Division Of Nano-robotic Manipulation Meta-tasks And Study Of CNT Picking Applications

As the integrated circuit characteristic size into the significant range of quantum effects,the development of integrated chip technology has come to the post-Molar era,where the power consumption bottleneck of chips makes it hard to maintain the current ratio of size reduction.Carbon nanotube(CNT)is one of the materials with a high potential to replace silicon.Peking University,Intel,IBM,and others have long been conducting indepth research on the performance of CNT and the use of CNT to manufacture micro and nanodevices.The diameter of CNT is usually several nanometers.Traditional equipment and techniques are not very suitable for characterizing and measuring the properties of individual CNT.However,with the aid of a scanning electron microscope and the construction of the nanomanipulation robot,a variety of precise treatments of CNT can be performed at the nanoscale using robotic manipulation techniques.Aiming at the problem of automatic CNT pickup for nanomanipulation robot,this paper proposes a method of automatic CNT pickup by learning the motion skills of the CNT pickup meta-tasks.Based on the meta-task division criterion of nanomanipulation robot,the meta-tasks motion model for nanomanipulation robot CNT pickup was established,and the invocation strategy of nanomanipulation robot meta-tasks was designed.And then the experiments of automatically picking up CNT by invoking multiple meta-tasks were implemented,which verified the feasibility of the nanomanipulation robot learning human manipulation skills through meta-tasks and reproducing the method.The successful picking of a single CNT of 4.8μm in length laid the foundation for the nanomanipulation robot to automatically perform other micro and nano manipulation tasks.The main research contents of this thesis are as follows:(1)In order to determine the key points of meta-task division for nanomanipulation robot,this paper has carried out several manual demonstration experiments to analyze the relationship between nanomanipulation robot and manipulated objects,as well as establish the criteria for dividing meta-tasks.The planar motion trajectory and clarity changes of the nanomanipulation robot are obtained by image processing techniques,and then a two-step division step was implemented to divide the complete task demonstration of human manipulation into multiple simple meta-tasks,which were successfully applied to the CNT pickup task.(2)To address the issue of how a nanomanipulation robot learns motor skills in human demonstrations,this paper developed the difference between the same meta-task divided by different demonstrations,and proposed a learning strategy for each type of meta-tasks motor skills based on the trajectory characteristics of the meta-tasks.According to the distinction between the meta-task trajectories,multiple motor trajectories of the meta-task were categorized,and an optimization process for optimizing each class of trajectories of the meta-task was developed,which can generate the optimal meta-task demonstration trajectories,these were learned using dynamic motor primitives.The test results show that the dynamic motor primitives can reproduce the demonstration trajectories of the meta-task.(3)The complete manipulation process of the CNT pickup task was analyzed,multiple combinatorial orders of each meta-task in the CNT pickup task under normal conditions were established,and a finite state machine model of the CNT pickup task of the nanomanipulation robot was built to solve the invocation problem of multiple metatasks.Several CNT pickup experiments were carried out to analyze the errors that may occur during the nanomanipulation robotic motion in the microscopic environment,and these errors were classified.For those errors that can be revised,correction meta-tasks were added to the nanomanipulation robot’s finite state machine model of the CNT pickup.(4)Several automatic CNT pickup experiments were carried out.Multiple meta-tasks for CNT pickup were obtained,optimized,and learned utilizing previous techniques.Methods for driving and controlling the nanomanipulation robot were examined,as well as an experimental platform for the automatic picking of CNT.Multiple meta-task automatic CNT pickup experiments were performed starting with the nanomanipulation robot reproduction of a single meta-task.The results showed that a single carbon nanotube with a length of 4.8m can be picked up automatically.This demonstrated that the nanomanipulation robot can learn and reproduce the motor skills in human manipulations through meta-tasks and provided the groundwork for the nano manipulator robot to perform other nanomanipulation tasks automatically in the future.