A Flexible Electrical Capacitance Tomography Sensor for Proximity Object Recognition

Abstract

The ability to recognize objects plays a crucial role in the application of robots. Electrical capacitance tomography (ECT) effectively addresses the limitations of other types of proximity sensors in object recognition. This study introduces a flexible ECT sensor, examines the characteristics of coplanar ECT sensors in measurement strategies, and utilizes Tikhonov regularization for position and target contour reconstruction. A detailed analysis of the sensor’s response and performance metrics is conducted. Under proximity conditions, the support vector machine (SVM) achieves a recognition accuracy of up to 93.1% for 15 different object items. To validate the practicality and effectiveness of the sensor, we design a simple physical experiment that combines close-range identification based on the ECT sensor with obstacle avoidance using a robotic arm. The experimental results indicate that the robotic arm can select appropriate avoidance strategies based on the potential impact of the object on itself, thereby enhancing obstacle avoidance efficiency. This indicates that through ECT sensors, effective proximity recognition of objects in the environment is achievable, offering a new solution to enhance the perceptual abilities and autonomy of robots in complex environments.