MIN-Fakultät
Fachbereich Informatik
TAMS

64-476 Oberseminar Technische Aspekte Multimodaler Systeme

Raum F-334
Zeit Dienstag 16 - 18 Uhr (c.t.)
Veranstalter Jianwei Zhang

Termine

18.10. Vorbesprechung

25.10. Hardwarebeschleunigte Berechnung von SURF Features
Vortragender: Andre Walter

01.11. ???
Vortragender: ???

08.11. Introduction to the Git Version Control System
Vortragender: Sebastian Rockel
Folien: GitIntroduction.pdf (English, 0.3 MB)

15.11. ???
Vortragender: Jianhua zhang

22.11. A Multi-camera Optical Tracking system and EKF-based Marker Tracking Method for Spine Surgical Robot
Vortragender: Bo Sun

29.11. ???
Vortragender: ???

06.12. ???
Vortragender: ???

13.12. ???
Vortragender: ???

20.12. fällt aus
Vortragender: ???

24.12. -- 08.01. Weihnachtsferien

10.01. Hierarchical Plan-Based Robot Control in Open-Ended Environments
Vortragender: Dominik Off
Zusammenfassung Autonomous robots need to plan their future course of action for the purpose of performing high-level tasks. Most of the existing AI planning approaches are based on the assumption that a complete plan can be generated prior to executing any action. Nevertheless, real-world robotic domains are inherently dynamic and open-ended. Therefore, often not all necessary information is initially available in order to generate a complete plan in advance. In this talk I present a robot control architecture that is based on a new hierarchical planning approach. The presented architecture is able to automatically switch between planning and acting so that missing information can be acquired by means of active knowledge acquisition. It releases domain engineers from the burden of dealing with the open-endedness of a domain by automatically integrating the execution of appropriate sensing actions. Real-world and simulation-based evaluation results demonstrate that this approach enables a robot to perform tasks even if no knowledge about the dynamic aspects of the environment is available a priori.

17.01. Adaptive Locomotion Control of a Caterpillar-like robot Based on Global Sensory Feedback
Vortragender: Guoyuan Li
Zusammenfassung In this presentation, I will present a framework to realize adaptive locomotion control for a caterpillar-like robot. A novel CPG model inspired from lampreys is employed as the controller of the robot. The advantage is that sensory feedback is easy to integrate into the model based on biological findings in lampreys. As a closed-loop control scheme, first the robot’s touch information is collected and processed, and then according to the analysis of module states, reactive strategies are generated. Finally, by using genetic algorithms, a group of parameters related to sensory input is evolved. A simulation is carried out to verify the feasibility of our approach, which requires the robot to climb over a complicated terrain. The result shows that after evolution, the robot achieves adaptive locomotion and succeeds to climb over the complex terrain.

24.01. In-hand Manipulation Action Gist Modeling and Generalization from Data-glove
Vortragender: Gang Cheng
Zusammenfassung: In-hand manipulation action gist is defined as the key figner motions between two adjacent states, through the guide of action gist sequence, the object is manipulated from the begin state to the end state. A Gaussian Markov Random Field based algorithm is proposed to extract the action gist of each finger, it can effectively decrease the negative impact froms the mentioned issues and provide a concise meta motion sequence. Besides, to manipulate an object, there are multiple methods to handle it, we have to evaluate the popularity of the action gist sequences in the demonstration set. A Meta Motion Occurrence Histogram Matrix is applied to describe the statistic feature of the motion order from all demonstration samples. This presentation is a step by step introduction to the interesting points in the model construction.

31.01. ???
Vortragender: Jian Chen

06.03. Introduction to ROS and the PR2 robot
Vortragender: Sebastian Rockel, Denis Klimentjew
Folien: ROS_PR2_Introduction.pdf (English, 11.3 MB)