Touch is one of the primary senses humans use when performing coordinated interaction, but the lack of a sense of touch in the majority of contemporary interactive technical systems, such as robots, which operate in non-deterministic environments, results in interactions that can at best be described as clumsy.
Observing human haptics and extracting the salient information from the gathered data is not only relevant if we are to try to understand the involved underlying cognitive processes, but should also provide us with significant clues to design future intelligent interactive systems. Such systems could one day help to take the burden of tedious tasks off our hands in a similar fashion to how industrial robots revolutionized manufacturing. The aim of the work in this thesis was to provide significant advancements in tactile sensing technology, and thus move us a step closer to realizing this goal.
The contributions contained herein can be broken into two major parts. The first part investigates capturing interaction patterns in humans with the goals of better understanding manual intelligence and improving the lives of hand amputees, while the second part is focused on augmenting technical systems with a sense of touch.
tacTiles, a wireless tactile sensitive surface element attached to a deformable textile, was developed to capture human full-body interactions with large surfaces we come into contact with in our daily lives, such as floors, chairs, sofas or other furniture. The Tactile Dataglove, iObject and the Tactile Pen were developed especially to observe human manual intelligence. Whereas iObject allows motion sensing and a higher definition tactile signal to be captured than the Tactile Dataglove (220 tactile cells in the first iObject prototype versus 54 cells in the glove), the wearable glove makes haptic interactions with arbitrary objects observable. The Tactile Pen was designed to measure grip force during handwriting in order to better facilitate therapeutic treatment assessments. These sensors have already been extensively used by various research groups, including our own, to gain a better understanding of human manual intelligence.
The Finger-Force-Linear-Sensor and the Tactile Bracelet are two novel sensors that were developed to facilitate more natural control of dexterous multi Degree-of-Freedom (DOF) hand prostheses. The Finger-Force-Linear-Sensor is a very accurate bidirectional single finger force ground-truth measurement device that was designed to enable testing and development of single finger forces and muscle activations mapping algorithms. The Tactile Bracelet was designed with the goal to provide a more robust and intuitive means of control for multi-DOF hand prostheses by measuring the muscle bulgings of the remnant muscles of lower arm amputees. It is currently in development and will eventually cover the complete forearm circumference with high spatial resolution tactile sensitive surfaces. An experiment involving a large number of lower arm amputees has already been planned.
The Modular flat tactile sensor system, the Fabric-based touch sensitive artificial skin and the 3D shaped tactile sensor were developed to cover and to add touch sensing capabilities to the surfaces of technical systems. The rapid augmentation of systems with a sense of touch was the main goal of the modular flat tactile sensor system. The developed sensor modules can be used alone or in an array to form larger tactile sensitive surfaces such as tactile sensitive tabletops. As many robots have curved surfaces, using flat rigid modules severely limits the areas that can be covered with tactile sensors. The Fabric-based tactile sensor, originally developed to form a tactile dataglove for human hands, can with minor modifications also function as an artificial skin for technical systems. Finally, the 3D shaped tactile sensor based on Laser-Direct-Structuring technology is a novel tactile sensor that has a true 3D shape and provides high sensitivity and a high spatial resolution. These sensors take us further along the path towards creating general purpose technical systems that in time can be of great help to us in our daily lives.
The desired tactile sensor characteristics differ significantly according to which haptic interaction patterns we wish to measure. Large tactile sensor arrays that are used to capture full body haptic interactions with floors and upholstered furniture, or that are designed to cover large areas of technical system surfaces, need to be scalable, have low power consumption and should ideally have a low material cost. Two examples of such sensors are tacTiles and the Fabric-based sensor for curved surfaces. At the other end of the tactile sensor development spectrum, if we want to observe manual interactions, high spatial and temporal resolution are crucial to enable the measurement of fine grasping and manipulation actions. Our fingertips contain the highest density area of mechanoreceptors, the organs that sense mechanical pressure and distortions. Thus, to construct biologically inspired anthropomorphic robotic hands, the artificial tactile sensors for the fingertips require similar high-fidelity sensors with surfaces that are curved under small bending radii in 2 dimensions, have high spatial densities, while simultaneously providing high sensitivity. With the fingertip tactile sensor, designed to fit the Shadow Robot Hands' fingers, I show that such sensors can indeed be constructed in the 3D-shaped high spatial resolution tactile sensor section of my thesis.
With my work I have made a significant contribution towards making haptics more observable. I achieved this by developing a high number of novel tactile sensors that are usable, give a deeper insight into human haptic interactions, have great potential to help amputees and that make technical systems, such as robots, more capable.