Autonomous Systems: A Cognitive-Oriented Approach Applied to Mobile Robotics
1st Edition
Published in August 2007
Book
Paperback/Softback
143 pages
978-3-8322-6497-0 (ISBN)
Description
Many problems in control - or more general in automation science - cannot be described by mathematical equations such that classical control approaches are not feasible. Humans have the amazing capability to cope with dynamic environments, to make goal-oriented decisions based on incomplete and incorrect knowledge, and to change their interacting behavior situation-dependent. The cognitive architecture presented in this thesis controls the interacting behavior of an autonomous system with its environment. The motivation of this thesis is to realize a cognitive-oriented controller which can be applied to a new range of problems in automation science, in particular problems where classical control approaches cannot cope with the complexity of the interaction and mathematical modeling is not feasible. This thesis describes the methodology to tackle this problem.
Recognizing the lack of true flexibility and intelligence in today's autonomous systems, the objective of this thesis is to realize a cognitive-oriented control approach with application to a mobile robot. To this end, the learning, planning, and execution processes are integrated in one architecture to build a cognitive technical system implemented on a mobile robot. Based on the information learned by interaction with the environment, plans to achieve different goals are built and executed by the agent. One important feature is the ability to change the goal without changing the knowledge base or any other part of the architecture. The main issue is the development of an architecture feasible to be implemented on a technical system and to detail the methodology. Consequently, the functionality of the autonomous learning system has to be tested in a real world environment and not just in a simulated one.
In [Söf03b], a situation-operator modeling technique is developed to formalize and model the human-machine interaction. The goal of this thesis is to use that method to design an autonomous system which is capable of learning by interacting with its environment. The initiating idea to build autonomous systems based on the situation-operator-model at the representation level, as outlined in [Söf01a], is detailed and technically realized for the first time. As mentioned before, the validation of the proposed approach to build a cognitive technical system is an important part of this thesis.
Recognizing the lack of true flexibility and intelligence in today's autonomous systems, the objective of this thesis is to realize a cognitive-oriented control approach with application to a mobile robot. To this end, the learning, planning, and execution processes are integrated in one architecture to build a cognitive technical system implemented on a mobile robot. Based on the information learned by interaction with the environment, plans to achieve different goals are built and executed by the agent. One important feature is the ability to change the goal without changing the knowledge base or any other part of the architecture. The main issue is the development of an architecture feasible to be implemented on a technical system and to detail the methodology. Consequently, the functionality of the autonomous learning system has to be tested in a real world environment and not just in a simulated one.
In [Söf03b], a situation-operator modeling technique is developed to formalize and model the human-machine interaction. The goal of this thesis is to use that method to design an autonomous system which is capable of learning by interacting with its environment. The initiating idea to build autonomous systems based on the situation-operator-model at the representation level, as outlined in [Söf01a], is detailed and technically realized for the first time. As mentioned before, the validation of the proposed approach to build a cognitive technical system is an important part of this thesis.