CPS: Medium: Learning for Control of Synthetic and Cyborg Insects in Uncertain Dynamic Environments

这项研究的目的是使合成昆虫和半机械昆虫能够在复杂的环境中运行，例如在户外或在坍塌的建筑中。由于移动平台和环境具有显著的不确定性，学习和适应能力至关重要。该方法包括三个主要方面：(I)通过在平台和环境的不同实例之间共享学习和适应，开发算法来高效地学习最优控制策略和动态模型。(Ii)创建可在低成本、低功率移动平台上运行的控制学习算法。(Iii)开发用于在最少数量的真实世界试验中在线改进政策绩效的算法。拟议的研究将促进实际网络物理系统的学习和适应能力。建议的方法将是普遍适用的，并导致一类新的学习和适应系统，这些系统能够利用多个任务之间的共享属性来显著加快向上学习和适应。这一研究项目的成功将使社会更接近于解决由移动、一次性、搜索和救援机器人组成的团队的巨大挑战，这些机器人可以在不确定和新颖的环境中顽强移动，在灾难情况下找到幸存者，同时消除救援人员的风险。该项目将通过研究和课堂作业为本科生和研究生提供跨学科培训，帮助他们创建将机器人和生活移动平台中的网络和物理方面紧密结合在一起的系统。通过SUPERB暑期项目，工程学专业中代表性不足的学生将体验学习和机器人技术方面的研究。

The objective of this research is to enable operation of synthetic andcyborg insects in complicated environments, such as outdoors or in acollapsed building. As the mobile platforms and environment havesignificant uncertainty, learning and adaptation capabilities arecritical. The approach consists of three main thrusts to enable thedesired learning and adaptation: (i) Development of algorithms toefficiently learn optimal control policies and dynamics models throughsharing the learning and adaptation between various instantiations ofplatforms and environments. (ii) Creation of control learningalgorithms which can be run on low-cost, low-power mobile platforms.(iii) Development of algorithms for online improvement of policyperformance in a minimal number of real-world trials. The proposed research will advance learning and adaptationcapabilities of practical cyberphysical systems. The proposedapproach will be generally applicable and lead to a new class oflearning and adapting systems that are able to leverage sharedproperties between multiple tasks to significantly speed uplearning and adaptation. Success in this research project will bring society closer to solvingthe grand challenge of teams of mobile, disposable, search and rescuerobots which can robustly locomote through uncertain and novelenvironments, finding survivors in disaster situations, while removingrisk from rescuers. This project will provide interdisciplinarytraining through research and classwork for undergraduate and graduatestudents in creating systems which intimately couple the cyber andphysical aspects in robotic and living mobile platforms. Through theSUPERB summer program, under-represented students in engineering willexperience research in learning and robotics.