Robustness and adaptivity: advanced control and estimation algorithms for the transverse dynamic atomic force microscope

观察单个生物分子的动态行为和相互作用是促进生物医学研究的长期目标。标准做法是使用几种扫描探针显微镜（SPM）之一，主要是原子力显微镜（AFM）。原子力显微镜的原理很简单：一个水平取向的悬臂，带有一个非常锋利的尖端，在感兴趣的物体上移动，从而捕获三维形貌图像。然而，生物医学科学中感兴趣的基本分子间事件的低力和快时间尺度通常远远超出商业AFM的操作范围，商业AFM通常需要几分钟的图像。因此，虽然AFM可以在生理条件下提供生物分子的亚分子分辨率（参见图1）。使用真空的电子显微镜），AFM仍有两个显著的缺点需要克服。成像速度慢：一个典型的256 x256像素的图像需要60秒才能产生。成像期间相互作用力过大：成像生物分子相互作用的一个重大挑战是探针和样品之间通常存在的力干扰甚至损坏生物分子。为了解决这些问题，我们将联合收割机与控制理论的最新进展相结合，新的SPM仪器，目前正在开发中的布里斯托，生产一种新的扫描探针显微镜，能够在不损坏这些脆弱样品的情况下对它们进行成像：因此，布里斯托的横向动态力显微镜（TDFM）将代表SPM仪器和生物分子研究的突破。在布里斯托的TDFM中，探针垂直于样品表面（而不是平行于样品表面，如在AFM中）对准，并在样品平面内振荡。振荡的幅度随着尖端-样品分离距离的减小而减小。探针振荡的幅度可以用作测量信号以控制具有亚纳米精度的探针-样品分离。在使用这种控制方法时，在扫描过程中，探针不应接触样品表面。新颖的控制方法通过控制悬臂高度的快速移动来创建高速TDFM（HS-TDFM）（z-运动）-控制快速放置任何待观察（生物）标本（x-y-motion）-估计样品悬臂力，例如范德华力，更好地了解丰富的测量信息，以便更快、更简单地融合和处理从HS-TDFM获得的数据。在实现这一切之前，TDFM将被重新设计，以采用高精度传感器技术，并获得最佳的动态性能。现代控制方法将包括线性鲁棒控制方法、非线性滑模控制方法、非线性自适应控制方法等。（神经网络）控制、使用滑动模式的现代估计/观测器技术和自适应原理。挑战将是-在1 MHz以上的带宽上实现实际控制;-理解和利用非线性HS-TDFM动态以更好地解释数据;- 开发新的估计器/观测器，结合自适应和滑动模式方法的范例，用于信号和参数识别;- 结合新的估计和控制方法，以改善HS-TDFM的控制系统。由此产生的HS-TDFM将是一个真正的非接触式成像技术的空间分辨率与AFM相当，相互作用力低于AFM。HS-TDFM将显示皮牛顿力灵敏度，并提供大量的信息，从直接观察相互作用的生物分子。它将根据观察生物过程的需要每秒收集多个图像。这不仅将有利于生命科学，而且还将支持材料科学中的SPM用户、纳米尺寸系统的生产者以及微处理器等纳米电子学。

Observing the dynamic behaviour and interactions of single biomolecules is a long-standing goal to facilitate bio-medical research. Standard practice is to use one of several scanning probe microscopes (SPMs), principally atomic force microscopy (AFM). The principle of an AFM is simple: a horizontally oriented cantilever with a very sharp tip is moved across the object of interest, allowing the capture of a three dimensional topographical image. However, the low forces and fast timescales of the fundamental inter-molecular events of interest in bio-medical sciences, generally lie far outside the operational range of commercial AFMs, which typically require several minutes per image. Thus, while AFMs can provide sub-molecular resolution of biomolecules under physiological conditions (cf. electron microscopy which uses a vacuum) there are two significant disadvantages of AFMs still to be overcome.- slow imaging rates: A typical 256x256 pixel image takes 60 seconds to produce.- excessive interaction forces during imaging: A significant challenge to imaging biomolecular interactions is that the forces typically present between the probe and the sample disturb or even damage the biomolecules.To counter these issues, we will combine the latest advances in control theory with the novel SPM instrumentation, currently in development in Bristol, to produce a new scanning probe microscope capable of imaging these fragile samples without damaging them: Thus, Bristol's transverse dynamic force microscope (TDFM) will represent a breakthrough in both SPM instrumentation and the study of biomolecules.In Bristol's TDFM, the probe is aligned perpendicularly to the sample surface (rather than parallel to it, as in AFMs) and oscillates in the plane of the sample. The amplitude of oscillation decreases as the tip-sample separation distance decreases. The amplitude of the probe oscillation can be used as a measurement signal to control the probe-sample separation with sub-nanometer precision. When using this control method, at no point during scanning should the probe come into contact with the sample surface.Novel control methods will create a high-speed TDFM (HS-TDFM) by-controlling the fast movement of the cantilever height (z-motion)-controlling the fast placement of any (biological) specimen to be observed (x-y-motion)-estimating sample-cantilever forces, e.g. Van-der Waals forces, to better understand the wealth of measured information for faster and simpler fusion & processing of data obtained from the HS-TDFM.Before any of this is possible, the TDFM will be redesigned to incorporate highly precise sensor technology and to obtain the best possible dynamic behaviour. The modern control approaches will include linear robust control approaches, nonlinear sliding mode control, nonlinear adaptive (neural network) control, modern estimation/observer techniques using sliding modes, and adaptive principles.The challenges will be to-achieve practical control at bandwidths above 1MHz;-understand & exploit the nonlinear HS-TDFM dynamics for better data interpretation;-develop novel estimators/observers combining the paradigms of adaptive and sliding mode methods, for signal and parameter identification;-incorporate novel estimation and control approaches for improving the control system of the HS-TDFM.The resulting HS-TDFM will be a true non-contact imaging technique capable of comparable spatial resolution and lower interaction forces than AFMs. The HS-TDFM will display pico-Newton force-sensitivity and provide a wealth of information from direct observation of the interacting biomolecules. It will collect multiple images per second as required for observing biological processes. This will not only benefit life-sciences but also support SPM users in material science, producers of nano-sized systems, and in nano-electronics, e.g. microprocessors.