Authors:
Matthias Rungger
1
;
Olaf Stursberg
1
;
Bernd Spanfelner
2
;
Christian Leuxner
2
and
Wassiou Sitou
2
Affiliations:
1
Institute of Automatic Control Engineering, Technische Universität München, Germany
;
2
Software and System Engineering, Technische Universität München, Germany
Keyword(s):
Hierarchical Planning, Autonomous Robots, Context Adaptation, Hybrid Models, Predictive Control.
Related
Ontology
Subjects/Areas/Topics:
Autonomous Agents
;
Informatics in Control, Automation and Robotics
;
Mobile Robots and Autonomous Systems
;
Reasoning about Action for Intelligent Robots
;
Robotics and Automation
Abstract:
This paper considers the behavior planning of robots deployed to act autonomously in highly dynamic environments. For such environments and complex tasks, model-based planning requires relatively complex world models to capture all relevant dependencies. The efficient generation of decisions, such that realtime requirements are met, has to be based on suitable means to handle complexity. This paper proposes a hierarchical architecture to vertically decompose the decision space. The layers of the architecture comprise methods for adaptation, action planning, and control, where each method operates on appropriately detailed models of the robot and its environment. The approach is illustrated for the example of robotic motion planning.