Safe Acting and Manipulation in Human Environments: A Key Concept for Robots in our Society Sami Haddadin, Sven Parusel, Rico Belder, Alin Albu-Sch¨ affer, and Gerd Hirzinger Abstract— In this paper we review our work on safe acting and manipulation in human environments. In order for a robot to be able to safely interact with its environment it is primary to be able to react to unforeseen events in real-time on basically all levels of abstraction. Having this goal in mind, our contributions reach from fundamental understanding of human injury due to robot-human collisions as the underlying metric for “safe” behavior, various interaction control schemes that ground on the basic components impedance control and collision behavior, to real-time motion planning and behavior based control as an interface level for task planning. A significant amount of this work has found found its way into international standardization committees, products, and was applied in numerous real-world applications. I. I NTRODUCTION LWR-I LWR-II LWR-III KUKA LWR Fig. 1. The generations of DLR light-weight robots (LWR-I, LWR-II, and LWR-III) and the commercialized version (KUKA LWR). Finally, first robotic systems gained sufficient control capabilities to perform delicate and complex manipulation and physical human-robot interaction (pHRI) tasks that re- quire the dynamic exchange of physical forces between the robot and its environment. The fully torque-controlled DLR Lightweight Robot III (LWR-III) is such a device [1] and was recently commercialized by the robot manufacturer KUKA (KUKA LWR) [2]. This step made it possible to automate difficult and up to now still manually executed assembly tasks. In particular, the achieved sensible and fast manipulation capabilities [3], [4], [5], [6] of the robot prevent damage from the handled potentially fragile objects and humans directly interacting with the device. Recently, there is strong interest in making classical safety barriers, as e.g. fences or light barriers, obsolete for these interactive devices in order to enable direct physical cooperation between human and robot. For understanding the risks of this undertaking we performed a series of safety investigations [7], [8], [9], [10], [11], [12], [13], which led to fundamental insight into the potential injury a human would suffer due to a collision with a robot. Furthermore, we developed human-friendly interaction control and motion schemes that enable the robot S. Haddadin, S. Parusel, R. Belder, A. Albu-Sch¨ affer, and G. Hirzinger are with the Institute of Robotics and Mechatronics, DLR - German Aerospace Center, Wessling, Germany, contact: sami.haddadin@dlr.de to show sophisticated real-time responses on interaction force level, motion planning, and real-time task planning [14], [4], [15], [16], [17], [18]. Generally, our approach of embodying reactivity on all levels of robot design and control is to our understanding the core to safe acting and manipulation in human environments. Consequently, the careful design and selection of methods that satisfy this requirement was our main premise. In this paper we give an overview of the developed anal- ysis tools, control schemes, motion planners, and real-time behaviors for robots that are sought to act and manipulate in human environments. We intend to give a “bird’s eye” view on the available repertoire of tools and how the developed methodologies, insights, and algorithms impact robotics in general. II. TECHNOLOGIES AND METHODS A. Lightweight & mechatronic robot design The most basic step for building robots that interact with dynamic environments is to design them compact, light- weight, and with high payload. Only light structures are capable of appropriate physical reaction to external forces, i.e. have low intrinsic impedance. Secondly, the robot’s proprioceptive sensorization is a key element. Apart from standard motor position sensing, joint torque sensing together with accurate flexible joint dynamics modeling enable real torque control and the sensation of contact forces. In this line of thinking we have developed a series of torque controlled lightweight robots at DLR that are suitable for a diverse range of applications involving space, industry, medical, and domestic use. Figure 1 shows the history of the DLR Lightweight robots, resulting in its commercialized version: the KUKA LWR [2]. Apart from minor modifications, this manipulator has exactly the same design as the 3 rd gener- ation of the DLR Lightweight robots [1], which are kine- matically redundant, 7-DoF, joint-torque controlled flexible joint robots. The current version is the result of 15 years of research that produced three consecutive generations. Since the LWR-III weighs 13.5 kg and is able to handle loads up to 15 kg, an approximate load-to-weight ratio of 1 is achieved 1 . The robot is a modular system and the joints are linked via carbon-fiber structures. The electronic parts, including power converting elements are integrated into the structure of the arm. Each joint is equipped with a motor position and a joint-torque sensor. Additionally, a 6-DoF force sensor can be embedded in the wrist. All electronics, motors, and gears are integrated into the arm, which makes the robot very compact and portable. 1 Please note that the nominal payload for the KUKA LWR is 7 kg, but it is able to handle up to 15 kg for research purposes.