Sonderforschungsbereich 331: Information Processing in Autonomous, Mobile Systems

Prof. Dr. Bernd Radig, Ordinarius
Tel.: +49-89-48095-121
Fax.: -120
Email: radig@informatik.tu.muenchen.de

Prof. Dr.-Ing. Hans-Jürgen Siegert, Ordinarius
Tel.: +49-89-48095-251
Fax.: -250
Email: siegert@informatik.tu.muenchen.de

Compared to today's stationary industry robots, autonomous mobile manipulation systems will have the capability to move and plan autonomously and thus are available to handle new kinds of tasks at varying places. Thereby, the features "mobility" and "autonomy" change the capabilities and the complexity of the manipulation systems decisively. The aim of the joint research project is the provision of the scientific and technical prerequisites needed to implement autonomous mobile manipulation systems and the evaluation of the prototypic solutions in demonstration scenarios. Therefore, information processing plays a central role - it is the principal subject of the SFB 331. Increasing the autonomy in respect to information and action increases the capability of mobile manipulation systems to find their way in changing environments even in cases of disturbances. This is achieved by using previously collected knowledge and sensor data. These capabilities use basic services that are typical for autonomous, mobile systems:

Platforms, that can move freely (locomotion) and that can solve manipulation tasks (manipulation) serve as a technical basis. Additionally, adequate sensors and the infrastructure required for information processing (computers, communication mechanisms) are used.
The systems are equipped with several sensors. Thus the integration of multi-sensor information and its model-based interpretation is important, for example in order to recognize disturbance situations.
The required knowledge for the autonomy ( e. g. environmental models, behavior rules, planning algorithms) are stored in the distributed, active and team-based knowledge base.
The autonomous planning, execution and supervision of the actions is done just in time, in order to be able to react on disturbance situations.
The communication to superior instances and the cooperation with other autonomous systems take place at the highest possible abstraction level (e. g. distribution of tasks, notification about disturbance situations). For this purpose new concepts are used, for example negotiation protocols for task distribution.

The joint research project SFB 331 started in 1986 and will run out at the end of 1997. In this last phase the integration of locomotion and manipulation, resulting in "mobile manipulation", especially in the area of service robots and the support of the cooperation of several mobile manipulation systems, has been attained. Thematically the various subprojects of the SFB 331 can be grouped as follows:

Global planning and coordination, communication and intelligent cooperation (Q1, Q6).
(Re-)planning of manipulation tasks, especially in fault situations, based on automatic simulation (M3).
Model-based interpretation and use of sensor information (L4, L9, Q5).
Application-oriented subprojects and provision of the platforms for the integration work (L7, L8, M1, M2).

The chairs of electrical engineering and information processing, mechanical engineering and computer science take part in this joint research project. The department of computer science works on two subprojects:

Subproject L9 Vision-based Object Recognition for Autonomous Mobile Systems (Prof. Dr. B. Radig): The aim of our research is to support autonomous mobile systems during navigation and manipulation tasks. This is done by a model-based interpretation of single perspective video images. The main topics of our work are:

Development of algorithms for the identification and localization of relevant 3D objects. These include articulated objects, that is objects with movable components. Based on these methods an AMS is e.g., able to pick up workpieces, dispose garbage, or open doors or drawers.
Utilization of all the available model information provided by an environmental model (see Q5). On the other side, the information gained by the image interpretation is fed back to the model.
Investigation of feature-based 3D reconstruction for exploration tasks.

Subproject Q6 Distributed Knowledge Base: Problem Oriented Services and Intelligent Cooperation (Prof. Dr. H.-J. Siegert): In this subproject a distributed real-time knowledge base is developed. This knowledge base is used in the SFB as an universal medium for data representation and cooperation. The knowledge base has been enlarged with problem oriented services and methods for intelligent cooperation. Problem oriented services use the basic interface of the distributed and team-oriented knowledge base. For example they allow the modeling of search-queries. The methods for intelligent cooperation support agreements between the autonomous systems. This includes task negotiations between the autonomous systems using contract-net-oriented protocols. It also contains support of data-structures and methods for cooperative task solution. Examples for cooperative tasks are handing-over tasks between autonomous mobile systems. In the demonstration scenarios of the SFB a product version of the knowledge base has been provided for the different computer platforms of the SFB. Further tasks of the subproject are supporting the integration of the knowledge base interfaces into the application programs of the different subprojects, coordinating the different requirements of the heterogeneous applications and supporting the conversion of the requirements into efficient knowledge-base-oriented communication mechanisms.

Back to contents