In the philosophy of behavior-based robotics, design of complex behavior needs the interaction of basic behaviors that are easily implemented. Action selection mechanism selects the most appropriate behavior among them to achieve goals of robot. Usually, robot might have one or more goals that conflict each other and needs a mechanism to coordinate them. Bayesian network represents the dependencies among variables with directed acyclic graph and infers posterior probability using prior knowledge. This paper proposes a method to improve behavior network, action selection mechanism that uses the graph of behaviors, goals and sensors with activation spreading, using goal inference mechanism of Bayesian network learned automatically. Experimental results on Khepera mobile robot show that the proposed method can generate more appropriate behaviors.
|Title of host publication||AI 2003|
|Subtitle of host publication||Advances in Artificial Intelligence - 16th Australian Conference on AI, Proceedings|
|Editors||Tamas D. Gedeon, Lance Chun Che Fung, Tamas D. Gedeon|
|Number of pages||13|
|Publication status||Published - 2003|
|Event||16th Australian Conference on Artificial Intelligence, AI 2003 - Perth, Australia|
Duration: 2003 Dec 3 → 2003 Dec 5
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Other||16th Australian Conference on Artificial Intelligence, AI 2003|
|Period||03/12/3 → 03/12/5|
Bibliographical notePublisher Copyright:
© Springer-Verlag Berlin Heidelberg 2003.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)