Article Preview
TopIntroduction
In the past decade, nanomanipulation, which aims at manipulating and handling nanometer size objects and structures with nanometer precision, has become a hot topic of research. As a first and critical step for achieving any complex functional nano devices, nanomanipulation can find numerous applications in various fields like medicine, biotechnologies (DNA and protein study) and data storage or material science (nanotube or surface film characterization) (Gauthier & Regnier, 2010; Requicha, 2008).
However, in order to manufacture nanotechnology products, the challenges in automatic nanomanipulation and handling of objects in nano scale require cross-disciplinary studies. Nowadays, assemblies of small nano structures built by nanomanipulation typically consist of ten to twenty components, and may take an experienced user a whole day to construct using Atomic Force Microscope (AFM) as the manipulator. To efficiently accomplish such tasks or even more complex ones, the manipulation process should be more automated with less human intervention (Liu et al., 2008).
One of the obstacles to achieve efficient and reliable nanomanipulation is that the physical and chemical phenomenon at this scale has not been well understood. Furthermore, visual feedback is not available while the AFM probe is used as a manipulation (pushing) tool (Li et al., 2003). Hence, a significant amount of work on modeling interactive forces during manipulation was introduced in Tafazzoli et al. (2005), Li et al. (2005), Yang and Jagannathan (2006). Additionally, AFM tips and some of the experimental samples used in the nanomanipulation are fragile. Improper applied force could damage these nano objects or even the tip. Thus, designing controllers for the manipulation and handling of nano scale objects poses a much greater challenge in terms of accommodating the nonlinearities and uncertainties in the system (Eslami & Jalili, 2011).
Meanwhile, the control of nonlinear uncertain dynamic systems has attracted extensive interests from the control community (Krstic et al., 1995; Qu, 1998). Among various methodologies, adaptive control techniques modify the parameters of the controller in response to the error feedback signal to pick up the prompt changes in system operating conditions (Slotine & Li, 1991). In an attempt to parameterize the unknown plant nonlinearities, either neural networks or fuzzy logic techniques have been utilized (Jagannathan & Lewis, 1996; Su & Stepanenko, 1994) due to their universal approximation ability (Lee & Tomizuka, 2000).
Typically, manipulation or pushing of micro/nano particles today is undertaken by simple open loop control strategy in Requicha et al. (2009) or by human operators in Li et al. (2005). Therefore, in this paper, an adaptive feedback controller integrating neural networks (NN) and error sign function is proposed to perform autonomous nanomanipulation tasks. A two-layer NN structure is employed to estimate the unknown system dynamics in an online manner and a robust term of error sign function is added to compensate the NN functional reconstruction errors and external unknown disturbances. Compared with traditional sliding mode designs, this method generates a continuous-time control signal while waiving the requirement of infinite bandwidth and chattering (Patre et al., 2008; Yang et al., 2011), which may deteriorate the nonlinearities of the piezoelectric actuators within AFM. Theoretically, the asymptotical stability performance and the boundedness of the NN weights and other signals in the closed-loop system are shown by using Lyapunov method.
This paper is organized as follows. First, the modeling of the nanomanipulation process is presented. Then, the robust integral of NN and error sign feedback control law is developed. The stability of the overall formation is presented and then we present numerical simulations. Finally, we will provides some concluding remarks.