer and a first integrated score based on a first query using the personal AI agent; and executing a second action plan to obtain a second integrated answer and a second integrated score based on a second query using the enterprise AI agent. A method for applying the LLM system includes: transforming a bitmap of a block diagram into a scalable vector graphics and integrating a semantic tag into the scalable vector graphics; training the large language model by using the scalable vector graphics integrated with the semantic tag and using a functional specification document; and generating a structured text description of the block diagram using the large language model which is trained.

This invention relates to the field of emergency response simulation technology, specifically disclosing a highly efficient emergency response plan optimization and intelligent reasoning method and system. The method includes: real-time access to data streams from physical sensors, social media, and public databases to construct a panoramic view of the emergency situation; constructing an emergency knowledge graph based on the panoramic view and a pre-defined domain ontology; identifying graph substructures that conform to pre-defined graph features within the emergency knowledge graph, and reading the corresponding emergency action plans to generate at least one comprehensive plan; performing deduction and optimization on the generated comprehensive plan, selecting and applying the optimal action plan, monitoring the application data of the optimal action plan in real time, and recursively updating the deduction process. This invention standardizes the data stream and determines the knowledge graph reflecting the scenario state based on the standardization results, resulting in highly readable data and extremely high efficiency in outputting emergency response plans.

A system is provided.SOLUTION: A system comprising: means for acquiring energy and water usage data in real time from sensors installed in each home or business; means for cleaning the acquired data, removing noise, and converting the data into a format suitable for learning; means for training a model using the cleaned data and generating an algorithm for proposing an optimal resource management method for each home or business; means for generating a specific action plan based on an analysis result and delivering the action plan to a user; and means for collecting feedback from the user and retraining the model.SELECTED DRAWING: Figure 1

The invention relates to the technical field of new energy charging piles, in particular to a loaning distribution strategy based on a high-power platform semi-matrix model. A charging system is divided into a group A and a group B, a full-matrix model is adopted in each group, semi-matrix connection is formed between the groups through DC + / DC-double contactors, and cross-group power borrowing is achieved. The method comprises the following specific steps: calculating a preliminary power module demand table, generating and arbitrating a borrowing demand, and generating an action plan table after redistributing modules. Through priority dynamic allocation and an arbitration rule (priority is high and under-power magnitude is high), dynamic power demand conflicts of different vehicles and the same vehicle in different charging stages are balanced, and the module utilization rate and the charging efficiency are improved. Experiments show that the strategy improves the overall charging efficiency, simplifies the module switching logic, enhances the system stability, and gives consideration to the guarantee of high-priority users and the rights and interests of low-priority users.