Research on Object Grasping Prediction Model of Flexible Robotic Arm Based on Feedforward Neural Network
Yixin Tang

School of Mechanical Engineering, Xihua University, Chengdu 610039；

Abstract：Object grasping is the core part of robot automation, which has irreplaceable value in the fields of industrial assembly, logistics sorting and service robot. However, the coupling effect of the physical characteristics of the object and the operating parameters in the actual scene leads to a significant fluctuation in the grasping success rate of the flexible manipulator, and the traditional method is difficult to achieve accurate prediction. Therefore, this paper proposes a prediction model of object grasping of flexible manipulator based on feedforward neural network. The model takes the two-dimensional position of the object, the grasping force, the angle and the type of the object as the input features. Through the adaptation of the double hidden layer structure to the ReLU and Sigmoid activation functions, the grasping success probability is output, and the two-class cross entropy loss and Adam optimizer are used to improve the training efficiency. The experimental results show that the accuracy is 89.2 %, and the mean square error is 0.073. It can effectively reveal that the success rate of grabbing items with handles is significantly higher than that without handles, and the stability can be improved by horizontal angle and moderate high intensity.

Keywords: Feedforward Neural Network; Flexible Robotic Arm; Object Grasping; Success Probability Prediction; Operation Parameter Optimization

References

[1]Zhou Z, Zuo R, Ying B, et al. A sensory soft robotic gripper capable of learning-based object recognition and force-controlled grasping[J]. IEEE Transactions on Automation Science and Engineering, 2022, 21(1): 844-854.

[2]Azizkhani M, Godage I S, Chen Y. Dynamic control of soft robotic arm: A simulation study[J]. IEEE Robotics and Automation Letters, 2022, 7(2): 3584-3591.

[3]Xiao W, Liu C, Hu D, et al. Soft robotic surface enhances the grasping adaptability and reliability of pneumatic grippers[J]. International Journal of Mechanical Sciences, 2022, 219: 107094.

[4]Liu Y, Su J, Li X, et al. A systematic automated grasping approach for automatic manipulation of fabric with soft robot grippers[J]. Industrial Robot: the international journal of robotics research and application, 2023, 50(4): 623-632.

[5]Xie Q, Wang T, Zhu S. Simplified dynamical model and experimental verification of an underwater hydraulic soft robotic arm[J]. Smart Materials and Structures, 2022, 31(7): 075011.

[6]Khadivar F, Billard A. Adaptive fingers coordination for robust grasp and in-hand manipulation under disturbances and unknown dynamics[J]. IEEE Transactions on Robotics, 2023, 39(5): 3350-3367.

[7]Zhang H, Peeters J, Demeester E, et al. A CNN-based grasp planning method for random picking of unknown objects with a vacuum gripper[J]. Journal of Intelligent & Robotic Systems, 2021, 103(4): 64.

[8]Hu W, Wang C, Liu F, et al. A grasps-generation-and-selection convolutional neural network for a digital twin of intelligent robotic grasping[J]. Robotics and Computer-Integrated Manufacturing, 2022, 77: 102371.