| dc.description.abstract | One of the grand challenges in robotics is robust grasping of unknown objects. This is particularly important when robots expand its territory from industry floors to domestic service applications where the object prior knowledge is not often available. As a result, sensor-based grasping is more desirable. Ideally, with the assistance of object sensing, robotic fingers can respond to subtle changes in object pose right before grasping and adjust operations dynamically. Moreover, the object material and structure information can help planners better estimate the force distribution, impact characteristics and friction coefficients for a more robust grasping.
However, current sensors have difficulties in satisfying these requirements. Tactile/force sensors may change object poses or even damage the object, which leads to slow or failed grasping. Non-contact long-distance sensors such as camera, LIDAR, radar, sonar suffer from occlusion or blind zones. Therefore, non-contact near-distance sensing is the optimal solution. Unfortunately, existing near-distance sensors based on optical, electric-field, and acoustic signals still cannot satisfy these grasping requirements. Electric-field sensors have difficulties in targets with low dielectric contrast to air. The optical ones lack lateral resolution and are not effective for optically-transparent or highly-reflective targets. Acoustic-based sensors could work on distance ranging and material/structure sensing, but fail on thin-film, porous, or sound-absorbing targets.
To address these issues, a new finger-mounted non-contact dual-modal and dual-sensing-mechanism (DMDSM) sensor for near-distance ranging and material/structure differentiation is studied and developed, which is based on two modalities and sensing mechanisms: pulse-echo ultrasound (US) and optoacoustics (OA). In both modalities, the object distance is estimated from the Time-of-Flight (ToF) of the US/OA signal, whose frequency spectra are used to extract the distinctive features of the material/structure. The development of the DMDSM sensor is conducted as follows. First, the prototype of the DMDSM sensor is designed, fabricated, and characterized. Testing is conducted on conventional objects and optically and/or acoustically challenging targets (OACTs) to characterize its performance. Second, to simplify the DMDSM sensor design and operation, a single wideband ultrasound transmitter and receiver is investigated where both US and OA collection can be initiated by a single laser pulse. Third, to expand to areal mapping or imaging, a new self-focused US/OA transceiver and a flat scanning mirror are studied to steer laser and ultrasound beams over the target with customized patterns. At last, optically-transparent focused (OTF) ultrasound transducers are explored, which are helpful to miniaturize the DMDSM sensors while enhancing their performances. | |
