EAGER/Collaborative Research:Bumpless Re-Engagement in Shared Control

这个早期概念探索性研究（EAGER）奖是一个跨学科项目，汇集了一位自适应控制专家和一位异常反应认知基础专家。重点是高度自动化、关键业务、智能数字服务的提供商，在这些服务中，人工主管通常监控自动控制器，但如果发生自动化无法处理的异常事件，则必须接管控制。目标是确保这种转换安全、平稳和可靠地发生。工业研究合作伙伴将提供一组来自实际关键业务软件服务的异常响应案例。智能服务越来越依赖于自主计算机制来快速处理大量事务。考虑到自动化的规模和依赖性，即使是很小的破坏性事件也会迅速串联起来，当异常情况发生时，人类的监督可能会导致延迟和错误的响应，从而挑战自动化的能力。该项目承诺通过提供预测自动化活动何时有饱和控制风险的异常可视化，使人类主管能够顺利地参与异常的开发，以及(b)快速地进行干预，以阻止级联问题。性能测试将基于异常响应的实际情况，并将评估异常后的重新参与是平稳还是颠簸的，对异常的响应能力的增加，以及可以通过新的共享控制形式有效监督的自主活动的范围。

This EArly-Concept Grant for Exploratory Research (EAGER) award is an interdisciplinary project that brings together an expert in adaptive control with an expert in the cognitive basis of anomaly response. The focus is on providers of highly automated, business critical, smart digital services in which human supervisors normally monitor automatic controllers, but must take over control if anomalous events occur that the automation is not designed to handle. The objective is to ensure that such transitions occur safely, smoothly, and reliably. An industrial research partner will provide a set of cases of anomaly response from actual business critical software services.Smart services increasingly rely on autonomous computational mechanisms to quickly handle extremely large numbers of transactions extremely. Given the scale and dependence on automation, even small disrupting events can cascade quickly, and human supervision may lead to late and erroneous responses when anomalies occur that challenge automated capabilities. This project promises to re-engage human supervisors in the developing anomaly (a) smoothly, by providing anomaly visualizations that anticipate when the automated activity is at risk of saturating control, and (b) quickly, so that supervisors can intervene to block cascading problems. Performance testing will be based on actual cases of anomaly response and will estimate how re-engagement following anomalies is smooth or bumpy, the increase in responsiveness to anomalies, and the range of autonomous activities that can be supervised effectively with the new form of shared control.