Failure State Estimation Using Soft Tactile Fingertip in Insertion Tasks

During assembly, failure states (FSs) may occur in insertion tasks due to factors such as inaccuracy in template matching algorithms and grasping position displacement. Therefore, it is essential to estimate the FS because the parts could undergo damaged if the robotic arm continues the insertion process during the failure insertion state. Therefore, this paper proposes a soft tactile fingertip to estimate FS during part insertion tasks. The proposed architecture employs two 3-axis Hall sensors and four cylindrical magnets as well as a feed-forward neural network, to estimate the contact state between the assembly parts, and the control flow is presented. AFS is defined as the state when a part of the grasped object collides with other objects, resulting in failure of an insertion task. Moreover, a FS is estimated in real-time based on the changes in the magnetic flux density measured via the sensors. The experimental results confirmed that the fingertip was capable of estimating the FS during part insertion of an electrical inverter. Moreover, the fingertip required a normal force of 0.5 N to classify the contact states. We demonstrated that the arm moved back to the initial assembly position, once a FS was detected.