An Optical Sensor Design: Concurrent Multi-axis Force Measurement and Tactile Perception.

PhD ThesesForce and tactile sensing have experienced a surge of interest over recent decades, as they convey useful information about the direct physical interaction between the sensor and the external environment. A robot end effector is a device designed to interact with the environment. End effectors such as robotic hands and grippers can be used to pick up, place or generally manipulate objects. There is a clear need to equip such end effectors with appropriate sensing means to be able to measure tactile and force information. Work to date has explored these two modalities separately. Tactile sensors have been developed for integration with gripper fingertips or as skins embedded with the outer side of manipulators, mainly to measure normal force and its distribution across a surface patch. On the other hand, force sensors have commonly been integrated with the joints of robotic arms or fingers to measure external multi-axis forces and torques via the connected links. We observe that a force sensor cannot measure tactile information, and current tactile sensors cannot accurately measure force information. This can become a particular issue when integrating force sensors remotely to measure forces indirectly, especially if the connecting link is flexible or, generally, difficult to model potentially impacting negatively on the force estimates. We aim to provide a solution for an integrated sensor capable of measuring tactile and force information at the point of contact, i.e., on the fingertip of a robot hand or arm. In this thesis, we explore the idea of integrating the two sensing modalities, tactile and force sensing, in one sensor housing with the signal acquisition being performed by a single monocular camera acting as the transducer. The hypothesis is that an integrated force/tactile sensor will perform in a better way than having these sensor modalities separated. This thesis shows that an integrated sensor achieves a tactile sensing performance comparable to existing vision-based tactile sensors and at the same time proves to provide more accurate force sensor information whilst extending the field of similar vision-based sensors from 3 DoF to 6 DoF. In addition, the tactile sensing element of our sensor is not affected by the patterns superimposed on to the flexible element of comparable vision-based sensors used to infer force information. In this thesis, we have implemented several sensor prototypes; designs and experimental analyses for each prototype are being provided. The manufactured sensor prototypes prove the validity of the proposed vision-based dual-modality sensing approach, and the proposed sensing principle and structure shows high versatility and accuracy, as well as the potential for further miniaturization, making the proposed concept suitable for integration with standard robot end effectors