Detect and Grasp Objects in Sand Guided by Tactile Sensor

Robots rely on vision and touch as primary means to perceive and interact with the external environment. In scenarios where visual perception is limited or unavailable, tactile sensing plays a crucial role in an interaction application. This paper presents a novel grasping framework that utilizes a vision-based tactile sensor to enhance the robot’s ability to explore and grasp invisible objects. The framework leverages the original tactile image acquired by the sensor to detect object contact and adjust grasping strength by analyzing the movement patterns of marker points in the tactile image. Using the framework, we achieve accurate positioning and non-destructive grasping of target objects in sand.