发现信息
- 作者:Wenting Zeng, Xiuhuai Xie, Xianyi Yan, Zhengquan Li, Yufei Wang, GuoFeng LUO, Cheng Wang, Longbiao Chen
- 年份:2025
- 发表源:IEEE Smart World Congress (SWC)
- OpenAlex ID:https://openalex.org/W7130688764
- DOI:https://doi.org/10.1109/swc65939.2025.00200
- arXiv:待核验
- 检索主题:self-evolving
- OpenAlex 引用数快照:0(2026-11-12)
摘要线索
In dynamic and complex campus environments, robot agents face challenges in providing efficient and accurate services due to their limited ability to learn and adapt to unforeseen situations. To address these issues, we propose Pandora, a large language model (LLM)-driven robot agent framework that enhances task execution through a structured approach of planning collection, insight extraction, and evolutionary inference. Specifically, the agent systematically collects and stores task planning using a task planning tree, extracts high-level insights to guide future actions, and integrates these insights and successful planning with supervised fine-tuning to achieve self-evolution. Evaluations on a digital twin platform and benchmark datasets demonstrate that Pandora significantly improves task success rate and task efficiency, offering a robust solution to develop adaptable and intelligent robot agents in campus settings.
核验与筛选
- 当前状态:待筛选、去重和一手来源核验。
- OpenAlex 元数据和引用数只用于发现,不作为正式引文或质量判断。
- 记录正式版本、venue、方法细节、实验协议和局限后,才迁入稳定论文卡片。