Research And Algorithm Of Precise Positioning For Taking And Putting Materials Based On Compound Robot

The application of highly efficient industrial robots plays an important role in the current critical period of industrial development.In order to improve the production efficiency and automation level of the production workshop to meet the requirements of current industrial robot development,this paper studies the track planning,time optimization,and intelligent material grabbing of the composite robotic arms in cigarette factories.Based on the studies,several algorithms are proposed to enable the efficient and precise grab and placement of the aluminum foil roll materials by composite robotic arms in the cigarette factory.To improve the efficiency of the robotic arm,this paper proposes a time-optimal trajectory planning method based on the 3-5-3 segmented polynomial interpolation algorithm based on the non-linear inertial weight particle swarm algorithm.First,we compare our algorithm with the existed polynomial interpolation method and prove that our 3-5-3 segmented polynomial interpolation algorithm achieves smoother motion parameter curves than the traditional cubic polynomial interpolation method,and lower computational complexity than the quintic method.Furthermore,we propose a weighted particle swarm optimization(WPSO)to optimize the convergence time of the 3-5-3 segmented polynomial interpolation algorithm of our trajectory planning method.Simulations on MATLAB confirm that compared to the traditional particle swarm optimization(PSO),WPSO achieves faster convergence speed(only requires 20-50 iterations)and better convergence level.We can thus conclude that with our algorithm applied in the trajectory planning of the robotic arm,both high planning efficiency and stable motion parameters can be achieved.To improve the grab precision of the robotic arm,this paper proposes a prediction model of the optimal grabbing position of the material based on generative grasping convolutional neural network(GG-CNN).We train the neural network with the Cornell data set and achieve a lost function of 0.115,which proves the good performance of our model.Furthermore,to solve the material placing problem,this paper proposes a localization method for the robotic arms to position the placement of materials,based on the you only look once(v5s)(YOLOv5s)detection table.We study the network architecture of YOLOv5 s and use the lightweighted network model of YOLOv5 s to train our self-collected data set.Furthermore,we train the network based on YOLOv5 s network evaluation indicators and verify the model applicability through the analysis of the mean average precision(MAP)and the loss function curves.The results confirm that our model can achieve high accuracy in the material placement of the robotic arm.To further evaluate the performance of our proposed models,we perform the experiment of material grabbing and placing on the robot operating system(ROS)platform,with the composite robot model built by the first generation Kinect camera.With our algorithms,the robotic arm can accurately grab and place the aluminum foil roll material in the cigarette factory,with different initial positions of the material and robotic arm.This confirms that the proposed methods in this paper can improve the efficiency and precision of the robotic arm and can be applied to the cigarette factory.