Katie Kang

About

Katie Kang is a leading researcher in robotics and artificial intelligence, specializing in deep reinforcement learning for vision-based autonomous systems. Her work addresses a critical challenge in robotics: enabling models to generalize across diverse environments and platforms despite limited real-world data. Kang’s most influential contribution, her 2019 paper “Generalization through Simulation,” has garnered 114 citations for its novel approach to integrating simulated and real data, dramatically improving the robustness of vision-based control for autonomous flight. This work demonstrates how combining synthetic training environments with limited real-world data can overcome the data scarcity that often hinders deep reinforcement learning in physical robots. More recently, Kang has advanced the field with her 2021 study “Hierarchically Integrated Models,” which tackles the challenge of learning navigation policies from heterogeneous robot platforms. By developing methods to aggregate data from multiple robot types, she reduces the prohibitive cost of single-robot data collection while enhancing policy generalization. Kang’s research is pivotal for scalable, real-world deployment of autonomous systems, offering practical solutions that bridge simulation and reality while maximizing the utility of diverse robotic data sources.

Research Focus

Key Achievements

Top Papers

1. 1
   Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight

   114 citations · 2019
2. 2
   Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight

   11 citations · 2019
3. 3
   Hierarchically Integrated Models: Learning to Navigate from Heterogeneous Robots

   3 citations · 2021

Key Collaborators

Contact & Links