Human-like Decision-making Model

Human-like Decision-making Model refers to decision-making mechanisms that simulate human psychological needs (safety, dominance, achievement, order, relatedness) for autonomous driving simulations. This model enables more realistic interaction-based testing, facilitating compliance with ISO 21448 (SOTIF) by addressing the 'unknown unsafe scenarios' that traditional-rule-based models fail to capture. It allows AD systems to be tested against unpredictable human behaviors, reducing the risk of real-world accidents during the development phase.

Implementation follows three steps: 1. Building a multi-dimensional behavior library based on human need theories. 2. Integrating these behaviors into simulation frameworks like OpenScenario to generate diverse edge cases. 3. Using the results to validate AD system robustness against human-like irrationality. For example, a Taiwanese automotive supplier could reduce real-world test-drive costs by 30% and increase SOTIF compliance by 40% within the first year of implementation, significantly lowering the risk of post-market safety recalls.

Three main challenges exist: Data-scarcity, cross-departmental silos, and evolving regulations. To overcome data-scarcity, enterprises should use GDPR-compliant synthetic data generation. For silos, establishing a cross-functional task force including AI engineers and risk managers is essential. Regarding regulations, companies must align with UNECE WP.29 requirements early. A 90-day agile implementation starting with high-risk scenarios (e.g., highway overtaking) is recommended to demonstrate ROI quickly to stakeholders.

Winners Consulting Services Co., Ltd. specializes in Human-like Decision-making for Taiwan enterprises, delivering compliant management systems within 90 days. Free consultation: https://winners.com.tw/contact