Robots Teaching Robots: Real-time Optimal Control of Complex Engineering Systems

点击翻译按钮获取中文摘要

This project introduces learner-helper robot pairs to enable the learner robot to use physical experimentation to improve its performance on repetitive task, without accurate analytical or numerical models. The specific challenge is that these tasks -- for example walking on two legs or riding a bicycle -- require a minimal necessary level of performance, below which the robot is unable to function. In the examples, this minimal level of ability corresponds to not falling over. The helper satisfies these minimal requirements while the learner uses repeated trials to improve its performance. For example, the helper might suspend the two-legged walker from a traveling harness or move alongside the bicycle robot providing an additional point of support. As the learner-helper team masters the task, the amount of assistance that the helper can apply is gradually reduced, until the learner is performing at a high level on its own. An analogy is a child learning to ride a bike with the help of an adult moving alongside. The new control technique will enable robots to teach robots in training lines of future factories similar to robots currently used in assembly lines of manufacturing companies. Therefore, the results of this research will benefit the U.S. economy and society. This research also involves several disciplines including mechanical, electrical, computer, and control engineering. The multi-disciplinary approach is expected to broaden the participation of underrepresented groups in research and positively impact engineering education.Optimal control is a branch of control theory that has the potential to revolutionize the creation of intelligent engineering systems, industrial robots, surgical robots, and assistive robots that can improve by repeated experience, somewhat similar to humans. There are many optimal control techniques to control engineering systems. However, almost all currently available techniques require high-fidelity models or a large amount of measured data to mitigate the so-called simulation-reality gap; the gap between the optimal performance predicted by computer simulations and the non-optimal performance observed in real engineering applications. This award supports fundamental research to close the simulation-reality gap when optimal control is applied to engineering systems. Model-based optimal control techniques enable efficient computation but they are subject to conservative control performance. Data-driven optimal control techniques mitigate the detrimental effect of uncertain models, but to do so, they require a large amount of training data. Therefore, scientific barriers must be overcome to realize the full application potential of optimal control techniques. This research will address the knowledge gap that limits the potential and theoretical promise of optimal control theory when applied to complex engineering systems. The new technique promotes optimization of system performance via real-time experiments guided by dedicated teacher robots, instead of optimizing system performance guided only by uncertain model-based predictions and measured data. The technique delivers a transformative approach to control the class of complex, underactuated, and unstable robots, for which obtaining high-fidelity models is challenging, while gathering training data is time-consuming. The research outcomes could potentially provide mainstream paradigms in creating next-generation intelligent machines.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.