A State-Matching-Based Method for Identifying Intrusive Object Data and Evaluating Collision Features Using Robotic E-Skin Proximity Perception

Abstract

Proximity perception is a crucial foundation for robotic collision safety control, and e-skin proximity offers unique advantages in this field. However, traditional e-skin proximity data struggle to effectively distinguish between intrusive object data and the data from the robot itself and the surrounding environment, making it accurately evaluate collision features of intrusive objects. This article proposes a state-matching-based method for identifying intrusive object data and evaluating collision features using e-skin proximity. By establishing a nonintrusive feature model, the process extracts the nonintrusive feature data corresponding to the current robot state through state matching and compares it with the current e-skin proximity data. This allows for the effective identification of intrusive object data and the accurate and rapid evaluation of collision features, such as approach distance (AD) and approach orientation (AO). In the static experiments, the proposed method significantly improves the accuracy of evaluating AD and AO. In the dynamic experiments, the method proposed in this article demonstrated a high degree of alignment between the evaluated values and the actual values for AD and AO. Furthermore, this article analyzes the impact of the sampling state differentiation (SD) threshold during the construction of the nonintrusive feature (NIF) model on the subsequent evaluation of the robot’s dynamic AD. It demonstrates that a lower threshold for sampling SD threshold will effectively enhance the stability of the robot’s dynamic feature evaluation. Through experiments on safe collision control of robots along predetermined trajectory, it is proven that the method proposed in this article can achieve safe collision speed control for robots in human-robot interaction (HRI) scenarios, where the robot operates at a speed of 0.5 m/s.