Stochastic Recruitment and Broadcast Feedback of Cellular Control Systems and Its Application to Muscle Actuators

在过去的几十年里，大多数研究团体一直在追求“小世界”科学和技术。纳米技术和智能材料社区一直在探索构建和操纵具有新功能的微小构建块的技术。生物化学和生物工程研究活细胞的分子水平行为。然而，为了从这些微小的构建块中获得有意义的功能，有必要处理大量微小事物的“集体行为”。我们面临的挑战是如何填补微观世界与执行有意义功能的宏观系统之间的空白。这个NSF项目的主要目标是开发一种新的方法来弥合两者之间的差距。大量独立单元的集体行为，称为细胞系统，将根据受生物系统启发的随机控制方法进行研究和控制。与今天的人造工程不同，生物系统中的细胞并不是决定性的。细胞的神经支配是通过生化扩散过程进行的，这一过程基本上是随机的。因此，单个细胞的个体行为是一个随机过程。尽管如此，大量细胞的集体行为是高度协调和可靠的。考虑到单个细胞接收的通信和控制命令的数量有限，这是一种惊人的行为，我们希望探索对巨大细胞系统的有效操纵。为此，提议的项目提出了两个关键概念，“随机招募”和“广播反馈”，这将揭开庞大的蜂窝系统的集成行为的神秘面纱。在这种随机控制系统中，中央控制器观察细胞单元的总输出，将输出与命令输入进行比较，并将差异均匀地广播给所有细胞。该控制器不要求单个单元服从确定性命令，而是通过全局广播通道将聚合错误通知单元，将最终控制决策留给单个单元。每个单元接收到相同的累计误差信号，随机地抛硬币做出控制决策。然而，与标准硬币不同的是，正面和反面的概率是用广播信号调制的。这反过来又允许整个细胞系统跟踪所需的总输出轨迹。整个系统的整体行为，虽然不是确定的，但是高度可预测和可靠的。尽管大部分细胞死亡或无功能，但整个系统仍能执行任务。所提出的控制方法具有许多高影响的应用领域。基于所提出的控制方法，将开发出一种由无数微小致动单元组成的新型肌肉致动器。这种随机控制有可能成为血管生成和组织工程的生物过程控制以及环境保护和监测的群体机器人控制的有效方法。在这个NSF项目中，将广泛探索生物工程、机器人和控制社区的合作研究。该项目的教育部分将包括发展关于随机细胞控制的新的大学间研究生科目和K-12推广工作。

In the last few decades the majority of research communities have been pursuing "small world" science and technology. Nanotechnology and smart materials communities have been exploring technologies for constructing and manipulating tiny building blocks that exhibit novel functionality. Biochemistry and biological engineering deal with molecular-level behavior of living cells. To attain meaningful functions out of these tiny building blocks, however, it is necessary to deal with "collective behavior" of vast numbers of tiny things. The challenge is to fill the void between the tiny world and macro-level systems that perform meaningful functions. The major objective of this NSF project is to develop a novel methodology for bridging the gap between the two. Collective behavior of vast numbers of independent units, called cellular systems, will be investigated and controlled based on a stochastic control method inspired by biological systems. Unlike today's engineered artifact, cells in a biological system are not deterministic. Cells are innervated through biochemical diffusion processes, which are fundamentally stochastic. The individual behavior of a single cell is therefore a random process. Nonetheless, the collective ensemble behavior of vast numbers of cells is highly coordinated and reliable. Considering the limited amount of communication and control commands that individual cells receive, this is an amazing behavior, which we would like to explore for effective manipulation of vast cellular systems. To this end, the proposed project presents two key concepts, "stochastic recruitment" and "broadcast feedback", which would demystify ensemble behavior of vast cellular systems. In this stochastic control system, a central controller observes an aggregate output of the cellular units, compares the output to a commanded input, and broadcasts the discrepancy to all the cells uniformly. Instead of demanding the individual cells to obey deterministic commands, the proposed controller only informs the cells about the aggregate error through a global broadcast channel, leaving the final control decision to the individual cells. Each cell, receiving the same aggregate error signal, flips a coin to make a control decision stochastically. Unlike a standard coin, however, the probabilities of heads and tails are modulated with the broadcast signal. This in turn allows the whole cellular system to track a desired aggregate output trajectory. The ensemble behavior of the overall system, although not deterministic, is highly predictable and reliable. Although a large fraction of the cells are dead or non functional, the overall system is still capable of performing the task.The proposed control method has a number of high-impact application areas. A new type of muscle actuators consisting of numerous tiny actuator cells will be developed based on the proposed control. This stochastic control has the potential to be an effective approach to biological process control for angiogenesis and tissue engineering as well as to swarm robot control for environment protection and monitoring. Collaborative research will be explored broadly across the bioengineering, robotics, and control communities during this NSF project. Educational components of the project will include development of new inter-university graduate subjects on stochastic cellular control and K-12 outreach efforts.