Repeated Deceptive Path Planning against Learnable Observer

We introduce Repeated Deceptive Path Planning (RDPP), a novel settingwhereanagentmustconcealitsdestinationfromalearnable observer that can adapt from historical trajectories. We show that existing deceptive planning methods, designed for static observers, fail in RDPP due to accumulated adaptation lag. To address this, we propose Deceptive Meta Planning (DeMP), a two-level optimization framework that anticipates and counteracts observer updates across episodes via meta-level learning. Experiments demonstrate that DeMP significantly outperforms traditional methods, enabling sustained deception against learning adversaries while maintaining efficient path costs.

Authors

• Shiyue Cao School of Artificial Intelligence, University of Chinese Academy of Sciences & Institute of Automation, Chinese Academy of Sciences
• Pei Xu Institute of Automation, Chinese Academy of Sciences
• Likun Yang University of Chinese Academy of Sciences
• Lei Cui University of Chinese Academy of Sciences
• Shizhao Yu University of Chinese Academy of Sciences
• Shiyu Zhang Institute of Automation, Chinese Academy of Sciences
• Yongjian Ren University of Chinese Academy of Sciences
• Xiaotang Chen Institute of Automation, Chinese Academy of Sciences
• Kaiqi Huang School of Artificial Intelligence, University of Chinese Academy of Sciences & Institute of Automation, Chinese Academy of Sciences

Keywords

• Deceptive Path Planning
• Goal Recognition
• Reinforcement Learning