31 results about "Safety policy" patented technology

The invention relates to the technical field of smart parking lots, in particular to a vehicle safety management method based on a smart parking lot, which comprises the following steps: S1, multi-source data acquisition: acquiring dynamic trajectory data of a target vehicle, parking lot environment parameters and adjacent parking space state data in real time, and generating a multi-source feature vector set; s2, data space-time alignment: carrying out space-time alignment processing on the multi-source feature vector set, and outputting a standardized vehicle behavior sequence; s3, abnormal behavior recognition: inputting the standardized vehicle behavior sequence into an abnormal behavior recognition model, and generating an abnormal behavior characteristic spectrum and a confidence coefficient parameter; s4, risk level assessment: constructing a risk assessment model based on the confidence coefficient parameter and the environmental parameter, and outputting the risk level of the parking lot in the current scene; and S5, security policy triggering: triggering a corresponding security policy execution instruction according to the risk level. According to the invention, potential risks can be effectively prevented, and the safety of vehicles and personnel is ensured.

The invention discloses an aircraft voice instruction security analysis method and system based on natural language processing, and aims to realize intelligent identification and execution control of aircraft operation instructions. The method comprises the steps of receiving a voice instruction audio input by an aircraft operator, and performing preprocessing to generate a standardized voice segment; converting the standardized voice segment into text content, and extracting structured semantic information related to the instruction; calculating a semantic intention confidence score of the instruction in combination with the structured semantic information, the aircraft operation state, the historical instruction and the current task mode; according to the confidence score and the security policy rule, jointly evaluating the security of the instruction and outputting an executable state identifier; when the state identification is executable, calculating a flight control command and submitting the flight control command to a flight control system for execution; and when the state identifier is non-executable, the interception operation is automatically executed, and abnormal prompt information is fed back to an operator through the man-machine interaction module.

The invention relates to the technical field of intelligent management and control of worker behaviors, and discloses an intelligent management and control method and system for worker behaviors based on multi-modal sign recognition, and the system is composed of a multi-modal sensor array and a behavior analysis server which are arranged at set intervals in an operation area. The method comprises the steps that a sensor array collects physical sign state and environment interaction data, and a behavior characteristic data set is generated; the server receives the real-time sign fusion data to construct a matrix, extracts mode features in combination with a data set after processing, predicts a behavior mode according to a physiological state change gradient and an action trajectory, outputs time-space distribution features through a behavior association model, updates the data set, and determines classification features; and dynamically configuring a security policy according to the classification characteristics, wherein the security policy comprises execution actions such as region identifier mapping and control protection equipment activation, and dynamically allocating protection resources. According to the scheme, multi-modal data fusion analysis and security policy dynamic adjustment are realized, and the intelligence and accuracy of operation security management and control are improved.

The invention discloses a railway marshalling station intelligent scheduling control method and system based on hierarchical reinforcement learning, and belongs to the technical field of rail transit. In the hierarchical reinforcement learning framework, the upper layer achieves a global marshalling priority strategy by using a deep Q network trained in stages and by means of dynamic reward function design; and the lower layer optimizes a PPO algorithm by means of a near-end strategy of dynamic action space cutting, and regulates and controls the state of the debugging equipment in real time. A security policy distillation technology is adopted in a marshalling station scheduling control process, and a security policy passing verification is extracted as a rule tree to serve as an initial policy for reinforcement learning, so that a training convergence process is accelerated. A transfer learning method is further applied, so that the method can adapt to the dynamic flow distribution requirements of marshalling stations of different scales. By adopting the technical scheme of the invention, the physical limitation and the operation safety standard of the railway marshalling station can be met as far as possible.

The invention discloses an off-line constraint reinforcement learning method for double-commenter regularization and dynamic coefficient scheduling based on a decision converter, and relates to the technical field of machine learning and intelligent control. The method is oriented to a safe and sensitive robot control task, and historical offline trajectory data is used as a training sample; on the basis of a backbone network of a constraint decision converter, a reward reviewer and a cost reviewer are introduced, and explicit modeling is carried out on state-action long-term return and accumulated cost; designing a joint loss function including optimal trajectory weighting, action prediction loss, a return consistency regular term and a cost consistency regular term, and dynamically scheduling a regularization coefficient by adopting a two-stage cosine annealing strategy; and by alternately optimizing parameters of the commenter and the decision converter, a security policy model considering high return and low cost is obtained through training. The model receives the current observation state of the robot and outputs a high-return motion control instruction meeting the cost constraint.

The present application discloses a safety monitoring and management method and system for an intelligent box with an integrated lock, which is used to enhance the safety protection ability of the intelligent box and eliminate potential safety hazards. The method of the present application includes: the edge computing unit obtains multi-modal sensor data of the box in real time; the edge computing unit performs real-time analysis on the multi-modal sensor data according to a preset abnormal threshold to detect and respond to single-point abnormal events of the box; the edge computing unit performs local primary feature extraction on the multi-modal sensor data and uploads the extracted primary feature vectors to the fog computing node; the fog computing node performs high-level feature extraction and spatio-temporal correlation analysis on the primary feature vectors uploaded by all boxes in the regional network to identify whether there are group safety risk events; if so, the fog computing node dynamically adjusts the safety policy level of all boxes in the regional network to the highest level.

Embodiments of the present invention provide computer-implemented methods, computer program products and computer systems. Embodiments of the present invention can identify a plurality of constraints on states of data and actions of data associated with a data model. Embodiments of the present invention can then identify constraints on safety policy parameters associated with a computing device. Embodiments of the present invention can then convert the identified constraints into a uniform domain syntax that considers coupled and decoupled constraints and introduce buffer data within the converted constraints, wherein the buffer data filters outlier constraints within the plurality of constraints. Embodiments of the present invention can then dynamically generate optimal safety policies associated with the computing device based on the remaining constraints.

The invention relates to a multi-target identification sensor array safety monitoring method, and aims to improve the safety perception and target identification capability in a complex environment, and the method forms an array by deploying at least two different types of sensors such as acoustic, vibration, electromagnetic, infrared and optical sensors. After collected environment data and target signals are preprocessed, high-representativeness fusion feature vectors are generated through multi-dimensional feature extraction and a dynamic fusion strategy. And in combination with multi-level thresholding methods such as a dynamic threshold and a fixed threshold, a potential target signal is accurately extracted and a preliminary event is generated. A hierarchical storage mechanism is adopted, a large table and a small table are created, high-confidence events are quickly retrieved through the small table, complete context information is reserved in the large table, and discrete events are aggregated into composite events through space-time window association analysis. And a multi-target classifier based on machine learning / deep learning performs type identification and attribute analysis on the composite event, and finally completes threat assessment and early warning in combination with a preset security policy library.

Disclosed in embodiments of the present disclosure are a method and apparatus for applying a large language model. A specific implementation of the method comprises: inputting an original question to be queried and a preset prompt into a large language model, and outputting a first question, wherein the prompt is used for instructing the large language model to extract, from the original question, a question reflecting a real intention; deleting non-key information from the first question to obtain a second question; and inputting the second question and a preset security prompt into the large language model, and outputting an answer, wherein the security prompt is used for indicating that the large language model needs to comply with a security policy when answering a question. The implementation can effectively defend against jailbreak attacks having deceptive and implicit intentions.

The invention discloses a research, development and operation method and device for security left shift, and relates to the field of software development processes and security management.The method comprises the steps that a security demand baseline is established in the demand analysis stage, security demands and business demands are fused, and risk protection measures are determined; security architecture design review is carried out in the architecture design stage, and a security team carries out threat modeling on the system architecture and carries out one-ticket negative right; in the code development stage, code specifications, open source component vulnerabilities and logic risks are scanned in real time through a static code analysis tool, and a code security closed-loop management mechanism is established; in the compiling and constructing stage, code review and component list filing are implemented, a security review report is formed by combining a dynamic penetration test and a running flow test, and a security strategy is updated through a continuous monitoring mechanism to deal with new threats. Security practice can be systematically integrated in the early stage of the software development life cycle, and the later vulnerability repair cost is remarkably reduced.

This invention discloses an adaptive control method for bearing grinding processes based on reinforcement learning, comprising: S1, real-time acquisition of multi-source state data; S2, construction of a high-dimensional state vector; S3, construction of a historical grinding dataset; S4, construction of an improved IQL model incorporating an uncertainty perception mechanism, offline training using the historical dataset, outputting the Q-value distribution, and learning a safety policy at the safety policy layer; S5, online deployment of the model, generation of candidate commands, and evaluation of uncertainty; S6, judgment by the safety constraint layer: issuing candidate commands when uncertainty is low, and invoking the safety policy to generate safe commands when uncertainty is high; S7, execution of commands and storage of new data in the dataset, periodically optimizing the model. This invention improves the safety and reliability of online decision-making through uncertainty quantification and a safety policy safety net, achieving adaptive and precise control of complex grinding conditions.

The invention discloses a multi-mode driving safety monitoring system based on a hydrogel electrode, and relates to the technical field of driving safety monitoring. The system comprises an edge device and a cloud, wherein the edge device comprises a synchronous acquisition module, a real-time monitoring module, a security policy module and a channel detection module; through high-precision data acquisition and synchronization, each physiological and environmental signal is ensured to be accurately aligned; the state change of the driver is captured from multiple dimensions, early warning is carried out in time, the accident risk is reduced, and comprehensive real-time monitoring and multi-dimensional evaluation are achieved; standby channels are switched in time during channel failure detection, and continuous and stable signal acquisition is ensured; the adaptability of the system to a new driving scene is continuously improved through cloud data management and incremental learning, and model recession is prevented; implementing; and carrying out continuous self-adaption and model iterative optimization. Generally speaking, the system not only can monitor and warn the driving fatigue state in real time, but also can continuously improve the monitoring precision and the system stability through the closed-loop feedback of the edge and the cloud, thereby providing guarantee for the driving safety.

The invention discloses a physical fresh-keeping abnormal working condition self-repairing system based on edge intelligence, and relates to the technical field of self-adaptive adjustment. The method comprises the following steps: a feature extraction module which performs processing based on operation state data collected by a fresh-keeping container in unit time to generate operation state feature data; the frost blockage working condition diagnosis module is used for inputting the running state characteristic data into a pre-training model and outputting a frost blockage grade result; the defrosting trigger module is used for generating and issuing a defrosting control instruction according to a safety strategy when severe frost blockage occurs; the pseudo-steady-state diagnosis module is used for numeralizing a diagnosis result under mild or moderate frost blockage, inputting the diagnosis result into a pseudo-steady-state diagnosis model in combination with the running state characteristic data, and identifying whether a pseudo-steady state caused by frost blockage exists or not; the adaptive adjustment module generates an optimal adjustment parameter set based on a genetic algorithm and issues and executes the optimal adjustment parameter set; according to the method, the frost blockage of the fresh-keeping container under the complex working condition and the pseudo-steady state caused by the frost blockage can be sensed, diagnosed and repaired in the whole process.

The invention discloses a vehicle multi-risk perception and hierarchical automatic takeover system and method, which can integrate multi-dimensional monitoring of vehicle state, driver state and environment information through a condition identification module, cover multiple dangerous scenes, break through the limitation of single function of the existing system, and effectively meet the safety protection requirement under a composite dangerous situation. Through accurate positioning of a tire burst position and a special dynamic control model, a customized stabilizing strategy is executed according to different tire burst conditions of left front tires, right front tires and the like, the out-of-control probability of the vehicle after tire burst is reduced, secondary accidents are reduced, and the emergency control stability is improved; a user-defined security policy is supported, and the system applicability is improved; the air bag is pre-started before collision to shorten the response time, the safety belt is automatically unlocked after collision, passive protection is converted into active front intervention, the passenger injury degree and the secondary disaster risk are effectively reduced, and passenger injury is reduced.

The application discloses a kind of self-adapting risk protection method and system of smart home equipment control, and relates to the technical field of smart home security.The method comprises the following steps: receiving the device control request of control end and extracting context information;Determine the risk analysis module to be called through decision scheduling module;Each risk analysis module carries out risk scoring;Comprehensive scoring engine carries out weighted aggregation and risk amplification to each risk score, and gives comprehensive risk score;According to the comprehensive risk score, the safety policy is executed on the device control request, including release, secondary authentication or interception operation;Finally, the device control request released by terminal is executed.The application realizes multi-dimensional dynamic risk assessment, and improves the security of smart home equipment control.

A virtual augmentation system includes: one or more head-up displays displaying virtual images over an environment forward of a host vehicle; sensors determining parameters indicative of a state of the host vehicle; a telematics module obtaining roadway information indicative of roadway markings and objects along a predicted path of the host vehicle; and a virtual marker and object modification module, based on the parameters and the roadway information, i) detecting one or more roadway markings and objects, ii) generating modified versions of the one or more detected roadway markings and objects as one or more virtual roadway markings and objects, and iii) via the one or more head-up displays, overlaying the virtual one or more roadway markings and objects over at least a portion of the one or more detected roadway markings and objects in a field of view of an occupant of the host vehicle.

This application relates to the field of vehicle active safety technology, and in particular to a vehicle emergency scenario safety prompt system and method. The system includes: a sensor module for real-time acquisition of current vehicle sensor data; a data processing module for determining the current emergency scenario type based on preprocessed sensor data; a safety policy generation module for calling a target safety policy corresponding to the current emergency scenario type from a pre-set safety policy library; and a prompting module for converting the target safety policy into voice prompts and video demonstrations to guide the user in taking appropriate actions. This solves the problem that existing automotive safety systems cannot provide proactive and immediate operational guidance in emergency scenarios. It achieves automatic identification of emergency scenarios through multi-sensor fusion and simultaneous output of voice and video operation guidance, thereby helping users respond correctly in critical moments, avoid misoperation, and reduce the risk of secondary accidents.

The present invention discloses a control method, device, vehicle, and computer-readable storage medium for an electric vehicle. The method comprises: obtaining a request value and a monitoring value corresponding to each of at least one requested torque; determining whether the requested torque is erroneous based on the monitoring value and the request value and generating an error message; and controlling the electric vehicle based on a safety policy corresponding to the error message. This method not only enables real-time monitoring of each requested torque, but also enables execution of a corresponding safety policy when an error occurs in the requested torque, thereby improving the driving safety of the electric vehicle.

The invention provides a power distribution network reconstruction method and system based on multi-agent safety reinforcement learning. The method comprises the following steps: constructing a Markov decision process in a power distribution network reconstruction demand state; constructing a security constraint condition meeting the power distribution network through a security policy embedding mode; a dynamic graph convolutional network GCN is adopted to encode the state characteristics of the power distribution network; constructing a parameterized strategy function by adopting a deep neural network; constructing a multi-objective value function to evaluate the long-term return of taking the action at under the current state st; and multi-agent cooperative training and optimization are carried out, and after the objective function is converged, a safe power distribution network reconstruction topological structure is output by the agents. Through the method, the stability and efficiency of power distribution network reconstruction are improved.

Method for monitoring a target system including the steps of obtaining input data relating to at least one current state of the target system, obtaining at least one functioning constraint of the target system, calculating, by a predictive control, an envisaged sequence of controls for monitoring the target system, from the input data, at least one target state of the target system, and the constraint; calculating, by a function readable by electronic equipment, a reliability value of the envisaged sequence of monitoring controls, according to at least one safety criterion, from the envisaged sequence of monitoring controls, and the constraint; generating, from the reliability value and a safety policy, at least one safety control; generating a monitoring control from the envisaged sequence of monitoring controls and the safety control, and applying this monitoring control to the target system.

The application discloses a kind of multimodal driving safety monitoring systems based on hydrogel electrode, it is related to driving safety monitoring technical field.The application includes edge device and cloud, wherein edge device includes synchronous acquisition module, real-time monitoring module, safety policy module and channel detection module;Through high-precision data acquisition and synchronization, ensure that each physiological and environmental signal is accurately aligned;From multiple dimensions, capture the change of driver state, timely warning reduces the risk of accident, realize comprehensive real-time monitoring and multidimensional evaluation;Channel failure detection switches standby channel in time, ensure that signal acquisition is continuous and stable;Data management and incremental learning of cloud constantly improve the adaptability of system to new driving scene, prevent model recession;Realize;Continuous self-adaptation and model iteration optimization.In general, the system can not only monitor and warn driving fatigue state in real time, and continuously improve monitoring accuracy and system stability through the closed-loop feedback of edge and cloud, provide guarantee for driving safety.

The application provides a task execution method for an agent, an agent, a device, a medium and a product, wherein the task execution method for the agent comprises: in response to receiving an evidence task package, generating an original plan based on the evidence task package and a corresponding formal constraint specification, and compiling the original plan into a structured execution plan; performing static verification on the structured execution plan based on the formal constraint specification; after the static verification is passed, scheduling the execution of the structured execution plan; recording evidence information during the execution of the structured execution plan, and outputting a task execution result comprising the evidence information. The above scheme defines a decision boundary through the formal constraint specification, cooperates with the compiled structured execution plan and the pre-execution verification mechanism, converts the agent from an open free reasoning mode to a constrained verifiable execution mode, improves the flexibility of task execution, and ensures that the execution process conforms to the predetermined safety policy.

The invention discloses a boarding bridge full-automatic control device and a control method, and relates to the technical field of airport ground equipment automation. The device comprises a physical isolation special network system, a bimodal sensing system, a PLC control system, an encryption storage system, a man-machine interaction system and a security policy execution system. Data isolation is achieved through an independent private network, cabin door recognition and obstacle detection are achieved through the dual-mode sensing technology of a camera and a laser radar, full-automatic approaching and bridge withdrawing operation is achieved in combination with a PLC control system, and operation safety of the system is guaranteed through multiple encryption and a safety strategy. The method can improve the operation efficiency of the boarding bridge, reduces manual intervention, enhances the operation safety, and is suitable for automatic upgrading of various airport boarding bridges.

The application discloses a safety reinforcement learning method for continuous control problems. A probabilistic ensemble model is used to approximate the dynamic environment, representing the cognitive uncertainty and accidental uncertainty of the environment. A safety policy optimization method is designed with a dangerous punishment mechanism as the core, which increases the cost of executing unsafe actions so that the value of safe actions is higher than that of unsafe actions. The safety policy optimization method is combined with model-based reinforcement learning, and the convergence and safety of the algorithm are theoretically proved. A model predictive shielding algorithm is designed, which uses model predictive control as a potential shielding layer during the interaction between the agent and the environment, filters high-risk actions, and selects high-value actions. Simulation experiments show that the safety reinforcement learning method for continuous control problems proposed by the application can achieve good performance while maintaining safety.

The invention belongs to the technical field of nuclear power, and particularly relates to an elastic capacity expanding and shrinking system applied to the nuclear industry. According to the accurate elastic response provided by the invention, the system adopts the K-ELI model to fuse the business indexes, the security factors and the dynamic weights, the business sensitivity is high, and the decision is more in line with the actual demand. Accuracy, timeliness and effectiveness of capacity expansion and contraction decision are greatly improved, and nuclear industry load fluctuation is effectively dealt with. The security policy adapter performs real-time verification and constraint before decision making and execution, and security context is accurately injected. The problem of conflict between security policies and capacity expansion and shrinkage is solved, a system full link adapts to various types of CPUs, OS, monitoring interfaces of container engines, APIs and security mechanisms, and optimization is carried out according to platform characteristics. And process group-level elastic expansion and contraction are supported. The refined resource control requirement of a special module in a nuclear industry scene is met, and the resource utilization rate is increased.

The invention discloses a hydrogen energy micro-grid dispatching method based on CVaR constraint safety reinforcement learning, and belongs to the technical field of power system operation and control. The method comprises the following steps: firstly, acquiring heterogeneous multi-microgrid community environment state spatial data; constructing a dynamic equivalent carbon intensity tracking model, and checking node carbon intensity and energy storage dynamic equivalent carbon intensity; then, a multi-agent flexible Actor-Critic distributed security policy network is constructed, the carbon emission long tail risk is captured through a non-cross quantile network, and a CVaR constrained original-dual optimization training model is adopted; and finally, outputting a scheduling instruction, and realizing cooperative scheduling by means of a carbon-energy cooperative time shifting mechanism. The method has the core advantages that the energy storage carbon flow is accurately tracked, the tail risk is explicitly optimized, economic and carbon quota constraints are balanced, the extreme carbon standard exceeding risk can be effectively avoided in an uncertain environment, the violation rate is remarkably reduced, the renewable energy consumption rate is increased, and community low-carbon cooperation and safe operation are achieved.

The application discloses a cold-chain logistics data protection method and system based on data analysis, relates to the technical field of data analysis, and guarantees the integrity and timeliness of cold-chain transportation data through an original parameter set Par; a characteristic deviation set Dev is generated, which can reveal implicit data deviations in the cold-chain transportation process due to network fluctuations or sensor abnormalities; a data credibility score Qdc is further calculated, so that each batch of cold-chain transportation data can obtain a quantitative credibility result; the data credibility score Qdc is compared with a system threshold value, an abnormality identification set Jud is generated, so that high-credibility data and low-credibility data can be quickly distinguished, and abnormal information is prevented from being directly mixed into a business database; finally, the abnormality identification set Jud is matched with a safety policy set of the cold-chain logistics, targeted control instructions are generated, and it is ensured that abnormal data can trigger immediate protection measures when being identified.

System for controlling a motor vehicle (MV) for autonomous driving, wherein the system comprises: a bank of control units (110) comprising a number of vehicle control units (ACU1, ..., ACUn) configured to generate target control signals (NCS) configured to cause the motor vehicle (MV) to move autonomously in accordance with a target path; and a monitoring unit (160) configured to receive sensor signals (SS) from the motor vehicle (MV) describing a current state of the motor vehicle (MV) and to generate commands ({cmd}) based on the control signals (SS) that influence how the bank of control units (110) generates the target control signals (NCS), characterized by the fact that the system further comprises: a data storage (140) comprising a set of boundary conditions ({bc}) that the target path must satisfy to be considered safe; and a first data interface unit (163) configured to provide at least one safety policy (P) describing an associated task-related rule to be followed during the operation of the motor vehicle (MV), wherein the first data interface unit (163) is configured to provide at least one safety policy (P) based on a safety event (SC) that prescribes how the motor vehicle (MV) should be controlled in order to meet a functional safety standard, wherein the monitoring unit (160) is configured to repeatedly generate the commands ({cmd}) to update the boundary conditions ({bc}), aiming to keep the target path within limits specified by the sensor signals (SS) and at least one safety policy (P), and wherein the bank of control units (110) is configured to: to read the set of boundary conditions ({bc}) from the data storage (140), and to control the motor vehicle (MV) to move in such a way that the desired path satisfies the boundary conditions ({bc}).

The present application relates to a kind of intelligent lock monitoring method and system based on machine vision, including the following steps, the method is by capturing the continuous video frame of intelligent lock area and extracting three-dimensional key point coordinates, the posture of user is tracked modeling, obtains action trajectory curve, identifies target interactive behavior event of approaching operation interface. Subsequently, the event is key frame extraction, obtains the action feature subsequence of candidate identity authentication stage, and whether there is multiple repeated attempt to trigger door lock behavior mode is detected. Finally, whether the access request is in accordance with the safety policy based on the behavior mode is judged, generates door lock control instruction and sends to controller module and executes corresponding operation. The technical problem that most current intelligent lock systems exist greater security risks when facing malicious probe, video playback attack and other advanced attack means is solved.