Vision Based Tactile Sensor Using Transparent Elastic Fingertip for Dexterous Handling

Humans have the ability to sense the weight and friction coefficient of the grasped object with their distributed tactile receptors. The ability makes it possible for humans to prevent from the slippage of manipulated object or collapsing the object. Such dexterous handlings are achieved by feeding back the signals from the receptors to muscle control system through neural networks. Therefore, it may be a key point for establishing dexterous handlings of robots when we try to mimic skilled human functions. For tactile sensing of robots, several methods and sensors have been proposed by using electrical resistance, capacitance, electromagnetic component, piezoelectric component, ultrasonic component, optical component, and strain gauge (Shinoda, 2002), (Lee & Nicholls, 1999). There exist many problems of these sensors to be solved for establishing practical ones. For an example, the sensor which consists of elastic body and strain gauges requires a lot of gauges and the wiring. Moreover, the signal processing is not simple to obtain the values of the contact forces and the friction coefficients (Maeno et al, 1998). On the other hand, optical sensors have been introduced because wiring is not required in the contact part to the object (Ohka et al, 2004), (Ferrier & Brockett, 2000), (Kamiyama et al, 2003). The introduction of optical sensor makes the size small and the wiring simple. However, the sensing of friction coefficient is not considered in those papers. Piezoelectric sensors have a certain potential to solve the problems of size and wiring but there has not been a practical solution yet for measuring friction coefficient. It is required for establishing dexterous handlings of robots to provide a purpose-built sensor for the measurement of friction coefficients between robot hand and the target surfaces. So as to avoid multiple usage of tactile sensors, we have proposed a new design of tactile sensors for multiple measuring of contact information including friction coefficient (Obinata et al, 2005).