Evolving Behavior Trees for Self-Adaptive Source Searching in Indoor Environments

Created by W.Langdon from gp-bibliography.bib Revision:1.7656

@InProceedings{Ma:2023:ICUS,

• author = "Xinyu Ma and Hai Zhu and Xiaozhou Zhu and Wen Yao",
• booktitle = "2023 IEEE International Conference on Unmanned Systems (ICUS)",
• title = "Evolving Behavior Trees for Self-Adaptive Source Searching in Indoor Environments",
• year = "2023",
• pages = "1003--1008",
• abstract = "This paper presents a self-adaptive source searching approach using evolved Behaviour Trees (BTs). Robots often face challenges in making comprehensive decisions during source searching due to difficulties in integrating sensor information and adjusting policies based on varying states and events. Behaviour Trees offer a modular framework for combining perceptions with corresponding actions, enabling robots to make informed decisions across different source searching stages. The proposed approach incorporates strategies from reactive methods and Infotaxis into BT nodes, with the source searching policies evolved through Genetic Programming. To enhance evolution, we introduce a substitution-based synthesis method for initial BTs and a parallel execution approach for fitness evaluation. Experimental results demonstrate that the proposed method outperforms Infotaxis with a 21.8percent increase in success rate and a 35.9percent decrease in average steps, and outperforms the reactive strategies with a 77.7percent increase in success rate and a 46.3percent decrease in average steps. While our method may exhibit lower performance compared to standard Infotaxis in certain wind conditions, it effectively reduces computational complexity without significant performance degradation.",
• keywords = "genetic algorithms, genetic programming, Degradation, Solid modelling, Robot sensing systems, Data models, Stability analysis, Behavioural sciences, source searching, Behaviour Trees, Infotaxis",
• DOI = "doi:10.1109/ICUS58632.2023.10318502",
• ISSN = "2771-7372",
• month = oct,
• notes = "Also known as \cite{10318502}",

}

Genetic Programming entries for Xinyu Ma Hai Zhu Xiaozhou Zhu Wen Yao

Citations