3 papers indexed
End-to-end deep reinforcement learning (DRL) for zero-shot object-goal visual navigation remains challenged by the sim-to-real gap, particularly variations in object appearance and restricted camera field-of-view (FoV). This letter proposes a Temporal Difference-Relational Networ…
Yongjian Zhu, Yusen Tao, Feitian Zhang
Miniature hybrid underwater gliders have attracted increasing attention for long-endurance ocean observation and confined-space inspection. Large-range wing reconfiguration offers a promising yet largely unexplored approach for simultaneously enhancing maneuverability and shape a…
Jiyue Tao, Shunheng Xin, Tongsheng Shen, Dexin Zhao, Feitian Zhang
Decentralized multi-robot coordination under partial observability remains challenging, especially in communication-free settings where agents must act solely from local sensor observations. Privileged policy distillation provides a promising approach by transferring knowledge fr…