Scholarlink

<p>This study aims to develop and validate a resilient distributed dynamic evacuation guidance system that maintains performance despite component failures. We investigate if a distributed algorithm combined with a persistent guidance strategy can perform comparably to a fully functional system, particularly when evaluated using agents that model realistic human cognitive processes during evacuation.&#xa0;A distributed algorithm (<i>UpdateSign2</i>) was integrated with a persistent guidance strategy. To simulate human behavior, a cognitive agent was developed incorporating herd behavior, risk aversion, and trust degradation. This trust model was derived from Virtual Reality (VR) experiments with human subjects, which measured adherence to dynamically changing signs. The system’s performance was assessed via multi-agent simulations under various failure scenarios, network delays, and with both simple and cognitive agents.&#xa0;The proposed system achieved evacuation performance nearly equivalent to that of a perfectly functioning system, even with component failures. The VR experiments confirmed that trust in guidance signs decays exponentially with increased directional changes. Simulations with cognitive agents validated the system’s robustness but indicated potentially longer evacuation times. The algorithm was found to inherently suppress frequent sign changes, thus mitigating the negative effects of trust degradation.&#xa0;The proposed distributed guidance system provides a robust and resilient framework for evacuations in harsh conditions. Validating systems with realistic cognitive agents is crucial. The system shows significant potential for real-world application, effectively guiding evacuees even when parts of the system fail.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Resilience in Evacuation Guidance: A Cognitive Agent Approach to System Failures

  • 期刊论文

  • 发表日期：

    2026年3月6日

查看全文
  • Akira Tsurushima,
  • Shuhei Miyano

摘要

This study aims to develop and validate a resilient distributed dynamic evacuation guidance system that maintains performance despite component failures. We investigate if a distributed algorithm combined with a persistent guidance strategy can perform comparably to a fully functional system, particularly when evaluated using agents that model realistic human cognitive processes during evacuation. A distributed algorithm (UpdateSign2) was integrated with a persistent guidance strategy. To simulate human behavior, a cognitive agent was developed incorporating herd behavior, risk aversion, and trust degradation. This trust model was derived from Virtual Reality (VR) experiments with human subjects, which measured adherence to dynamically changing signs. The system’s performance was assessed via multi-agent simulations under various failure scenarios, network delays, and with both simple and cognitive agents. The proposed system achieved evacuation performance nearly equivalent to that of a perfectly functioning system, even with component failures. The VR experiments confirmed that trust in guidance signs decays exponentially with increased directional changes. Simulations with cognitive agents validated the system’s robustness but indicated potentially longer evacuation times. The algorithm was found to inherently suppress frequent sign changes, thus mitigating the negative effects of trust degradation. The proposed distributed guidance system provides a robust and resilient framework for evacuations in harsh conditions. Validating systems with realistic cognitive agents is crucial. The system shows significant potential for real-world application, effectively guiding evacuees even when parts of the system fail.