Electronic device for artificial intelligence ethical decision-making model and operating method thereof

Abstract

An electronic device for an artificial intelligence ethical decision-making model may be provided. The electronic device comprises a storage unit and a processor. The processor may detect whether a pedestrian is a subject of consideration when a stop behavior is determined, extend the stop time if the pedestrian is a subject of consideration, and perform an additional safety check process when every pedestrian has crossed a road.

Description

Electronic device and its operating method for an ethical decision-making model of artificial intelligence

[0001] The present invention relates to an electronic device for artificial intelligence and its operating method. More specifically, the present invention relates to an electronic device for an ethical decision-making model for artificial intelligence and its operating method.

[0002]

[0003] With the advancement of artificial intelligence technology, concerns are growing about the ethical issues it may pose, such as unfairness, bias, the spread of misinformation, and misuse. Consequently, interest is growing in technologies that enable ethical decision-making by AI. In particular, in the case of autonomous driving technology, AI decisions can potentially impact human life, necessitating the need for autonomous driving systems capable of safe and ethical decision-making.

[0004] For example, Republic of Korea Patent Publication No. 10-2248705 (registered on April 29, 2021) discloses a method for evaluating AI ethics. However, this patent publication does not disclose a method for AI ethical decision-making.

[0005]

[0006] The challenge the present invention seeks to address is to integrate ethical standards into the decision-making process of artificial intelligence, thereby enabling it to operate in a fair and transparent manner.

[0007] The challenge the present invention seeks to address is to minimize bias in artificial intelligence and ensure that services can be provided fairly to individuals with diverse demographic backgrounds.

[0008] The challenge the present invention seeks to address is to equip artificial intelligence with the ability to understand the context of a situation and respond appropriately.

[0009] The challenge the present invention seeks to address is to transparently disclose the decision-making process of artificial intelligence to users, thereby building trust and reducing misinformation and misunderstandings.

[0010]

[0011] According to embodiments of the present invention, an electronic device for an ethical decision-making model of artificial intelligence includes a storage unit and a processor, wherein the processor can detect whether a pedestrian is a person subject to consideration when a stop action is decided, extend the stop time if the pedestrian is a person subject to consideration, and perform an additional safety verification process when all pedestrians have crossed the road.

[0012] According to embodiments of the present invention, an operating method of an electronic device for an ethical decision-making model of artificial intelligence may include a step of detecting whether a pedestrian is a person to be considered when a stop action is decided, a step of extending the stop time if the pedestrian is a person to be considered, and a step of performing an additional safety verification process when all pedestrians have crossed the road.

[0013] According to embodiments of the present invention, a non-transitory computer-readable medium storing computer instructions that, when executed by a processor of an electronic device, cause the electronic device to perform operations, the operations may include: when a stop action is determined, detecting whether a pedestrian is a person to be considered; if the pedestrian is a person to be considered, extending the stop time; and if all pedestrians have crossed the road, performing an additional safety check process.

[0014]

[0015] According to embodiments of the present invention, transparency can be enhanced by clearly explaining the decision-making process of an AI system to users. This can help users understand and trust the AI's decisions and reduce potential misunderstandings.

[0016] According to embodiments of the present invention, the fairness of generative AI systems can be enhanced through bias detection and correction algorithms. This provides all users with an equal level of service and decision-making, while minimizing unreasonable bias against specific groups.

[0017] According to embodiments of the present invention, the ethical responsibility of technology can be enhanced by supporting AI systems to act in accordance with ethical standards. This can be crucial for AI to make more responsible decisions by considering social and cultural contexts.

[0018]

[0019] FIG. 1 is a block diagram of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0020] Figure 2 is a conceptual diagram of an ethical decision-making framework according to one embodiment of the present invention.

[0021] FIG. 3 is a conceptual diagram illustrating the first step of an ethical reinforcement learning cycle according to one embodiment of the present invention.

[0022] FIG. 4 is a conceptual diagram illustrating the second step of an ethical reinforcement learning cycle according to one embodiment of the present invention.

[0023] FIG. 5 is a conceptual diagram illustrating the third step of the ethical reinforcement learning cycle according to one embodiment of the present invention.

[0024] FIG. 6 is a conceptual diagram illustrating the fourth step of the ethical reinforcement learning cycle according to one embodiment of the present invention.

[0025] Figure 7 is a conceptual diagram for explaining a data processing flow according to one embodiment of the present invention.

[0026] Figure 8 is a process for ethical decision-making according to one embodiment of the present invention.

[0027] Figure 9 is a bias detection algorithm according to one embodiment of the present invention.

[0028] Figure 10 is an additional safety verification process algorithm in a bias detection algorithm according to one embodiment of the present invention.

[0029] Figure 11 is a scene analysis process algorithm in a bias detection algorithm according to one embodiment of the present invention.

[0030] Figure 12 is a continuous monitoring process algorithm in a bias detection algorithm according to one embodiment of the present invention.

[0031] FIG. 13 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0032] FIG. 14 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0033] FIG. 15 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0034] FIG. 16 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0035] Figure 17 is a flowchart of a scene analysis process of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0036] FIG. 18 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0037] FIG. 19 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0038] FIG. 20 is a flowchart of a process for continuously monitoring the surroundings of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0039] FIG. 21 is a block diagram of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0040]

[0041] Hereinafter, the operating principles of preferred embodiments of the present invention will be described in detail with reference to the attached drawings. Furthermore, when describing embodiments of the invention, detailed descriptions of related known functions or configurations will be omitted if they are deemed to obscure the gist of the present disclosure. Furthermore, the terms used below are defined based on their functions in the present invention and may vary depending on the intent or custom of the user or operator. Therefore, the definitions of the terms used should be interpreted based on the contents and corresponding functions throughout this specification.

[0042] FIG. 1 is a block diagram of an electronic device (100) for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0043] Referring to FIG. 1, an electronic device (100) for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a storage unit (110) and a processor (120). In one embodiment, the electronic device (100) may be mounted on an autonomous vehicle or an edge computing device for an autonomous vehicle.

[0044] The storage unit (110) can store various data and programs. The processor (120) can control the overall operation of the electronic device (100). For example, the processor (120) can control the storage unit (110).

[0045] In one embodiment, the processor (120) may, when deciding to stop, detect whether the pedestrian is a person of interest. If the pedestrian is a person of interest, the stop time may be extended, and if all pedestrians have crossed the street, additional safety checks may be performed. Persons of interest may include, for example, children, patients, the elderly, pregnant women, etc.

[0046] In one embodiment, the processor (120) may perform object recognition of items carried by a pedestrian to determine whether the pedestrian is a person of interest. For example, the processor (120) may recognize items such as a daycare bag, a cane, a stroller, or a wheelchair. By recognizing objects, the person of interest can be easily identified.

[0047] In one embodiment, in an additional safety check process, the processor (120) may extend the stopping time if the pedestrian exhibits unsteady movements, inform the pedestrian that it is safe, and sound an alarm to the pedestrian if the pedestrian is looking away.

[0048] In one embodiment, the processor (120) may, if it determines an emergency stop action, activate the emergency brakes, alert surrounding vehicles, contact emergency services, perform scene analysis, and wait until safety is confirmed.

[0049] In one embodiment, in scene analysis, the processor (120) can analyze hazards in the scene, predict the movement paths of pedestrians, assess collision risks, and determine an optimal avoidance route.

[0050] In one embodiment, the processor (120) may, if it determines a deceleration action, reduce the speed of the vehicle, increase the stopping distance, activate a pedestrian alarm system, and yield to pedestrians.

[0051] In one embodiment, the processor (120) may perform a process of maintaining the speed of the vehicle and continuously monitoring the surroundings when normal operation is determined.

[0052] In one embodiment, in the process of continuously monitoring the surroundings, the processor (120) can monitor whether there is a pedestrian who suddenly appears, and determine and perform an action based on the status of the pedestrian and the crosswalk who suddenly appears.

[0053] Figure 2 is a conceptual diagram of an ethical decision-making framework according to one embodiment of the present invention.

[0054] Referring to Figure 2, the ethical decision-making framework, as an integrated part of the ethical middleware, can function independently of core processes. The ethical decision-making framework is crucial for coordinating methods to mitigate bias and prevent errors. Starting from the initial "sensing" stage, bias detection algorithms can proactively identify and eliminate biases, thereby preventing the spread of misinformation.

[0055] Figure 2 illustrates a three-stage hierarchical structure for applying ethical controls in the process of transmitting data collected from sensor units, such as Internet of Things devices, to a generative AI platform.

[0056] The first stage may include a data input layer and a data processing layer, which correspond to the "sensing" stage. In this layer, ethical middleware can collect initial data, identify bias in the input data using bias detection algorithms and self-inspection algorithms, and perform self-verification. The data processing layer can improve the quality of the collected information through data filtering, imputation, and transformation, and can take measures to protect personal information, ensuring that generative AI operates on ethically sound data.

[0057] The second stage is the Analysis & Processing Layer, categorized as the "Thinking" stage. Data that has passed the sensing stage can undergo additional bias-detection analysis and self-diagnosis evaluation in the Analysis & Processing Layer. This process can be conducted with reference to Korean cultural, social, and ethical standards. Through this analysis, the system can establish guidelines for appropriately interpreting and processing data, supporting the generation of AI to produce ethically valid results.

[0058] The third stage is the Response & Adjustment Layer, which falls within the "Action" phase. This layer can correct biases and implement behavioral adjustments. Generative AI can make appropriate decisions based on previous analysis results through an ethical decision-making mechanism. This allows generative AI systems to recognize domain-specific ethical boundaries, operate efficiently within these boundaries, generate ethically valid content for user interactions, and proactively manage potential ethical risks.

[0059] To meet the ethical judgment requirements of generative AI, bias detection algorithms can incorporate advanced technologies such as contextual awareness and situational awareness. This can ensure careful bias management during data processing and enhance the system's ability to make ethical decisions. A contextual awareness module can play a fundamental role by analyzing the causes and conditions under which data was generated, thereby aligning data interpretation with environmental and contextual facts. These meticulously analyzed intuitions can be fed into bias detection algorithms and form the basis for the central analysis process. Using middleware technology to document the analysis flow and results as immutable records can enhance the reliability and openness of the system's judgmental actions. These technological and structural advancements not only increase the completeness and precision of generative AI systems, but also establish proactive systems for improving current biases and solidify the core principles of fairness and accuracy, which are crucial for ethical AI operation.

[0060] Within middleware technologies, data processing for self-censoring algorithms can determine from the outset whether the actions and decisions of generative AI systems are ethical, privacy-preserving, transparent, and explainable. Self-censoring modules can be built by referencing guidelines based on the social context surrounding the state and behavior of generative AI systems, and can assess how their decisions and actions may be interpreted within specific cultural and social contexts. Furthermore, generative AI systems can perceive changes in their environment and, through contextual awareness, elicit appropriate responses. To ensure transparency, data processing processes can be recorded and managed, and to ensure explainability, errors can be identified and corrected. Algorithms can therefore regularly monitor data processing and identify potential ethical violations. Thus, middleware can balance technical utility and ethical responsibility by strengthening the autonomy and accountability of generative AI systems and continuously improving cultural and contextual appropriateness.

[0061] As illustrated in Figure 2, ethics can be considered across sensing, thinking, and behavior in the decision-making process of a generative AI module to ensure safe decision-making. Ethical considerations can ensure fairness by utilizing criteria that can detect biases based on culture, gender, age, and other factors, while also incorporating legal and social standards. These ethical guidelines can be designed in detail to be applied and judged in each processing step.

[0062] In the sensing phase, input from sensor data and multimedia data, such as images, videos, and language, can be detected, and potential biases can be identified by analyzing patterns and context within the data. Contextual awareness involves understanding the cultural and contextual nuances of the data being analyzed.

[0063] In the inference stage, algorithms process detected data contextually, determine the cause, distinguish between true bias and contextual content, and securely record the inference process using middleware technology to ensure transparency and reliability. This allows for the use of complex algorithms that treat all data fairly and clearly establish accountability.

[0064] At the action stage, algorithms for ethical decision-making based on analysis can ensure the consistency of actions and changes by identifying biased data and suggesting corrections, or by adjusting AI models to mitigate bias. This can minimize negative impacts by assessing the social repercussions of decisions, reviewing the results, and adjusting guidelines or models to balance values.

[0065] An ethical decision framework can provide a structured approach to assisting generative AI systems in resolving complex ethical issues and making appropriate decisions. This framework can involve four steps: defining the goals and scope of the generative AI system; establishing the social, cultural, and legal standards the system must adhere to; establishing decision-making rules; and evaluating and refining the results. These decision-making rules can be based on ethical standards that reflect relevant cultural characteristics. In the result evaluation and improvement step, the system's decisions are evaluated, and, if necessary, the system's ethical guidelines or decision rules can be continuously refined. The decision-making framework can be integrated across all layers of the ethical middleware, ensuring that all decisions adhere to established ethical standards. As a result, generative AI systems can recognize ethical risks and develop strategies to address ethical considerations while operating in diverse cultural and social contexts.

[0066] Bias detection algorithms can be designed to precisely detect and manage biases that may arise during data processing within generative AI systems. Bias detection algorithms can utilize context-aware capabilities to analyze the causes and context of data generation, and situational awareness to assess the significance and relevance of data based on current environmental conditions. These intuitions can be integrated into the core analysis process of bias detection algorithms, enabling them to accurately identify and assess potential biases within the data. By utilizing blockchain technology for distributed processing, bias detection algorithms can store the analysis process and results as immutable records, thereby ensuring the reliability and transparency of algorithmic decisions and actions. This record-keeping method allows the system to continuously monitor and improve biases. If bias is detected at the data input layer, bias detection analysis can be performed at the analysis and processing layer, followed by bias correction at the response and coordination layer.

[0067] Self-censorship algorithms can have autonomous structural features that enable generative AI systems to independently evaluate and coordinate their decisions and actions. The initial self-censorship module can analyze the current state and behavior of the generative AI and assess whether the system's decisions and actions can be interpreted across various cultural and social contexts. Furthermore, context-aware capabilities can be used to determine whether the system is responding appropriately to changes in its surroundings. Self-censorship results can be securely recorded using blockchain technology, which provides integrity and traceability of the system's actions. This structural framework can strengthen the autonomy and explainability of generative AI systems and serve as a foundation for continuously improving cultural and contextual alignment. The self-censorship process can begin at the data input layer, continue with evaluation in the analysis and processing layer, and culminate in behavioral coordination using reinforcement learning in the response and coordination layer.

[0068] FIG. 3 is a conceptual diagram illustrating the first stage of an ethical reinforcement learning cycle according to one embodiment of the present invention. FIG. 4 is a conceptual diagram illustrating the second stage of an ethical reinforcement learning cycle according to one embodiment of the present invention. FIG. 5 is a conceptual diagram illustrating the third stage of an ethical reinforcement learning cycle according to one embodiment of the present invention. FIG. 6 is a conceptual diagram illustrating the fourth stage of an ethical reinforcement learning cycle according to one embodiment of the present invention.

[0069] Referring to Figures 3 through 6, utilizing reinforcement learning techniques in an ethical decision framework can significantly enhance the process by which generative AI systems enhance their ethical decision-making capabilities. This can help generative AI systems dynamically make appropriate ethical decisions in complex, real-world environments. By applying these techniques, generative AI systems can dynamically adjust their decision strategies based on feedback from real-world interactions, thereby improving their behavior to more closely adhere to ethical standards. The ethical reinforcement learning cycle can include four stages, each of which can be linked to a different layer of ethical middleware: Stage 1: Establishing ethical benchmarks and bias detection; Stage 2: Scenario-based learning and self-censorship; Stage 3: Integrating human-centered assessment and feedback; and Stage 4: Ethical coordination and reinforcement.

[0070] Referring to Figure 3, a bias detection algorithm can be activated at the data input layer, where it can play a crucial role in initial data processing. In the first stage, the source and context of data creation can be analyzed based on standards established through ethical benchmarks. By utilizing context-aware capabilities, the bias detection algorithm can review the input data for compliance with these established ethical standards, forming the basis for the initial bias detection process. This initial analysis can be crucial for identifying and filtering potential biases before the data is processed in subsequent stages.

[0071] The results of assessing the ethical context of data can directly influence the next step in scenario-based learning. Data identified as biased can undergo further analysis and modification, transforming them into scenarios that enable generative AI to learn realistic and ethical responses. This process can enable bias detection algorithms to effectively assess the inherent value and meaning of data, thereby supporting the generative AI learning process within ethical middleware. Once scenarios are established, they can be updated through self-censoring algorithms, ensuring that each scenario adheres to current ethical standards and accurately reflects evolving cultural and social contexts. This continuous update cycle, driven by self-censoring algorithms, can enhance the responsiveness and adaptability of generative AI systems to new ethical issues and dilemmas.

[0072] Referring to Figure 4, during the scenario-based learning phase, a self-censorship algorithm is activated in the analysis and processing layer to evaluate the ethicality of responses generated by the generative AI. The self-censorship algorithm can perform self-evaluations to determine how well the generative AI's decisions and actions align with established ethical standards, and can serve as a crucial self-evaluation mechanism in the generative AI's learning process. By analyzing the appropriateness of the generative AI's responses to each scenario in relation to the cultural and social context, the generative AI can develop improved and more sophisticated ethical judgment capabilities.

[0073] The results of this self-censorship can be integrated into the human-centered evaluation phase, providing a solid foundation for human evaluators to conduct thorough examinations of generative AI responses. By leveraging the detailed analysis results of the self-censorship algorithm, evaluators can more accurately assess the ethical appropriateness of generative AI responses. This rigorous evaluation process can be crucial to ensuring that the data and responses generated during the generative AI training process accurately reflect human ethical values ​​and standards.

[0074] Referring to Figure 5, incorporating human-centered evaluation and feedback within an ethical decision-making framework can be crucial. Such evaluations can occur at the response and coordination layers, and the results can directly influence the real-time ethical coordination of generative AI and improve decision-making algorithms. Human evaluators can assess whether generative AI's responses meet established ethical standards and provide specific feedback on perceived shortcomings. This feedback can be crucial for the continuous improvement of generative AI's ethical learning process.

[0075] When feedback is received, generative AI can adjust its responses based on the input and optimize ethical behavior through reinforcement learning. In this iterative process, generative AI can apply the knowledge gained from the feedback to refine future responses to more closely align with the ethical decision framework. This step can be crucial for allowing generative AI to continuously modify and improve its behavior to meet ethical standards.

[0076] Referring to Figure 6, finally, the generative AI can synthesize all feedback and learning outcomes to fine-tune the ethical decision-making framework. This process can be repeated iteratively across the entire ethical middleware system, leveraging the intuition and data accumulated from each step to continuously improve the generative AI's decision-making capabilities. To ensure transparency and reliability throughout the entire decision-making process, each decision made by the generative AI can be recorded on the blockchain. By learning and adapting through these iterative feedback cycles, the generative AI can evolve into a more sophisticated and trustworthy ethical decision-maker.

[0077] Figure 7 is a conceptual diagram for explaining a data processing flow according to one embodiment of the present invention.

[0078] Referring to Figure 7, the middleware may include the following configuration.

[0079] 1) Data collection module:

[0080] Collect data from various sensors (e.g. cameras, radar, LiDAR, etc.)

[0081] Combined to provide flexible processing capabilities without being limited by sensor type or vehicle type.

[0082] 2) Data preprocessing:

[0083] The collected data goes through a preprocessing process to remove noise, unify the format, and extract only the necessary information.

[0084] In this process, data analysis functions are implemented to enable real-time processing.

[0085] 3) Ethical Analysis:

[0086] Preprocessed data is analyzed using a bias detection algorithm to detect potential bias toward pedestrians.

[0087] Analysis results are evaluated based on ethical guidelines, and certain sensitive information is protected by encryption and anonymization technologies.

[0088] 4) Decision Support:

[0089] The decision-making module determines action guidelines based on the results of the ethical analysis.

[0090] This module prioritizes the computational power needed for critical ethical decision-making processes, making efficient use of the vehicle's computational resources.

[0091] 5) Behavioral Control and Execution:

[0092] The determined instructions are transmitted to the vehicle control system through the action control module so that the vehicle can respond appropriately.

[0093] Resource management is optimized to perform the necessary calculations to minimize the load on vehicle resources.

[0094] Figure 8 is a process for ethical decision-making according to one embodiment of the present invention.

[0095] Referring to Figure 8, this system can be broadly divided into three main stages: sensing, thinking, and action. The sensing stage detects objects, including pedestrians, by multiprocessing data collected through each sensor device using a peripheral scheduler. In the thinking stage, various tasks can be processed in parallel using a parallel processing scheduler to make ethical decisions. At this point, a bias detection algorithm can be incorporated to enable general decision-making. Ethical commands can then be executed, and vehicle operation and complexity can be implemented based on AI decisions.

[0096] 1) Multimodal sensor input (sensing): Data is collected from various sensors, and the collected data is preprocessed through signal processing.

[0097] 2) Artificial Intelligence Decision-Making Module (Thinking): This module determines the decisions to be reflected in autonomous driving and provides fair and quick support for ethical judgment.

[0098] [Data Analysis]

[0099] Contextual Analysis: Pedestrian Identification

[0100] Pedestrian Detection: Detecting the location and status of pedestrians

[0101] Risk Assessment: Assess the level of risk posed by the detected pedestrian.

[0102] [Bias Detection Algorithm] Analyzes collected data to detect possible bias in the AI ​​decision processor.

[0103] [Decision-making] Final decision based on detected data and bias analysis

[0104] [Ethical Control] Review whether the decision-making process meets ethical standards.

[0105] 3) Actuator control (action): Directly executes ethical commands received from the artificial intelligence decision-making module.

[0106] [Vehicle Control] Performing AI decisions through the operation of actual vehicles.

[0107] [Complexity and Accessibility] Assess and implement system complexity and technical accessibility.

[0108] These systems can be designed to accurately recognize pedestrians in dangerous situations and take precise action, utilizing diverse input data, sophisticated processing capabilities, and complex algorithms for ethical decision-making. These systems can play a crucial role in ensuring the safety of autonomous vehicles.

[0109] Figure 9 is a bias detection algorithm according to one embodiment of the present invention.

[0110] Referring to Figure 9, a bias detection algorithm can recognize and analyze bias in data or artificial intelligence systems. Such algorithms can play a crucial role in detecting unfair or biased patterns that may appear in machine learning models or decision-making systems. Bias can occur unconsciously or systematically during data collection, processing, and model training, ultimately negatively impacting the accuracy and fairness of results. For example, this disclosure describes a system configuration and functions designed to enable ethical decision-making and minimize bias in autonomous driving.

[0111] Bias detection algorithms can be part of the decision-making system of autonomous vehicles, integrated seamlessly from sensor data collection to decision-making. They can be crucial for recognizing and correcting bias within the conscious process of sensor data analysis, while ensuring that decisions adhere to ethical standards.

[0112] The operation of a bias detection algorithm can begin with the collection of multidimensional data from various sensors, such as cameras, LiDAR, and radar. This data can be normalized and classified based on predetermined criteria, such as pedestrian age, mobility aids, and crowd dynamics. This initial processing can prepare the data for further analysis and bias detection.

[0113] After data preparation, bias detection algorithms can actively scan the labeled data to identify patterns indicative of bias, such as the over- or under-representation of certain pedestrian groups. These biases could stem from biased training data or sensor inconsistencies. Once potential biases are detected, the algorithms can perform risk assessments and contextual analysis to understand how these biases could impact decision-making. This analysis can assess risks associated with biased decisions, such as unsafe pedestrian crossings or inefficient traffic flow.

[0114] Based on thorough analysis, bias detection algorithms can adjust their decision-making processes to mitigate perceived biases. For example, if the algorithm detects a bias toward slow pedestrians, it can adjust the vehicle's speed or stopping pattern to safely accommodate all pedestrians.

[0115] The final step could involve ethical controls, where decisions adjusted to account for biases are checked against the vehicle's ethical guidelines to ensure they are not only technically sound but also ethically justifiable. This ensures that decisions are fair and protect the safety of all road users.

[0116] Furthermore, bias detection algorithms can be dynamic systems that continuously learn from new data, adjusting parameters and improving accuracy and fairness over time. This continuous monitoring and learning can be crucial for adapting to unpredictable urban environments and changing pedestrian behavior.

[0117] This structured, ongoing, and ethically guided approach can ensure that autonomous vehicle decision-making processes are robust, fair, and adaptable, thereby enhancing the reliability and social acceptance of autonomous driving technology. Bias detection algorithms can therefore bridge the gap between technical efficiency and ethical responsibility in autonomous vehicle operation.

[0118] In this scenario, an autonomous vehicle approaches a busy urban crosswalk with pedestrians of various ages, abilities, and mobility aids. The vehicle's bias detection algorithm detects bias in sensor data that interferes with the recognition of wheelchair-bound pedestrians. The bias detection algorithm quickly adjusts sensor sensitivity to accurately recognize all pedestrians, regardless of their characteristics, and yields the right of way. This allows the vehicle to make an ethically sound decision to safely stop and allow all pedestrians to cross.

[0119] The algorithm illustrated in Figure 9 can describe the initial response phase. Based on real-time monitoring, the autonomous vehicle can detect pedestrians, execute a determined action, such as stopping or slowing down, and perform additional safety checks prioritizing pedestrian safety. The ExecuteAction function is the central function in this process and can execute the action determined by the vehicle's decision-making system. Based on the evaluation, the vehicle can choose to stop, initiate an emergency stop, slow down, or maintain its current speed. If a stop decision is made, the vehicle can come to a halt. If a vulnerable pedestrian is present, the stopping time can be extended to allow all pedestrians to safely cross the road. Through these measures, pedestrian safety can be prioritized and the vehicle's intention to protect other road users can be demonstrated.

[0120] Figure 10 is an additional safety verification process algorithm in a bias detection algorithm according to one embodiment of the present invention. Figure 11 is an algorithm for a scene analysis process in a bias detection algorithm according to one embodiment of the present invention.

[0121] Referring to Figures 10 and 11, safety checks and analysis can focus on thoroughly examining pedestrian movements and adapting vehicle responses to maintain a safe distance and effectively handle emergency situations. The PerformAdditionalSafetyChecks function can play a crucial role by extending stopping time for unstable pedestrians. This function can notify pedestrians when it is safe to cross and activate an audible warning for pedestrians who are distracted, thus improving overall safety.

[0122] The PerformExtensiveSceneAnalysis function can play a crucial role in assessing risk within a scene by predicting pedestrian paths, assessing the risk of collision with suddenly appearing pedestrians, and determining the optimal path to avoid accidents. This capability can be crucial for autonomous vehicles to safely navigate complex urban environments while prioritizing pedestrian safety.

[0123] Figure 12 is a continuous monitoring process algorithm in a bias detection algorithm according to one embodiment of the present invention.

[0124] Referring to Figure 12, the monitoring mechanism can ensure road safety by explaining how an autonomous vehicle continuously monitors its surroundings and makes real-time decisions about newly detected or suddenly appearing pedestrians while moving, and adapts to changes, and can respond quickly to unexpected situations.

[0125] The ContinuouslyMonitorSurroundings function can be crucial for vehicle adaptability and safety in dynamic urban environments. This function continuously scans the surroundings while the vehicle is moving, compares new and existing pedestrians, identifies pedestrians who appear unexpectedly, and determines immediate actions to ensure the safety of the vehicle and all road users.

[0126] FIG. 13 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0127] Referring to FIG. 13, an operating method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a step of detecting whether a pedestrian is a person to be considered when a stop action is determined (S1310), a step of extending the stop time if the pedestrian is a person to be considered (S1320), and a step of performing an additional safety verification process when all pedestrians have crossed the road (S1330).

[0128] FIG. 14 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0129] Referring to FIG. 14, the step of detecting whether a pedestrian is a person to be considered (S1310, see FIG. 13) may include a step of performing object recognition of an item carried by the pedestrian (S1410).

[0130] That is, the operating method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a step of performing object recognition of an item carried by a pedestrian (S1410), a step of extending a stop time if the pedestrian is a subject of consideration (S1420), and a step of performing an additional safety confirmation process if all pedestrians have crossed the road (S1430).

[0131] FIG. 15 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0132] Referring to FIG. 15, an additional safety check process (S1330, see FIG. 13) may include a step of extending the stopping time and informing the pedestrian that it is safe if the pedestrian shows unstable movements (S1531), and a step of sounding an alarm to the pedestrian if the pedestrian is looking away (S1532).

[0133] That is, the operating method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a step (S1510) of detecting whether a pedestrian is a person to be considered when a stop action is decided, a step (S1520) of extending the stop time if the pedestrian is a person to be considered, a step (S1531) of extending the stop time and informing the pedestrian that it is safe if the pedestrian shows unstable movements, and a step (S1532) of sounding an alarm to the pedestrian if the pedestrian is looking away.

[0134] FIG. 16 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0135] Referring to FIG. 16, the method of operating an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may further include a step (S1640) of activating an emergency brake, alerting surrounding vehicles, contacting emergency services, performing scene analysis, and waiting until safety is confirmed when an emergency stop action is determined.

[0136] That is, the operating method of the electronic device for the ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a step of detecting whether a pedestrian is a person subject to consideration when a stop action is determined (S1610), a step of extending the stop time if the pedestrian is a person subject to consideration (S1720), a step of performing an additional safety confirmation process when all pedestrians have crossed the road (S1630), and a step of activating the emergency brake, warning surrounding vehicles, contacting emergency services, performing scene analysis, and waiting until safety is confirmed (S1640) when an emergency stop action is determined.

[0137] Figure 17 is a flowchart of a scene analysis process of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0138] Referring to FIG. 17, scene analysis (see FIG. 16 S1640) can analyze risk factors in a scene, predict the movement path of pedestrians, assess collision risk, and determine an optimal avoidance route (S1700).

[0139] FIG. 18 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0140] Referring to FIG. 18, the method of operating an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may further include a step (S1840) of reducing the speed of the vehicle, increasing the stopping distance, activating a pedestrian alarm system, and yielding to a pedestrian when a deceleration action is determined.

[0141] That is, the operating method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a step of detecting whether a pedestrian is a person subject to consideration when a stop action is determined (S1810), a step of extending a stop time if the pedestrian is a person subject to consideration (S1820), a step of performing an additional safety confirmation process when all pedestrians have crossed the road (S1830), and a step of reducing the speed of the vehicle, increasing the stopping distance, activating a pedestrian alarm system, and yielding to the pedestrian when a slow-down action is determined (S1840).

[0142] FIG. 19 is a flowchart of an operation method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0143] Referring to FIG. 19, the method of operating an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may further include a step (S1940) of maintaining the speed of the vehicle and continuously monitoring the surroundings when normal operation is determined.

[0144] That is, the operating method of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention may include a step of detecting whether a pedestrian is a person subject to consideration when a stop action is determined (S1910), a step of extending the stop time if the pedestrian is a person subject to consideration (S1920), a step of performing an additional safety confirmation process when all pedestrians have crossed the road (S1930), and a step of maintaining the speed of the vehicle and continuously monitoring the surroundings when normal operation is determined (S1940).

[0145] FIG. 20 is a flowchart of a process for continuously monitoring the surroundings of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0146] Referring to FIG. 20, the process of continuously monitoring the surroundings (see FIG. 19 S1940) may include a step of monitoring whether there is a pedestrian who suddenly appears, and determining and performing an action based on the status of the pedestrian who suddenly appears and the crosswalk (S2000).

[0147] Meanwhile, the method of operating a node for an autonomous driving data management system using a distributed ledger technology according to an embodiment of the present invention described above can be implemented as a computer-executable program code and provided to an electronic device so as to be executed by a processor in a state stored in various non-transitory computer readable media.

[0148] For example, in a non-transitory computer-readable medium storing computer instructions that, when executed by a processor of an electronic device, cause the electronic device to perform operations, the operations may include, when a stop action is determined, detecting whether a pedestrian is a person of interest, if the pedestrian is a person of interest, extending the stop time, and if all pedestrians have crossed the street, performing additional safety checks.

[0149] FIG. 21 is a block diagram of an electronic device for an ethical decision-making model of artificial intelligence according to one embodiment of the present invention.

[0150] Referring to FIG. 21, the electronic device (2100) may include a storage unit (2110), a communication unit (2130), and a processor (2120).

[0151] The storage unit (2110) may include at least one of volatile memory and non-volatile memory. For example, the volatile memory may include DRAM, SRAM, SDRAM, DDR SDRAM, FeRAM, MRAM, PRAM, PoRAM, or ReRAM. For example, the non-volatile memory may include flash memory, mask ROM, PROM, OTPROM, EPROM, EEPROM, a hard disk, or an optical disk.

[0152] The communication unit (2130) can enable communication between the electronic device (2100) and another electronic device through at least one of a short-range communication module (2131), a wireless communication module (2132), and a wired communication module (2133).

[0153] The processor (2120) may include a RAM (2121), a ROM (2122), a main CPU (2123), a GPU (2124), first to n interfaces (2125-1 to 2125-n), and a bus (2126). Here, the RAM (2121), the ROM (2122), the main CPU (2123), the GPU (2124), and the first to n interfaces (2125-1 to 2125-n) may be connected to each other via the bus (2126).

[0154] A command set for system booting, etc. may be stored in ROM (2122). When a turn-on command is input and power is supplied, the main CPU (2123) may copy the operating system stored in the storage unit (2110) to RAM (2121) according to the command stored in ROM (2122) and execute the operating system to boot the system. When booting is complete, the main CPU (2123) may copy various stored application programs to RAM (2121) and execute the application programs copied to RAM (2121) to perform various operations.

[0155] The main CPU (2123) can access the storage unit (2110) and perform booting using the operating system stored in the storage unit (2110). In addition, the main CPU (2123) can control various operations of the electronic device (2100) using various programs and data stored in the storage unit (2110).

[0156] The GPU (2124) can create a screen containing various objects such as icons, images, and text, and can perform various calculations.

[0157] The first to nth interfaces (2125-1 to 2125-n) may be connected to the various components described above. One of the interfaces may be a network interface that connects to an external device via a network.

[0158] While the embodiments of the present invention have been illustrated and described above, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the appended claims and their equivalents.

[0159]

[0160] <Explanation of symbols>

[0161] 100, 2100: Electronic devices

[0162] 110, 2110: Storage

[0163] 120, 2120: Processor

[0164] 2130: Ministry of Communications

Claims

1. In an electronic device for an ethical decision-making model of artificial intelligence, storage; and Processor; including; The above processor, When deciding to stop, detect whether the pedestrian is a subject of consideration, If a pedestrian is a subject of consideration, the stopping time is extended, An electronic device for an AI ethical decision-making model that performs additional safety checks when every pedestrian crosses the street.

2. In paragraph 1, The above processor, An electronic device for an artificial intelligence ethical decision-making model that performs object recognition of items carried by a pedestrian to detect whether the pedestrian is a person subject to consideration.

3. In paragraph 1, In the above additional safety verification process, The above processor, If the above pedestrian shows unsteady movements, extend the stopping time and inform the pedestrian that it is safe. An electronic device for ethical decision-making in artificial intelligence that sounds an alarm to pedestrians when they are distracted.

4. In paragraph 1, The above processor, An electronic device for an ethical decision-making model of artificial intelligence that activates emergency braking, alerts surrounding vehicles, contacts emergency services, performs scene analysis, and waits until safety is confirmed when an emergency stop action is decided.

5. In paragraph 4, In the above scene analysis, The above processor is an electronic device for an artificial intelligence ethical decision-making model that analyzes risk factors in a scene, predicts the movement paths of pedestrians, assesses collision risks, and determines the optimal avoidance route.

6. In paragraph 1, The above processor, An electronic device for an artificial intelligence ethical decision-making model that reduces the vehicle's speed, increases the stopping distance, activates a pedestrian alarm system, and yields to pedestrians when a deceleration action is decided.

7. In paragraph 1, The above processor, An electronic device for an artificial intelligence ethical decision-making model that performs the process of maintaining the vehicle's speed and continuously monitoring the surroundings when deciding to operate normally.

8. In paragraph 7, In the process of continuously monitoring the above surroundings, The above processor, An electronic device for an ethical decision-making model of artificial intelligence that monitors whether a pedestrian suddenly appears and decides and performs actions based on the status of the pedestrian and the crosswalk that suddenly appears.

9. In the method of operation of an electronic device for an ethical decision-making model of artificial intelligence, When a stop action is decided, a step of detecting whether the pedestrian is a subject of consideration; If a pedestrian is a subject of consideration, steps are taken to extend the stopping time; and A method of operating an electronic device for an ethical decision-making model of artificial intelligence, comprising the step of performing an additional safety verification process when all pedestrians cross the road.

10. In paragraph 9, A method of operating an electronic device for an ethical decision-making model of artificial intelligence, wherein the step of detecting whether the pedestrian is a person to be taken into consideration includes a step of performing object recognition of an item carried by the pedestrian.

11. In paragraph 9, The above additional safety verification process is: If the above pedestrian shows unstable movements, a step of extending the stopping time and informing the pedestrian that it is safe; and An operating method of an electronic device for an ethical decision-making model of artificial intelligence, comprising: a step of sounding an alarm to a pedestrian when the pedestrian is looking away; 12. In paragraph 9, An operating method of an electronic device for an ethical decision-making model of artificial intelligence, further comprising the steps of activating emergency brakes, alerting surrounding vehicles, contacting emergency services, performing scene analysis, and waiting until safety is confirmed, when an emergency stop action is decided.

13. In paragraph 12, The above scene analysis is, An operating method of an electronic device for an ethical decision-making model of artificial intelligence, comprising steps of analyzing risk factors in a scene, predicting the movement paths of pedestrians, assessing collision risks, and determining an optimal avoidance route.

14. In paragraph 9, An operating method of an electronic device for an ethical decision-making model of artificial intelligence, further comprising steps of reducing the speed of the vehicle, increasing the stopping distance, activating a pedestrian alarm system, and yielding to pedestrians, if a deceleration action is determined.

15. In paragraph 9, An operating method of an electronic device for an ethical decision-making model of artificial intelligence, further comprising the step of maintaining the speed of the vehicle and continuously monitoring the surroundings when a normal operation is determined.

16. In paragraph 8, A method of operating an electronic device for an ethical decision-making model of artificial intelligence, wherein the process of continuously monitoring the surroundings includes a step of monitoring whether there are pedestrians who suddenly appear, and determining and performing actions based on the status of the pedestrians and crosswalks who suddenly appear.

17. A non-transitory computer-readable medium storing computer instructions that, when executed by a processor of an electronic device, cause the electronic device to perform an operation, the operation comprising: When a stop action is decided, a step of detecting whether the pedestrian is a subject of consideration; If a pedestrian is a subject of consideration, steps are taken to extend the stopping time; and A non-transitory computer-readable medium comprising: a step of performing an additional safety check process when all pedestrians have crossed the street;

Images