Localization and Mapping are the mandatory functionalities of an autonomous mobile robot. These functionalities are implemented mostly using expensive laser range sensors which are characterized by long range, wide opening angle and precision. Recent developments in low-cost RGBD cameras make them suitable for robot mapping and localization. However, limitations like viewing angle, range and accuracy prevent them from replacing laser rangers in large areas. With such RGBD cameras, mapping large areas is hard but localization in large areas can be achieved by augmenting them with additional information. This makes the low-cost RGBD cameras suitable for a cost-effective localization solution, especially for the fleets of robots operating in large areas, where one robot equipped with a laser ranger is used in mapping while the other robots equipped with low-cost RGBD cameras for localization. Adaptive Monte Carlo Localization(AMCL) is the most widely used algorithm for mobile robot localization. Our proposed solution extends the AMCL to use a virtual laser generated from RGBD camera and visual markers to solve the issues concerning the global localization, reducing the uncertainty in pose estimates during tracking and kidnapped robot problem. Our approach is tested in large areas and the uncertainty during the localization is compared against the result from using a laser ranger.

Coronary heart disease is a common cause of death despite being preventable. To treat the underlying plaque deposits in the arterial walls, intravascular optical coherence tomography can be used by experts to detect and characterize the lesions. In clinical routine, hundreds of images are acquired for each patient, which require automatic plaque detection for fast and accurate decision support. So far, automatic approaches rely on classic machine learning methods and deep learning solutions have rarely been studied. Given the success of deep learning methods with other imaging modalities, a thorough understanding of deep learning-based plaque detection for future clinical decision support systems is required. We address this issue with a new data set consisting of in vivo patient images labeled by three trained experts. Using this data set, we employ the state-of-the-art deep learning models that directly learn plaque classification from the images. For improved performance, we study different transfer learning approaches. Furthermore, we investigate the use of Cartesian and polar image representations and employ data augmentation techniques tailored to each representation. We fuse both representations in a multi-path architecture for more effective feature exploitation. Last, we address the challenge of plaque differentiation in addition to detection. Overall, we find that our combined model performs best with an accuracy of 91.7%, a sensitivity of 90.9%, and a specificity of 92.4%. Our results indicate that building a deep learning-based clinical decision support system for plaque detection is feasible.

The high system flexibility necessary for the full automation of complex and unstructured tasks leads to increased technological complexity, thus to higher costs and lower performance. In this paper, after an introduction to the different dimensions of flexibility, a method for flexible modular configuration and evaluation of systems of systems is introduced. The method starts from process requirements and, considering factors such as feasibility, development costs, market potential and effective impact on the current processes, enables the evaluation of a flexible systems of systems equipped with the needed functionalities before its actual development. This allows setting the focus on those aspects of flexibility that add market value to the system, thus promoting the efficient development of systems addressed to interested customers in intralogistics. An example of application of the method is given and discussed.

Fast re-organization of a production facility and its production flows enables the automated and efficient production of individualized products up to a lot size of one single instance in large-scale. Flexible service robots replacing hard wired transportation belts are one of the enablers of the Industry 4.0 idea. Within this paper we propose a distributed approach for autonomous robots forming a fleet capable of performing transportation tasks within an industrial production system. The system is able to adapt to changes in the production flow by changing the software configuration only. The presented approach combines the predictability and efficiency of path based navigation with the flexibility of fast reactive local obstacle avoidance. The combination enables reliable, predictable and robust navigation of many robots in a fleet, that are able to cope with the challenges of a dynamic environment such as obstacles blocking the way and persons moving around.

Reliable object recognition is a mandatory prerequisite for Service Robots in everyday environments. Typical approaches for object recognition use single algorithms or features. However, none is yet able to classify across all types of objects and the field of object recognition is thus still an open challenge. We propose an approach for object recognition and pose estimation that combines existing algorithms. Probabilistic methods are used to fuse the classification and pose estimation results, considering the error introduced by the measurements, actuators (sensor on manipulator) and algorithms. Since integration is one of the real challenges from the laboratory towards the real world, we demonstrate the approach in two fully integrated scenarios. We run the experiments on two platforms and focus on the distinction of few but similar objects.