This technology concerns explaining the behavioral intent of AI agents to users in systems applying reinforcement learning and reflecting those results in system operation. Specifically, it is a technology that enhances the explainability of reinforcement learning-based decision-making by outputting the AI's intent before it acts, through an XAII provision module.
Existing reinforcement learning systems made it difficult for users to understand the purpose or intent of an agent's chosen actions in advance, which could limit reliability, controllability, and field applicability. To address this, this technology proposes a method that places an XAII provision module and an AI mode provision module in the control unit of an AI terminal, providing behavioral intent information to the input/output unit, and reflecting user feedback or result values.
Accordingly, this technology enables pre-explanation of AI agent actions, user confirmation, and adjustment of intervention levels, making it applicable to safety-critical automation systems, robots, defense and operational support systems, and explainable AI services.
This technology relates to a system and method for analyzing application installation information. Specifically, it is an application installation information extraction technique that analyzes Android installation logs and DB files.
Existing technologies faced limitations in malware detection within current systems, often addressed by analyzing system call logs and proposing methods to detect malicious activity in Openstack and HDFS using natural language sequence models and LSTM (Long Short-Term Memory). To overcome this, the present technology proposes a configuration that includes a log collection unit, a log analysis unit, and a database generation unit for collecting installation logs, analyzing configuration information, and recording application configuration information.
Accordingly, this technology can improve the ability to provide application configuration information by analyzing log extension files and DB extension files related to configuration data. It holds value for application in the software, IT, and internet sectors.
This technology relates to a harmonic compensation apparatus and method for a single-phase grid-connected inverter. Specifically, it is a single-phase grid-connected inverter harmonic compensation technology combining PI·PR control.
Conventional technologies needed to address the complexity of designing controllers to eliminate harmonics in a stationary reference frame and improve the power factor in grid-connected power systems. Accordingly, this technology proposes a configuration that includes a filtering unit, a current controller, harmonic compensation, and a pulse width modulation unit for a single-phase system interconnected inverter.
As a result, this technology can improve the output quality of a single-phase system interconnected inverter by reducing the Total Harmonic Distortion (THD) value. It has practical value in the fields of electricity, power generation, and energy.
This technology relates to an artificial neural network rule extraction apparatus and method. Specifically, it is an explainable AI technology that extracts the decision basis of artificial neural networks into logical rules.
Existing technologies faced limitations with traditional methods and required a more efficient and accurate approach, especially in extracting rules from artificial neural networks for data with continuous attributes. To address this, the present technology proposes a configuration that includes an artificial neural network rule extractor for learning input datasets, comprising a data learning unit, a first rule extraction unit, a binary classification unit, and a second rule extraction unit.
Accordingly, this technology can improve the accuracy and efficiency of extracting rules from artificial neural networks for data with continuous attributes, overcome the limitations of traditional methods, and provide a more efficient and accurate approach. It has value for utilization in the software, IT, and internet fields.
This technology relates to an event-operation-based multi-action intention inference device and method. Specifically, it is a multi-action intention inference technology that utilizes time-series perceptual information and conflict rules.
Existing technologies sought to address the difficulty of models predicting multiple intentions simultaneously and needed to resolve the lack of information at specific points in time for single intention prediction. To address this, this technology proposes a configuration that includes an input unit, an intention inference module, a heartbeat determination module, and verification of the error module.
Accordingly, this technology can improve the accuracy of action intention prediction and reduce the computational load required for decision-making. It has value for utilization in the healthcare, pharmaceutical, and software sectors.
This technology relates to a naval combat system device and its implementation method for controlling the combat operations of a naval vessel based on artificial intelligence. Specifically, it is a technology that processes reinforcement learning-based tactical and combat execution phases by linking sensors, consoles, platform equipment, and weapon systems with a system cabinet.
Existing naval combat systems faced challenges in providing quick and consistent judgments in complex combat situations when sensor information, platform control, weapon control, and tactical recommendations were operated separately. To address this, this technology proposes a method that integrates simulation data-based learning and combat execution algorithms by configuring sensor interface modules, multi-functional console modules, platform interface modules, weapon interface modules, and a system cabinet.
Accordingly, this technology can link AI-based tactical recommendations with equipment control during detection, control, and engagement processes, which is expected to have ripple effects in naval vessel operation automation, combat system advancement, and virtual training-linked defense systems.
This technology concerns a device and method for preventing the freezing of washing machine drain hoses. Specifically, it is a device technology that prevents washing machine drain hose freezing through ultrasonic vibration.
Existing technologies needed to address the frequent failure of washing machine drainage due to residual freezing in cold environments. This technology proposes a configuration that includes a drain body, a vibrator, an excitation unit, and a controller to prevent freezing of the drain pipe.
Accordingly, this technology can improve the efficiency of preventing drain pipe freezing and residual water in washing machines by using ultrasonic vibration. It has practical value in the fields of TVs, home appliances, and electronic devices.
This technology relates to an indoor positioning system and an indoor positioning method using the same. Specifically, it is an indoor positioning technology utilizing Wi-Fi and device sensor fusion.
Existing technologies, particularly conventional indoor positioning systems, needed to address precision issues in environments with obstacles such as Wi-Fi interference, walls, and human bodies. To address this, this technology proposes a configuration that includes a user terminal with a sensor, a wireless communication receiver, an AP selection unit, a distance measurement unit, a threshold comparison unit, and a position compensation unit.
Accordingly, this technology can improve the precision of indoor positioning systems by utilizing sensor data and Wi-Fi RSS values to estimate the user's location. It is valuable for applications in the IT, internet, software, and smartphone sectors.
This technology relates to a frame for mounting multiple laser radars, a laser radar system including the frame, and a method for integrating sensing data coordinates from the multiple laser radars. In particular, it is an automotive laser radar system technology that integrates coordinates from multiple low-cost LiDARs.
Existing technologies faced the challenge of high manufacturing costs and low precision associated with expensive multi-channel laser radars. To address this, this technology proposes a configuration comprising multiple laser radars positioned on the upper exterior of a vehicle, a frame unit for supporting the laser radars, and a controller for integrating sensed data and virtualized coordinates.
Consequently, this technology can improve precision and reduce the manufacturing cost of laser radar systems by implementing multi-channel laser radars using multiple low-cost laser radars. It offers significant value for applications in the automotive, mobility, electronic device, and electronic component sectors.
This technology relates to a monitoring system that monitors services deployed using a service mesh architecture. Specifically, it is a Service Mesh-based Microservice QoS Monitoring and Recovery Technology.
Existing technologies needed to address the complexity and challenges of service monitoring in microservice architectures by providing a service monitoring system that ensures service quality and efficient recovery procedures. Therefore, this technology proposes a configuration that includes an execution unit, an analysis section, a decision unit, and service discovery for monitoring services and ensuring service quality within a service mesh architecture.
Accordingly, this technology can improve monitoring and recovery processes in microservice architectures by providing a service monitoring system that ensures service quality and efficient recovery procedures. It offers valuable applications in the IT, internet, and software sectors.
