Tactile sensing system design issues in machine manipulation

Research in robotics and automation is gradually revealing the importance of tactile information in the control of machine manipulation systems. Substantial research efforts have been devoted to the construction of compact, high resolution force sensing arrays which employ sophisticated transduction and processing techniques. A variety of systems have been experimentally investigated and a widespread optimism exists regarding the potential use of complex tactile sensing systems in robotic end effectors. Unfortunately, relatively few multi-detector systems have seen actual have seen actual use in real manipulation systems. Those designs that have been applied in automatic environments have almost invariably been used in static circumstances for simple contact imaging. As a consequence of these efforts, it is becoming clear that much work remains to be done before machine touch can be understood and then used to enhance the performance of a dynamic manipulation process. The slow progress in the development of comprehensive tactile sensing systems indicates that the fundamental problem is not simply one of transducer array design and fabrication. Advancements in this area will require: (1) understanding new concepts related to contact detection and image formation as well as the use of contact information to control grasp and to aid in task planning; and (2) the development of actual sensing systems which can be used first to experimentally explore important issues in machine manipulation, and later as a basis for the future design of practical and economic tactile sensing systems. The development of appropriate tactile sensors for research applications will require an exhaustive design effort aimed at understanding the architecture of these systems at all levels, including: (1) transducers and preprocessors which acquire data indicating the type of contact between end effector surfaces and an object and which prepare that data for transmission; (2) multiplexing and transmission systems which efficiently supply sensor data to the controller; and (3) tactile focus control systems which, in order to maximize system transmission efficiency, will dynamically select which sensors will be interrogated for information which will be integrated with other sensory input. This paper reviews preliminary work aimed at understanding the general issues and trade-offs governing the design of real tactile sensing systems. Also, specific designs emphasizing practical necessities such as simplicity, reliability, and economy will be discussed along with plans to integrate this system into the Utah/MIT Dextrous Hand.