Yeseong Park

About

Yeseong Park is a robotics researcher specializing in deep reinforcement learning for visual grasping and manipulation. His work focuses on overcoming the fundamental limitations of actor-critic deep RL methods when applied to robotic grasping tasks, particularly in cluttered environments with diverse, unseen objects. Park’s major contribution lies in developing state representation learning techniques that accelerate actor-critic deep reinforcement learning by preprocessing and disentangling raw visual input. His most cited paper (2021, 14 citations) demonstrates how state representation learning from preprocessed images significantly improves the performance and speed of actor-critic methods for visual grasping. His follow-up work (2020, 7 citations) extends this approach to cluttered scenes by disentangling raw input images, enabling more robust grasping of novel objects. Park also developed a real-world platform for actor-critic deep RL-based robotic grasping (2020, 3 citations), bridging simulation and physical deployment. His research addresses a critical bottleneck in applying RL to robotics—poor performance with diverse objects—by making learned representations more efficient and generalizable. Park’s work has practical implications for industrial automation, warehouse robotics, and assistive manipulation, offering a pathway to more adaptive and faster-learning robotic systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Acceleration of Actor-Critic Deep Reinforcement Learning for Visual Grasping by State Representation Learning Based on a Preprocessed Input Image

   14 citations · 2021
2. 2
   Acceleration of Actor-Critic Deep Reinforcement Learning for Visual Grasping in Clutter by State Representation Learning Based on Disentanglement of a Raw Input Image

   7 citations · 2020
3. 3
   Development of an Actor-Critic Deep Reinforcement Learning Platform for Robotic Grasping in Real World

   3 citations · 2020

Key Collaborators

Contact & Links