Jinyu Zhu

About

Jinyu Zhu is a researcher at the forefront of integrating formal methods with deep reinforcement learning for autonomous robotics. Their work focuses on multi-task motion planning under stochastic environments, where they address the critical challenge of reward sparsity in continuous state spaces. Zhu’s most cited paper, “GR(1)-Guided Deep Reinforcement Learning for Multi-Task Motion Planning under a Stochastic Environment” (2022, 8 citations), introduces a novel framework that leverages Generalized Reactivity (GR(1)) specifications to guide DRL agents, enabling them to learn robust, safe behaviors across multiple tasks without handcrafted reward engineering. This contribution bridges the gap between symbolic planning and learning-based control, offering a principled approach to handling complex, uncertain environments. By combining temporal logic constraints with deep RL, Zhu’s work has significant implications for real-world robotics applications, from autonomous navigation to manipulation. Their research continues to push the boundaries of how robots can reason about and adapt to dynamic, unpredictable settings, making Zhu a promising voice in the intersection of AI, control theory, and robotics.

Research Focus

Key Achievements

Top Papers

1. 1
   GR(1)-Guided Deep Reinforcement Learning for Multi-Task Motion Planning under a Stochastic Environment

   8 citations · 2022

Key Collaborators

Contact & Links