Fachbereich Informatik

64-476 Oberseminar Technische Aspekte Multimodaler Systeme

Raum F-334
Zeit Dienstag 16:15 - 17:45 Uhr
Veranstalter Jianwei Zhang


04.04. Vorbesprechung und Terminvergabe

18.04. Department and TAMS IT infrastructure
Introduction for new colleagues -- TAMS internal presentation
Vortragender: Andreas Mäder
Folien tamsIT170418.pdf 7.3 Mi pdf

25.04. Apriltags Docking
Vortragender: Kolja Poreski
Zusammenfassung: The robot this work refers to is a TurtleBot from Kobuki. In this work, an optical method for position determination should be used in order to create a new automated docking algorithm for this TurtleBot. This new docking approach should replace the old one, from Kobuki that still uses Infrared. Another goal is to answer the question wheter optical position determinations are precise enough to be used permanently for autonomous docking. Finally this new approach should be compared to the old one, by using a benchmark designed for docking approaches.
Folien Slides 3.2 M pdf

02.05. Previous research work - future plans
Vortragender: Shuang Li
Zusammenfassung: The previous work mainly focused on the global path planning technology of UGV to efficiently find the shortest collision free path from the start point to target point in two-dimensional static environments by means of proper selection of the environmental modeling methods and further study on two improved genetic algorithms.
Abstract: Slides 1.7 M pdf

09.05. Towards using ROS in the Robocup humanoid league
Vortragender: Marc Bestmann
Abstract: Sharing software modules between teams in the RoboCup Humanoid League is difficult since all teams use different frameworks. This leads to reimplementation of software which slows the research process. A common framework for the league would resolve this. Therefore, this thesis proposes a ROS-based architecture which is defined by a set of ROS messages. The teams can decide on the specific implementation of nodes since the messages provide the interfaces. Different tools, especially for visualization, are implemented to be used in conjunction with this architecture. Furthermore, the robot control module of the team Hamburg Bit-Bots is transferred into the new framework to show its usability. The architecture is compared to others and its performance is evaluated.
The presented architecture makes sharing software modules easier and can thereby accelerate the research in the RoboCup Humanoid League. Furthermore, the entry of new teams is simplified, due to the availability of shared modules.
Folien Slides 4.3 M pdf

16.05. Virtuelle Tastatur
Vortragender: Carolin Konietzny, Paul Bienkowski
Zusammenfassung: In this project we tackle the problem of keyboards being constraining and not ergonomic. We set out to design a system for recording hand and finger movements and the corresponding keyboard input and feed those into a machine learning algorithm. We use IMUs as sensors for capturing the movement and different types of artificial neural networks for the machine learning part. Our work includes the design of a preprocessing algorithm that extracts relevant information from the raw sensor data. We found Convolutional Neural Networks (CNNs) to be suitable for this kind of task and suggest a working configuration for those. We were able to achieve good accuracy when typing a few keys slowly. We show some experiments on the effectiveness of our approach and discuss their results.
Folien Slides 6.2 M pdf

13.06. Haptic recognition based on deep learning
Vortragender: Dong Han
Zusammenfassung: In order to improve the abilities of unknown environment perception and flexible manipulation for robots, this work combines the multi-modal haptic signals from BioTac into haptic images, and builds a multi-class and multi-label haptic image recognition model based on CNN. This model can extract four haptic features, namely hardness, thermal conductivity, roughness, and texture, and recognize objects.
Abstract: Slides 1.6 M pdf

30.05. Bio-IK
Vortragender: Philipp Ruppel
Zusammenfassung: The talk presents a ROS version of the memetic BioIK algorithm for multi end-effector inverse kinematics, implemented as a kinematics solver plugin compatible with Moveit. The algorithm was tested on a variety of robots, including the PR2 service robot, Kuka LWR-4+ with Shadow Dexterous Hand, and the Valkyrie humanoid. Due to careful optimizations, the refined algorithm is shown to outperform classical IK solvers (including trak-ik) on most tests both in terms of success rates and performance.
Folien Slides 6.3 M pdf

06.06. Pfingstferien

13.06. Interactive Reinforcement Learning for HRI
Vortragender: Zhen Deng
Zusammenfassung: This talk is about interaction reinforcement learning for HRI. Two important problems will be addressed. That is how to improve the sample-efficiency of RL and how to learn from human teaching.
Folien Slides 1.0 M pdf

20.06. Adaptive Slicing for the FDM Process Revisited
Vortragender: Florens Wasserfall
Zusammenfassung: Adaptively computing the layer heights for 3D-printed parts has the potential to achieve high quality results while maintaining a reasonably short printing time. The basic concept, several error measures and variations of the algorithm have been around in the literature for two decades now, but never showed significant impact on widely used slicing software. Users of our early test implementations reported two major drawbacks of the existing approaches: the control measures are not intuitively usable and the resulting height distribution in many cases is not optimal for an object, requiring extensive post-editing. In this talk, I introduce a more intuitive control measure and implementation based on the volumetric surface error and a subsequent manual refinement of layer heights by manipulating an interpolated height-curve. All implementations are available as ready-to-use open source software.
Folien Slides 3.3 M pdf

27.06. Fuzzy Semantics Understanding for natural HRI
Vortragender: Jinpeng Mi
Folien Slides 4.4 M pdf

04.07. Capturing and tracking motion with PhaseSpace
Vortragender: Eugen Richter
Zusammenfassung: Things you need to know in order to get started with the PhaseSpace Impulse X2E motion capture/tracking system. You will be introduced to the general workflow, the available software tools and the benefits/constraints of the system.
Folien Slides 5.0 M pdf

11.07. Grasping with the Shadow Hand
Vortragender: Yannick Jonetzko
Folien Slides 18.0 M pdf