A Comprehensive Pattern Recognition Neural Network for Collision Classification Using Force Sensor Signals

In this paper, force sensor signals are classified using a pattern recognition neural network (PRNN). The to show if there is collision or not. our previous work, the joints positions of 2-DOF robot were used estimate external signal, which was attached at end-effector, and joint torques based on multilayer feedforward NN (MLFFNN). current estimated signal joints’ from work as inputs proposed designed PRNN, its output whether found PRNN trained scaled conjugate gradient backpropagation algorithm tested validated different data training one. results prove that effective in classifying signals. Its effectiveness for cases 92.8%, non-collisions 99.4%. Therefore, overall efficiency 99.2%. same methodology repeated another algorithm, Levenberg–Marquardt (PRNN-LM). structure PRNN-LM also signals, 99.3%, slightly higher than first PRNN. Finally, comparison with other classifiers included. This shows PRNN-LM.