Junjie Zeng

National University of Defense Technology

Papers

Total Citations

H-Index

About

Junjie Zeng is a robotics researcher specializing in autonomous navigation, deep reinforcement learning (DRL), and motion planning in dynamic, unknown environments. His major contributions center on developing novel hierarchical frameworks that integrate path planning with adaptive motion control to enable non-holonomic robots to navigate safely among moving obstacles. In his highly cited 2019 paper (70 citations), Zeng introduced the MK-A3C algorithm, a memory- and knowledge-based DRL approach that achieves continuous control in complex, unpredictable settings. Building on this, his JPS-IA3C framework (18 citations) combines Jump Point Search with an improved Asynchronous Advantage Actor-Critic architecture, demonstrating that hybrid planning-control strategies outperform individual methods. Zeng has also addressed practical challenges in grid-based distance map updating, proposing speed optimization techniques for incremental updates that enhance computational efficiency in robotics and game AI. His work bridges the gap between theoretical DRL advances and real-world robotic deployment, with applications ranging from autonomous vehicles to service robots. Zeng’s research is particularly notable for its focus on continuous control and memory-augmented learning, pushing the boundaries of how robots perceive and react to dynamic environments.