ATRACT: A Trustworthy Robotic Autonomous system to support Casualty Triage

在越南战争中，随着“黄金时刻”的出现，美国撤离直升机改变了士兵的生存能力。这依赖于空中优势和相对自由的行动，并且从那以后一直是英国/美国/北约处理战场伤亡的方法。然而，最近低成本、精确、肩扛式地对空导弹的扩散和有效性极大地扰乱了乌克兰的直升机行动，从而增加了伤亡撤离行动的风险。此外，前线军医在世界上最恶劣和最恶劣的环境中工作，他们经常冒着生命危险出诊，在战斗部队附近需要他们时挺身而出。他们经常被要求在给定的时间内监测多起伤亡，并根据受伤的严重程度优先考虑他们应该首先照顾的人。因此，在竞争激烈的环境中，在传统直升机运送伤员的速度缓慢或不可用的情况下，迫切需要提高伤员的存活率。人工智能（AI）和机器人自主系统（RAS）的最新进展为应对这一挑战提供了新的和未来的机会。与此相一致，拟议的attract系统是一项颠覆性创新，通过设计、开发和现场测试一个值得信赖的无人机驱动的RAS，以一种新颖的方式解决了这一未满足的需求，以帮助一线医务人员在创伤后的第一个“白金十分钟”内做出决策。吸引力将采用跨学科和变革性的研究方法，重点是：1)利用无人机在复杂地形上的先进机动，对受伤士兵进行精确搜索和定位；2)结合先进多模态传感的新平台，超越机器人系统检测前线士兵的最先进算法；3)实时监测他们的受伤严重程度和生命体征，以实现有效的分诊预测/更新；4)医疗应急响应团队可用的地方。在途中医疗队接近时向其提供实时伤亡信息，从而实现更有效的人员资源管理和伤员优先排序，从而减少在地面的时间，最大限度地提高生存能力，并最大限度地减少一线医务人员受到攻击的风险。人工智能和RAS是许多行业（例如制造业、农业、运输、医疗保健等）的驱动力，并有助于解决人类面临的一些最紧迫的问题。许多此类技术在可信度（技术上稳健，道德上遵守和合法）方面存在重大限制，主要是因为它们通常使用“黑箱”方法，其中人工智能元素在使用和从多个来源操作数据的方式中通常不太可见和透明，并且经常表现出无意识的偏见，导致决策缺乏控制。此外，它们没有针对不断变化的条件和/或环境设置提供情境化服务或定制化干预措施。attract将通过设计和开发过程来解决这些限制，这些过程符合最新的道德和法律国防部人工智能标准，以及军事医疗实践，并结合世界卫生组织手术清单的原则，将医疗考虑与数据质量、避免偏见和系统可靠性因素结合起来。我们将确保《吸引力》是透明的、一致的和可解释的，以便在设计、开发和测试的每个阶段都能系统地解决潜在的偏见、法律和医疗合规以及国防部伦理问题。在这方面取得的成功成果将彻底改变一线保健服务、伤员后送以及提供紧急和挽救生命的医疗援助的方式，从而产生重大的健康、社会和经济效益。

In the Vietnam War, American evacuation helicopters transformed soldier survivability with the emergence of the 'Golden Hour'. This relied on air superiority and relative freedom of movement and has been the UK/US/NATO approach to battlefield casualty treatment since. However, recent proliferation and effectiveness of low-cost, accurate, shoulder-launched ground-to-air missiles has significantly disrupted helicopter operations in Ukraine and thus, presenting a heightened risk to Casualty Evacuation (CASEVAC) operations. Moreover, frontline army doctors work in world's most harsh and hostile environments, and often risk their lives while marching out and stepping in when they are needed near fighting forces. They are often required to monitor multiple casualties at a given time and prioritise whom they should be attending first based on the severity of injuries. Thus, there is an urgent unmet need for enhancing casualty survival in a contested environment where conventional helicopter CASEVAC is slow or unavailable.Recent advancement in Artificial Intelligence (AI) and Robotic Autonomous System (RAS) provides new and future opportunities to meet this challenge. In line with this, the proposed ATRACT system is a disruptive innovation to address this unmet need in a novel way by designing, developing and field-testing a trustworthy drone-driven RAS to help frontline medics in decision-making in the first 'platinum ten minutes' following trauma. ATRACT will adopt an interdisciplinary and transformative research approach focusing on: 1) accurate search and localisation of injured soldiers using advanced manoeuvring of a drone in difficult terrains, 2) a novel platform that combines advanced multimodal sensing, beyond state-of-the-art algorithms for a robotic system to detect frontline soldiers, 3) real-time monitoring of their injury severity and vital signs for effective triage prediction/update, and 4) where medical emergency response team is available, real-time casualty information to the enroute medical team as it approaches, enabling more effective crew resource management and casualty prioritisation, thereby reducing time on the ground to maximise survivability and to minimise risk of the frontline medics being attacked. AI and RAS are the driving forces in many industries (e.g., manufacturing, agriculture, transport, healthcare, etc.) and helping to address some of the most pressing issues facing humankind. Many such technologies have a major limitation of trustworthiness (technically robust, ethically adherent and lawful) and mainly because they typically use a "black box" approach, where AI elements are often less visible and transparent in the way data is used and operationalised from multiple sources, and frequently exhibits unconscious biases resulting in lack of control in decision-making. Moreover, they do not provide contextualised services or customised interventions to changing conditions and/or environmental settings. ATRACT will address these limitations via design and development processes which comply with the latest ethical and legal MoD AI standards, and military medical practice, incorporating principles from the WHO Surgical Checklist to align medical considerations with data quality, bias avoidance and system reliability factors. We will ensure that ATRACT is transparent, consistent and interpretable so that potential bias, legal and medical compliance, and MoD ethics can be addressed systematically at every stage of design, development and testing with expert-in-the-loop. Successful results in this context will revolutionise the way frontline health services, casualty evacuation and the delivery of emergency and lifesaving medical aid is delivered, resulting in significant health, social and economic benefits.