Software Architecture and Integration of Cognitive Systems |
The development of autonomous systems that learn through interaction with humans is an experimental research activity that combines research efforts in mechanics, electronics, informatics and the cognitive sciences. Creating, e.g., cognitive robots requires building systems that can adapt their behavior to environments that are complex, rapidly changing, and that can not be completely modeled in preface. Nowadays, many of the resulting challenges can be successfully addressed on the level of individual algorithms or by advanced robotics hardware tuned for specific scenarios.
In order to provide an avenue for cognitive systems to become useful in every-day human settings and as a prerequisite for entering the consumer market, these attributes must be ensured on an architectural level that encompasses the web of skills embodied cognitive systems must be capable of.
Hence, in the Applied Informatics we investigate
- architectures and engineering principles allowing to efficiently assemble, implement and test systems,
- survey how software engineering is actually performed in the real world, and
- improve existing development processes and tools to address robotics challenges.
Current examples for architectural principles that are investigated are so-called memory architectures that provide memory functionality useful across different tasks and scenarios in order to support learning processes, the transfer of current concepts of event-driven architectures for efficient software integration or the exploitation of autonomic computing concepts for self-regulation in order to increase the autonomy and dependability of cognitive systems.
Eventually, these research efforts lead to the development of advanced software concepts and implementations, e.g., of middleware for cognitive systems or frameworks for computer vision and pattern recognition that are used for the construction of novel cognitive systems. Examples for these efforts that are made available to the public domain are:
These software architectures were developed and evaluated in various collaborative research projects on cognitive systems, e.g., COGNIRON, DESIRE, VAMPIRE, as well as sucessfully applied for the integration of advanced demonstration scenarios, e.g., in the RoboCup@Home mobile robotic system of Bielefeld University or diverse student projects.