SCH: INT: Pediatric motor rehabilitation via socially interacting robot swarms

每年，仅在美国就有1万名婴儿患上脑瘫，4000名婴儿患有先天性脊柱和大脑缺陷。这些婴儿的运动迟缓不仅对家庭，而且对整个社会都有终身的社会和经济后果。目前，有行动障碍的婴儿无法获得移动机器人所能提供的适应性、动态和自主的游戏活动。我们目前对发育中儿童显著的神经可塑性和神经肌肉变化的科学理解与儿童康复的最新水平不同步，这仍然不能满足数百万有特殊需求的学龄儿童的需求。到4岁或5岁时，这些孩子中的大多数已经永久地失去了早年每天丰富的运动和认知刺激。此外，绝大多数的治疗活动是由成年照顾者发起和指导的；然而，排他性和持续的成人控制可能会使孩子筋疲力尽，并剥夺孩子自由探索她的动态环境的机会。该项目的研究活动将引入机器人玩具群高剂量儿童康复所必需的科学创新。机器人玩具群将被设计成对儿童的大脑活动做出反应，这样他们就可以在身体和智力上都参与进来。设想的创新不仅将打破婴儿运动康复的现状，而且可以减轻目前特殊需要儿童照顾者的压力。通过这种方式，该项目将直接促进国家科学基金会确定的关键期望社会成果之一，即改善社会中个人的福祉，并间接提高美国的经济竞争力，主要是通过预期提高运动残疾个体在其一生中的生产力。该提案提出了一项研究计划，该计划能够开发一个丰富的婴儿运动康复环境，该环境涉及到在游戏活动中与婴儿进行社交互动的机器人玩具群的刺激。我们的假设是，适当设计的群体运动行为可以使婴儿参与积极的游戏，并且基于非侵入性图像和脑电图（EEG）测量，儿童群体互动可以实时协调。研究计划沿着四个关键轴扩展，旨在实现以下具体目标：(i)设计专门用于社交儿童机器人交互的集体运动行为；（ii）开发资源感知感知算法，用于实时监测和跟踪这种动态交互；（iii）产生脑电图数据处理方法，以提供受试者参与的实时反馈；最后（iv）将这项技术整合到一个适应性和反应性丰富的环境中，提供干预和评估运动康复结果。该建议通过在丰富的环境中引入协调的社交互动智能玩具作为自适应人机界面，推动了儿童运动康复的发展。智力上的贡献包括(a)利用定向的基于游戏的互动，对发育中儿童的神经可塑性和神经肌肉变化进行新的干预途径；(b)一种新的数学架构，用于调节人类群体的相互作用，而不是基于直接的人类输入或明确的命令；(c)婴儿参与高移动和互动智能玩具的定量神经生理学指标；(d)利用深度网络的新的资源感知、健壮和一致的多目标跟踪算法。这些贡献将是重要的，因为它们将确定协调物体运动行为与婴儿感知和行动之间的具体基本关系，并利用它们来推动治疗干预。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Each year, 10,000 infants in the U.S. alone develop cerebral palsy and 4,000 infants have birth defects of the spine and the brain. The motor delays for these infants have lifelong social and economic consequences, not only for the families, but also for the society as a whole. Currently, infants with mobility impairments do not have access to the type of adaptive, dynamic, and autonomous play-based activity that mobile robots can provide. Our current scientific understanding of the significant neuroplastic and neuromuscular changes in developing children is not at pace with the state-of-the-art in pediatric rehabilitation, which still cannot meet the needs of millions of school-aged children with special needs. By the age of 4 or 5, most of these children have permanently lost the daily rich motor and cognitive stimulation of the early years. In addition, the vast majority of therapy activities are initiated and directed by adult caregivers; yet exclusive and continuous adult control can be exhausting and deprive the child from the chance to freely explore her dynamic environment. This project’s research activities will introduce the scientific innovations necessary for robotic toy swarm-enabled high-dosage pediatric rehabilitation. The robotic toy swarms will be designed to be responsive to children’s brain activity so they can be both physically and intellectually engaged. The envisioned innovations will not only disrupt the status quo in infant motor rehabilitation, but can also alleviate current stressors to special need children caregivers. In this way, this project will promote directly one of the key desired societal outcomes identified by NSF, which is the improvement of well-being of individuals in society, and indirectly that of increasing the economic competitiveness of the U.S. primarily through the expected increase in productivity for individuals with motor disabilities over the course of their life.This proposal puts forth a research plan that enables the development of an enriched infant motor rehabilitation environment, which involves stimulation by swarms of robotic toys that are socially interacting with infants in play-based activities. Our hypothesis is that appropriately designed swarm motion behaviors can keep infants engaged in active game-play, and that child-swarm interaction can be coordinated in real-time based on non-invasive imagery and electroencephalograpy (EEG) measurements. The research plan expands along four key axes that aim at the following specific goals: (i) the design of collective motion behaviors specifically for social child-robot interaction; (ii) the development of resource-aware perception algorithms for real-time monitoring and tracking of this dynamic interaction; (iii) the production of EEG data processing methods to provide real-time feedback on subject engagement; and finally (iv) the integration of this technology into an adaptive and reactive enriched environment that delivers intervention and assesses motor rehabilitation outcomes. This proposal pushes the envelope in pediatric motor rehabilitation by introducing coordinated socially-interacting smart toys as an adaptive human-machine interface within an enriched environment. The intellectual contributions include (a) new intervention pathways to neuroplastic and neuromuscular changes in developing children that utilize directed play-based interaction; (b) a new mathematical architecture for regulating human-swarm interaction that is not based on direct human input or explicit commands; (c) quantitative neurophysiological metrics of infant engagement with highly mobile and interactive smart toys; and (d) new resource-aware, robust and consistent multi-target tracking algorithms that leverage deep networks. These contributions will be significant because they will identify specific fundamental relationships between coordinated object motion behavior and infant perception and action, and exploit them to drive therapeutic interventions.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.