Research On Acoustic Manipulation Technology Of Small Objects Based On Depth Reinforcement Learning

As a new type of non-invasive control means without direct contact with the driver,acoustic manipulation has attracted extensive attention of scholars in the field of non-invasive control of small objects,while the use of acoustic manipulation of small objects on the Chladni plate has broad application prospects in precision medicine,industrial control of droplets and particles,etc.Traditional acoustic wave manipulation believes that the impact of acoustic wave on particles is random in the process of motion.In recent years,some studies have shown that the impact of acoustic vibration on small objects on the Chladni plate is orderly,but the lack of accurate and effective modeling means limits its application.In order to solve this problem,this paper introduces the artificial intelligence method into the modeling of the acoustic force field on the Chladni plate,and on the basis of the acoustic field model,the particle motion is controlled by giving different acoustic frequencies.This paper focuses on the modeling of acoustic field and the control of particle motion.The main research contents of this paper are as follows:(1)In order to observe the impact of sound waves on small objects,we combined Faster R-CNN algorithm and Local Weighted Regression(LOESS)algorithm to model the acoustic field.Firstly,image recognition technology based on Faster RCNN algorithm is used to identify the position of particles on the Chladni plate,and calculate the displacement of small particles at different positions on the Chladni plate under a fixed acoustic frequency.After accumulating a large amount of displacement data,based on the local weighted regression(LOESS)algorithm,a complete acoustic displacement model on the plate is established to analyze how the acoustic force field affects the movement of small objects on the Chladni plate.(2)Different acoustic frequencies have different effects on the control of small objects on the Chladni plate.Because the traditional algorithm only selects the local optimal solution each time and does not carry out backtracking processing,it is difficult to achieve accurate control because the optimal solution is obtained only in a few cases.To solve this problem,this paper introduces artificial intelligence algorithm after establishing acoustic displacement model,uses PPO algorithm for learning,and realizes acoustic wave controlling particle motion by switching different frequencies.In this paper,we also set up a simulation model to conduct the simulation experiment of particle manipulation on the Chladni plate.Finally,this paper will control the fragile instant coffee granules based on the acoustic manipulation platform built.By comparing with the simulation experiment,the feasibility of the modeling and control method is verified,which proves that the model can realize the motion control of small particles.