Behavior trees (BTs) have gained recognition for their modularity and scalability, establishing them as a robust framework for task automation and planning in various domains, including robotics, game AI, and autonomous systems. This paper presents a comprehensive review of the integration of BTs with reinforcement learning (RL) and learning-from-demonstration (LfD) to enhance decision making and task planning in robotic systems. The review emphasizes the advantages of BTs, such as increased adaptability and efficiency through RL integration and the simplification of robot programming via LfD. Despite these benefits, challenges persist in the areas of computational complexity, scalability in multi-agent systems, and the automatic generation of BTs. This paper concludes by identifying key areas for future research to address these challenges and further advance the development of autonomous robotic systems.
This work was supported by the Technology Innovation Program (1415187715, Development of AI learning platform for intelligent excavators based on expert work data) funded By the Ministry of Trade, Industry & Energy(MOTIE, Korea)
