Object Pose Estimation Using Soft Tactile Sensor Based on Manifold Particle Filter with Continuous Observation

It is desired for wider application of robotic manipulation system that the robot can estimate object pose using simple tactile sensor without relying strongly on visual sensors. In this paper, we propose C-MPF, an extension of the manifold particle filter that can handle continuous and multidimensional tactile information. Multiple channels of soft tactile sensors can be used in the proposed estimation method to identify the object pose by multiple contact motions. The proposed method was experimentally validated using a 5-DOF manipulator equipped with a sponge material tactile sensor with four channels. It was confirmed that C-MPF estimated the position and posture of the cylindrical object by multiple contact motions even with a simple soft tactile sensor by repeating contact observations.