The term Industry 4.0 – manufacturing with exploitation of digital technologies – comprises several challenging topics, among others intuitive machine programming, advanced maintenance-enabling technologies as well as flexible logistics. This paper gives an overview on recent research and development of our institute (Zentrum für Telematik, ZfT) in this field, the results of which are combined in an Industry 4.0 demonstration factory for the assembly of small satellites systems. We present a total of six tools to be used in such advanced manufacturing systems and their individual advantages. Based on our experience with industrial project partners, customers and visitors of our demonstration factory as well as on the evaluations of the jurors of several awards, we give a qualitative estimate of the effort required to port the individual tools to new production environments. Finally, utilizing our in-house expertise in the New-Space and Industry-4.0 sectors, we give an insight into the benefits achievable using digital manufacturing for small satellite assembly.

The overview of a situation in a search and rescue disaster is the key aspect of an effective assistance. In the recent years the utilization of multicopter with various photogrammetry systems is an upcoming trend and an open field of research. This paper discusses the technical aspects of an automated integral system that will support rescuers during the strategic mission planning and will give situational awareness by instantaneous 3D mapping. The approach combines sensors including a 3D Laserscanner, a thermal camera and an attitude system as a payload unit on an Multicopter. The continuous data fusion and the down link are providing an instant 3D environment map that is continuously revised and updated.

Rescue and security applications call for fast mapping of complex environments. For this purpose we propose a planner that generates paths for multiple unmanned aerial vehicles (UAVs) that need to cover multiple disjoint non-convex polygonal areas with nadir pointing cameras of varying field of view. These fulfill a user specified overlap to allow for taken images being used in different map generation procedures. The proposed planner first decomposes the given areas into suitably sized convex subareas. These are allocated to the available UAVs by a metaheuristic solver while minimizing the longest path of all involved UAVs. Given the allocation, subareas are adjusted by a developed heuristic to satisfy the overlap constraints at the inner borders of the split areas. Finally paths allowing for scanning the areas are generated. Experiments are conducted to demonstrate and validate the results of the planner.

We used a brain-computer interface (BCI) system controlled with event-related potentials (ERPs) evoked by tactile stimulation to control a mobile platform. Eight tactile stimulators were attached in four pairs to the arms, legs and back of the participants (N=12). The electroencephalogram (EEG) was recorded via a modified Emotiv headset. All participants were trained in the laboratory, then four participants controlled the mobile platform in an outdoor environment. Inside the laboratory the participants achieved average accuracies of 72%. Outside four participants achieved average accuracies of 61% (range 52-88%). Technical problems with the responses of the mobile platform and high outside temperatures prevented higher levels of control with the mobile platform. A mobile platform better suited for the discrete control implemented with the BCI or a different control scheme will be needed for future experiments. Nevertheless, subjects were able to control the mobile platform with tactile ERPs using low-cost EEG equipment in a real world environment.

This paper extends the well known efficient RRT* algorithm to handle special environmental structures, such as narrow passages. By applying the watershed algorithm, inspired by image processing, a segmentation component is added to solve the given problem. Therefore the traditional watershed segmentation competes against the newly developed center marker watershed algorithm and the results of different experiments, combining both segmentation methods with the Rapidly Exploring Random Tree* (RRT*) algorithm to the Segmented Rapidly Exploring Random Tree* (SRRT*), are compared for several critical environments. It is shown, the whole trajectory planning process benefits from the segmentation and the SRRT*, using the new center marker watershed, outer performs the traditional RRT* method.

In many future applications autonomous mobile robots will play an important role. One of the key functionalities is the autonomous navigation of these robots. At the moment only physical constraints of the robot itself are considered in planning algorithms. But in real world applications also other kind of constrains like direction constrains or speed limits exist. This paper shows an extension to the RRTCAP* algorithm in order to cope with this kind of constrains. In experimental results this extension is evaluated according some performance measures.

In archaeological studies the use of new technologies has moved into focus in the past years creating new challenges such as the processing of the massive amounts of data. In this paper we present steps and processes for smart 3D modelling of environments by use of the mobile robot Irma3D. A robot that is equipped with multiple sensors, most importantly a photo camera and a laser scanner, enables the automation of most of the processes, including data acquisition and registration. The robot was tested in the Würzburg Residence. Methods for automatic 3D color reconstructions of cultural heritage sites are evaluated in this paper.

In archaeological studies the use of new technologies has moved into focus in the past years creating new challenges such as the processing of the massive amounts of data. In this paper we present steps and processes for smart 3D modelling of environments by use of the mobile robot Irma3D. A robot that is equipped with multiple sensors, most importantly a photo camera and a laser scanner, enables the automation of most of the processes, including data acquisition and registration. The robot was tested in two scenarios, Ostia Antica and the Wurzburg Residence. The paper describes the steps for creating 3D color reconstructions of these renown cultural heritage sites.

The localization of mobile robots is very important for any outdoor application such as search and rescue, reconnaissance, surveillance, and monitoring. Already a lot of research was done in the field of localization; nevertheless an accurate and reliable positioning of a mobile robot is very challenging. The common Global Positioning System (GPS) is a very common and popular for the outdoor localization. But GPS has well known drawbacks, like the limited accuracy of the positioning. For this reason, various sensor data fusion algorithms were developed, which fuse the GPS positioning information and dead reckoning sensors to overcome the drawbacks of GPS and dead reckoning sensors. A fundamental aspect of these fusions is the resulting accuracy of the determined position, which depends on the sensor accuracy as well as on the filter parameters. Usually, the output of such Filter are compared to the rare sensor data, but it is not compared to the real position of the vehicle. In the following, a test bed for localization methods will be introduced. To demonstrate the usage, Kalman Filter were implemented to fuse GPS and odometry sensors to determine the position of a mobile robot in an outdoor environment. The focus of this paper is the calibration and evaluation of the Kalman Filter using the test bed based on a high precision optical positioning system. Therefore, the experimental setup, the Kalman Filter, the synchronization of the high precision positioning system, and the transformation of one system into the other is explained.

Recently, several backpack-mounted systems, also known as personal laser scanning systems, have been developed. They consist of laser scanners or cameras that are carried by a human operator to acquire measurements of the environment while walking. These systems were first designed to overcome the challenges of mapping indoor environments with doors and stairs. While the human operator inherently has the ability to open doors and to climb stairs, the flexible movements introduce irregularities of the trajectory to the system. To compete with other mapping systems, the accuracy of these systems has to be evaluated. In this paper, we present an extensive evaluation of our backpack mobile mapping system in indoor environments. It is shown that the system can deal with the normal human walking motion, but has problems with irregular jittering. Moreover, we demonstrate the applicability of the backpack in a suitable urban scenario.

We present an algorithm for estimating the six degrees of freedom 6DOF pose for a rigid object of arbitrary shape, which can move fast in cluttered environments. The RGBD data input is obtained by fusing a time-of-flight TOF camera and a color camera. The proposed approach is composed of a coarse estimation stage for prealignment and an accurate estimation stage for refining the coarse pose. The most important contribution is the textured iterative closest point ICP in the accurate stage, where the Lukas-Kanade method is incorporated into the point-to-plane ICP framework, by which geometrically symmetric objects can be handled. In addition, the pose estimation performance under motion artifacts that are common for a TOF sensor can be significantly improved. Another important contribution is the tailored sparse representation under an annealed particle filtering framework for effectively extracting the target from the background clutter and providing a coarse 6DOF pose. The entire algorithm is implemented with graphics processing unit acceleration and shows real-time performance. The approach is verified on a variety of targets in both indoor and outdoor scenarios.

This work presents the RRT-planning algorithm adapted to a mobile car-like robot. Based on an short overview of the RRT-approach, all necessary components, adapted to target vehicle, are shown. A possible state space model, including a valid metric, is presented, as well as a control input selection approach. In order to avoid undesired circular motion, we introduced a circle limitation for the control input selection algorithm. Several simulation results are presented for different environments. We show the limitation and propose a solution implementing this algorithm on real hardware. Finally a conclusion is drawn, summarizing the results and giving an outlook to future work.

The well-known potential field method for obstacle avoidance in the scope of mobile robots is discussed in this paper. Particular attention is on the car-like mobile robots, which impose practical limitations on the application of potential field method due to its limited speed and curvature in motion. Along with the review of some recent studies on this topic, we point out the necessity of implementing a nonholonomic motion planner and propose some extensions to other potential-field-related methods to deal with the constraints of car-like robots. Two exemplary scenarios based on our extensions are simuated to prove their feasibility in application.

This paper presents a narrow passage assistance function to support older adults with degenerated mental and physical abilities. The assistance function is implemented on a mobility scooter which already provides several assistance functions. The narrow passage algorithm is the newest assistance function on the mobility scooter and reliefs the operator from the challenging control of the mobility scooter in narrows, such as doorways, while the operator is controlling the vehicle. This is a very important aspect, because the elderly people do not want to be controlled by the vehicle, they want to control the vehicle by themselves. Therefore, the algorithm determines the prospective path of the vehicle according to the current steering angle. A Laser Range Finder (LRF) is used to detect narrows on the prospective path of the vehicle. If the vehicle can drive through the narrow, this assistance function prevents a slow down caused by the collision avoidance function. The algorithm is evaluated in several test runs using a high precision positioning system. Finally, the results of the experiments are presented and discussed.

In this work, I examine the problem of understanding American football in video. In particular, I present several mid-level computer vision algorithms that each accomplish a different sub-task within a larger system for annotating, interpreting, and analyzing collections of American football video. The analysis of football video is useful in its own right, as teams at all levels from high school to professional football currently spend thousands of dollars and countless human work hours processing video of their own play and the play of their opponents with the aim of developing strategy and improving performance. However, because football is an extremely challenging visual domain, with difficulties ranging from the chaotic motion and identical appearance of the players to the visual clutter on the field in the form of logos and other markings, computer vision algorithms developed towards the end goal of understanding American football are broadly applicable across a variety of visual problems.I address four specific football-related problems in this thesis. First, I describe an approach for registering video with a static model (i.e. the football field in the American football domain) using a novel concept of locally distinctive invariant image feature matches. I also introduce a novel empirical registration transform stability test, which we use to initialize our registration procedure.Second, I outline a novel method for constructing mosaics from collections of video. This method takes a greedy utility maximization approach to build mosaics that achieve user-definable mosaic quality objectives. While broadly applicable, our mosaicing approach accomplishes several tasks specifically relevant to the analysis of football video, including automatically constructing reference image sets for our video registration procedure and for computing background models for initial formation recognition and player tracking algorithms. Third, I present an approach for recognizing initial player formations. This approach, called the Mixture-of-Parts Pictorial Structure (MoPPS) model, extends classical pictorial structures to recognize multi-part objects whose parts can vary in both type and location and for which an object part’s location can depend on its type. While this model is effective in the American football domain, it is also broadly applicable.Finally, I address the problem of tracking football players through video using a novel particle filtering formulation and an associated discriminative training procedure that directly maximizes filter performance based on observed errors during tracking. This particle filtering framework and training procedure are also broadly applicable.For each of these algorithms, I also present a series of detailed experiments demonstrating the method’s effectiveness in the American football domain. As a further contribution, I have made the data sets from most of these experiments publicly available.

This work gives an overview of the project SiNafaR, safe navigation for autonomous robot systems, which aims at identifying requirements, legal aspects in Germany, and finding technological solutions for heterogeneous multi robotic systems in the civil sector. Based on a security services scenario a surveillance system consisting of UAVs and UGVs is proposed. Main technological gaps, such as assessable localization, reliable communication and consistent HMI are identified and solutions are presented.

Automatic transfer of learned knowledge from one task or domain to another offers great potential to simplify and expedite the construction and deployment of intelligent systems. In practice however, there are many barriers to achieving this goal. In this article, we present a prototype system for the real-world context of transferring knowledge of American football from video observation to control in a game simulator. We trace an example play from the raw video through execution and adaptation in the simulator, highlighting the system's component algorithms along with issues of complexity, generality, and scale. We then conclude with a discussion of the implications of this work for other applications, along with several possible improvements.

This paper presents the calibration procedure of a car-like mobile robot with a high-precision positioning system. The mobile robot, which was calibrated, is a small outdoor vehicle designed for teleoperated missions. The aim of the calibration is the achievement of an accurate locomotion, therefore the incremental propulsion sensor, the gyroscope, the propulsion actuator and the steering actuator were analyzed with the help of an active optical high-precision positioning system.

A novel system for real-time tumor tracking and motion compensation with a robotic HexaPOD treatment couch is described. The approach is based on continuous tracking of the tumor motion in portal images without implanted fiducial markers, using the therapeutic megavoltage beam, and tracking of abdominal breathing motion with optical markers. Based on the two independently acquired data sets the table movements for motion compensation are calculated. The principle of operation of the entire prototype system is detailed first. In the second part the performance of the HexaPOD couch was investigated with a robotic four-dimensional-phantom capable of simulating real patient tumor trajectories in three-dimensional space. The performance and limitations of the HexaPOD table and the control system were characterized in terms of its dynamic behavior. The maximum speed and acceleration of the HexaPOD were 8 mm/s and 34.5 mm/s2 in the lateral direction, and 9.5 mm/s and 29.5 mm/s2 in longitudinal and anterior-posterior direction, respectively. Base line drifts of the mean tumor position of realistic lung tumor trajectories could be fully compensated. For continuous tumor tracking and motion compensation a reduction of tumor motion up to 68% of the original amplitude was achieved. In conclusion, this study demonstrated that it is technically feasible to compensate breathing induced tumor motion in the lung with the adaptive tumor tracking system.

The task of registering video frames with a static model is a common problem in many computer vision domains. The standard approach to registration involves finding point correspondences between the video and the model and using those correspondences to numerically determine registration transforms. Current methods locate video-to-model point correspondences by assembling a set of reference images to represent the model and then detecting and matching invariant local image features between the video frames and the set of reference images. These methods work well when all video frames can be guaranteed to contain a sufficient number of distinctive visual features. However, as we demonstrate, these methods are prone to severe misregistration errors in domains where many video frames lack distinctive image features. To overcome these errors, we introduce a concept of local distinctiveness which allows us to find model matches for nearly all video features, regardless of their distinctiveness on a global scale. We present results from the American football domain-where many video frames lack distinctive image features-which show a drastic improvement in registration accuracy over current methods. In addition, we introduce a simple, empirical stability test that allows our method to be fully automated. Finally, we present a registration dataset from the American football domain we hope can be used as a benchmarking tool for registration methods.