Acoustic Detection of Contact State Changes of Deformable Linear Objects

Introduction In assembly, objects have to be brought in contact with other objects. In order to deal with the imprecisions in the geometry of the mating objects and in the manipulating system, a lot of research concentrates on describing assembly tasks using contact states and on finding corresponding algorithms to detect contact state transitions or even to determine the current contact state based on sensor data [1-3], [6], [8], [9], [13]. Typical force-based approaches only evaluate the low frequency effects of changes in the contact situation. Most often, contact forces are measured to decide on contact state changes using simple thresholds. The sound caused by collisions or by different friction coefficients in different contact situations is not evaluated. In this paper, the sound and the vibration generated by a contact is used to “hear” changes in contact situations. The sound may be recorded using an ordinary microphone. However, the use of a microphone is impractical as the noise of the whole enviroment is recorded as well. As this noise can overlay the signal of interest, the microphone must be placed very close to the expected point of contact. Because of that we use an impact sound sensor that only records the vibration of an object in touch with the sensor and is insensitive to airbourne sound. This sensor is mounted on the robots gripper so we can measure the impact sound of the held object directly, without having to place one or more microphones in the workspace (Fig. 1, left). Using this setup we will try to detect three typical changes in the contact situation between a leaf spring and its environment.