KIT – The Research University in the Helmholtz Association Institute for Anthropomatics and Robotics (IAR), High Performance Humanoid Technologies (H 2 T) www.kit.edu Humanoid Robotics @ KIT Tamim Asfour http://www.humanoid.kit.edu http://h2t.anthropomatik.kit.edu
10
Embed
Humanoid Robotics @ KIT · KIT –The Research University in the Helmholtz Association Institute for Anthropomatics and Robotics (IAR), High Performance Humanoid Technologies (H2T)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
KIT – The Research University in the Helmholtz Association
Institute for Anthropomatics and Robotics (IAR), High Performance Humanoid Technologies (H2T)
Grasping and manipulationIntegration of vision and haptics to deal with unknown objects Active perception for object segmentation Vision-based localisationMobile manipulation
Learning for human observationMarker-based (and markerless) observation of human actions Learning motion primitives from human demonstration Motion primitives for grasping, walking and whole-body locomotion and manipulation tasks
8x
Humanoids@KIT
Praxis der Forschung in WS 16/17 am H2T
Humanoids@KIT
Two Topics
1. Deep Learning for Object Manipulation
2. Reactive Grasping of Unknown Objects
Humanoids@KIT
RGB-D
Camera
MemoryX
Deep Learning Framework
for Object Detection
1. Deep Learning for Object Manipulation
Deep
Action
Planner
Humanoids@KIT
GoalApplying RNN-based deep learning techniques to allow ARMAR to explore the possible actions that objects suggest
Collecting and annotating data for training the proposed deep architecture
RequirementsKnowledge in computer Vision, robotics, and machine learning
Solid programming skills in C/C++ (or Python)
1. Deep Learning for Object Manipulation
Humanoids@KIT
2. Reactive Grasping of Unknown Objects
TaskDesign a closed-loop reactive grasping controller
Data fusion of vision and haptics
Online grasp execution
MethodsComparison of different data-drivenmethods (SVM, Gaussian Processes, …)
Data collection with ARMAR-III
RequirementsSolid programming skills in C++
Solid background in math
Humanoids@KIT
H2T „special“ requirements for PdF
Candidates must spend at least one day per week in the H2T labs