|
|
|
|
|
|
|
Human Robot Interaction
Main Aspects:
Multi-modal user detection and tracking
- Vision-based approaches
- Sonar-/ laser based approaches
- Audio-based approaches
- Probabilistic coding and aggregation techniques
Vision-based identification of user instructions
- Estimation of user's head and eye gaze
- Detection of user's pointing poses
- Identification of command gestures
- Command words, continuous speech
Vision-based recognition of current user state
- Age and gender
- Gaze direction
- Facial expression
- Interest on interaction
- Body language
- Recognition of authorized users
User adaptive multi-modal dialog
- Verbal and non-verbal dialog
- Dialog regimes and coding principles
- Learning efficient Human-Robot dialog strategies from
"Action-Perception Cycle"
- Social interaction with robot companions
|
|
|
|
CompanionAble-Project
There are widely acknowledged imperatives for helping the
elderly live at home (semi)- independently for as long as possible. Without
cognitive stimulation support the elderly dementia and depression sufferers
can deteriorate rapidly and the carers ill face a more demanding task. Both
groups are increasingly at the risk of social exclusion. The FP7 Integrated
Project CompanionAble will provide the synergy of Robotics and Ambient
Intelligence technologies and their integration to provide for a
care-giver's assistive environment. This will support the cognitive
stimulation and therapy management of the care-recipient. This is mediated
by a robotic companion (mobile facilitation) working collaboratively with a
smart home environment (stationary facilitation).
Positive effects of both individual solutions shall be combined to
demonstrate how the synergy between a stationary smart home solution and an
embodied mobile robot companion can make the care and the care-recipient’s
interaction with her assistive system significantly better.
|
 |
SERROKON - SERvice ROboter KONzeption
Based on the results achieved within the previous
projects PERSES and SERROKON-V (supported by TMWFK), the aim is to develop a
flexible, mobile, and interactive service robot that can be reconfigured for
a variety of application domains. Particularly, the arising platform shall
be investigated for two ambituous prototypical applications: a shopping
assistant operating in a home depot store, and a home robot for domestic
tasks (entertainment, edutainment, security). Besides the development of the
robot platform, special emphasis is placed on the accomplishment of field
experiments concerning usability and acceptability within the mentioned
prototypical applications.
|
|
|
|
Robot Navigation
Main Aspects:
Our research activities in robot navigation methods are mainly focussed
on the following problems:
- Probabilistic techniques for multimodal map building and
self-localization
- Simultaneous localization and map building (SLAM)
- Omnivision-based self-localization and SLAM
- Path planning and movement control
- Neural and probabilistic sensor fusion techniques
- Consistent integration of navigation and HRI techniques into overall
control architectures
The following application fields are of particular interest:
- Mobile robotics for large-scale, un-engineered and crowded
environments
- Mobile indoor and outdoor service robotics
- Navigation in context of Human-Robot Interaction (HRI)
- Robust navigation in domestic environments
|
|
|
|
Neural Computing and Machine Learning
Main Aspects:
Our methodological research activities in Neural Computing and Machine
Learning are focused on the following areas:
- Reinforcement learning and active learning within the
"Action-Perception Cycle"
- Learning of visuomotor mappings
- Dynamical organization of behaviors in Multi-Agents Systems (MAS)
- Generative character of perception - as internal simulation of
hypothetical actions and anticipation on their sensory consequences
- Task-relevant control of visual and auditory attention
In this context, the following application fields are of particular
interest:
- Learning efficient Human-Robot dialog strategies from
"Action-Perception Cycle"
- Learning visually guided behaviors for mobile robot navigation and
process control
|
|
|
|
SOFCOM
Selfoptimizing combustion control for CO2 emission reduction in industrial
coal-fired power plants
(SelbstOptimierende Feuerungsführung zur CO2-EmissionsMinderung in
großindustriellen Kohlekraftwerken)
The scope of the SOFCOM project is to develop a
selfoptimizing combustion control suitable for industrial coal-fired power
plants. The primary goal is to increase the effectiveness of the combustion
process to reduce the coal consumption and therefore minimize the production
of greenhouse gases. The realization is focused on intelligent control of
the underlying process. Selforganizing, hence selfoptimizing, techniques are
used to handle the input of spatial-temporal video data, to produce an
optimal control strategy, and to adapt to changes in the combustion process.
|
|
ASOP
Adaptive Self Optimization Process control system
This project contains the following tasks:
- Further development of our model-based anticipative process control
system with respect to the following subtasks:
- Data-mining of visual and conventional process data
- Implementation of appropriate process models using artificial
neural networks
- Design, test und fine-tuning of the process evaluation
- Development of simulated demonstrators for process evaluation and
control
- Development of tools for data-visualization based on neural
clustering algorithms
- Feasibility-analysis for new methods of data-mining to extract
control-relevant visual information out of the video-data of a
combustion process (ICA and nonlinear PCA)
|
|
|
|
| Navigation |
 |
|
 |
| Activities |
The research of the Department of Neuroinformatics and Cognitive Robotics is
focused on the development of behavior-oriented sensorimotor systems, especially
mobile robots, with lifelong learning capabilities to adapt perception and
behavioral control to changing environmental conditions. Behaviors considered
are especially from the areas of
- visually guided mobile robot navigation and
- multi-modal interaction between semi-autonomous agents and human users.
Against this background, the methodological research activities of the
department are focused on the following areas:
- Evolution and organization of navigation- and interaction-behaviors in
multi-modal Human-Robot systems
- Probabilistic and neural techniques for mobile robot navigation,
self-localization, and Human-Robot-Interaction (HRI)
- Vision-based, non-verbal Human-Robot-Interaction by means of gestures,
head and eye gaze, facial expression, and body language
- On-line learning within the "Action-Perception Cycle" (Reinforcement
Learning)
- Life-long learning handling the Stability-Plasticity Dilemma
- Dynamical organization of behaviors in Multi-Agent Systems (MAS)
- Task-relevant visual and auditory attention
- Generative character of perception - as internal simulation of
hypothetical actions and anticipation of their sensory consequences
In the department's research, the following application fields are of
particular interest:
- Autonomous and interaction-based robots in concrete "Real-World"-
applications (Service robots, entertainment and education robots, etc.)
- Vision-based Human-Machine-Interfaces (HMI) for interactive robots
- Robust person detection, tracking and recognition in real-world
video-data streams
- Visually-guided, self-optimizing process control for combustion
processes
The robotics equipment of the department consists of:
- Three mobile robot SCITOS
- One mobile robot PERSES (PERsonalSErvice System - based on a
B21-platform)
- Two mobile robots HOROS 1+2 (HOme RObot System - based on Pioneer
II-platforms)
- One mobile robot MILVA (Multi-sensory Intelligent Learning Vehicle with
Autonomy)
- Seven miniature robots KHEPERA
|
|
|
|
|
K 
