Artificial Skin in Robotics -- A Comprehensive Interface for System-Environment Interaction

Kurzfassung

Modern robotic systems are gradually escaping their fenced workspace and begin to physically interact with humans. If the physical barriers between the workspace of the robotic system and the human fall away, new safety and interaction concepts for the physical human robot interaction have to be developed. These concepts have to make sure, that, e.g. during a joint assembly task within a car body neither the human nor the interacting robotic system are endangered. Sensor systems are a key element to detect changes in the environment and sense actions of the human user. One aspect herein is the detection of direct physical contact between the robotic system and the environment or the human user. In order to ensure safety of human and robotic system a close surveillance of the covering structure of the robotic system with respect to physical contact is desired. A promising solution are spatially distributed tactile sensors that are mounted on the covering structure of the robotic system. The acquired tactile information can be utilized to initiate adequate collision reaction strategies or as a means of intuitive human machine communication. A major challenge for the development of tactile sensors for robotic systems is the adaptation to the complex, often 3D-curved surfaces of modern robotic systems. This adaptation is not or only insufficiently possible with the available tactile sensors. The analysis of the state-of-the-art reveals, that the applied materials and manufacturing technologies restricts the majority of tactile sensors to the application on planar or developable surfaces. If robotic systems operate in an unstructured and time varying environment collisions can no longer be avoided. That is why, amongst other means, the surface of the covering structure of the robotic system has to be equipped with a passive mechanical damping layer. In order not to constrain the sensitivity of the tactile sensors, the tactile sensors have to be integrated on top of the mechanical damping layer. Consequently, the tactile surface sensors have to be stretchable to allow for the underlying mechanical damping layer to deform in case of a collision. In addition, the tactile surface sensors have to be overload proof to withstand the high indentation forces that occur in case of a collision. In the past, the majority of approaches towards tactile sensors for robotic systems focussed on high spatial resolution and sensitivity. A future integration into a robotic system or the mechanical robustness of the tactile sensors have often been neglected. However, if the development was focussed on mechanical robustness, the resulting tactile sensors lack the required sensitivity. The analysis of the current state-of-the-art of science and technology impressively shows that, considered individually, all requirements can be fulfilled – but not their combination. Therefore the focus of this thesis is the derivation of a solution of the goal conflict between the desired high sensitivity and the required mechanical robustness. Human skin is considered as a design metaphor for the development of a multi functional artificial skin concept that, next to providing the required sensory capabilities, exhibits mechanical deformation and damping properties required for the operation on a robotic system. Based on the analysis of the current state-of-the-art of science and technology and the outcome of own previous work design paradigms for the solution of the described goal conflict are proposed. The development and the structure of this thesis are based on the design methodology for mechatronic systems presented in VDI 2206. Herein the V-model, known from software engineering, is adapted for the development of mechatronic systems. During the system design an overall concept for a scalable artificial skin is derived. The concept allows the adaptation of the properties of the artificial skin to the respective application site on the robotic system. Besides the scalability of sensor surface area, spatial resolution and sensitivity the pursued approach accounts for scalability of the underlying manufacturing processes that is required for the successful integration into robotic systems. In addition, a concept for the acquisition and preprocessing of the tactile data is proposed. Based on the functional partitioning, known from software design, the desired functional range of an artificial skin is divided in adjustable functional components. During the domain specific design concepts for the individual functional components are derived based on the integrative design of tailored materials and scalable manufacturing processes. As an example, novel electrically conductive polymer based circuit tracks are developed in order to enable the required elastic deformability of the sensitive surface area of the artificial skin. The properties of a future artificial artificial skin system can be anticipated based on the conducted FEM simulation. Currently no standardized test procedure for the assessment of the functionality of tactile sensor exists, therefore a simplified test procedure for the verification of the desired properties of the artificial skin is proposed. Within the system integration exemplarily a tactile surface sensor for the acquisition of normal indentation forces is implemented. For this, the required functional components are combined and the underlying manufacturing processes are field-tested. The resulting artificial skin prototypes are identified on a specialized testbed and applied on a robotic system. The suitability for daily use of the artificial skin prototypes is examined in a collision detection scenario on the DLR LWR III. The conducted tests demonstrate, that the proposed overall design concept allows for the development of an artificial skin that is scalable with respect to sensor surface area, spatial resolution and sensitivity. The prototypes and the conducted experiments verify that the presented artificial skin can be operated on 3D-curved surfaces of modern robotic systems and that the goal conflict between sensitivity and a collision tolerant design can be solved.