Papers
5
Total Citations
34
H-Index
4
About
Houqiang Li is a leading researcher in deep reinforcement learning (DRL) and computer vision, with a focus on bridging the gap between simulated training and real-world deployment. His most impactful work centers on robust policy learning, where he developed **State-Conservative Policy Optimization** to address performance degradation caused by disturbances in transition dynamics—a critical challenge for real-world robotics and autonomous systems. Li also advanced sparse-reward environments through **Soft Hindsight Experience Replay** and **Quantile Regression Hindsight Experience Replay**, enabling more efficient learning in continuous control tasks like robotic arm manipulation. In computer vision, he proposed a **Two-stage Hough Transform** for lane detection systems, demonstrating practical deployment on embedded platforms like the TMS320DM6437. His recent work on **multi-target active object tracking** using Monte Carlo Tree Search (MCTS) and motion modeling has applications in UAVs and intelligent robotics. With over 30 citations across his top papers, Li’s contributions are shaping the reliability and efficiency of DRL algorithms for real-world tasks, making him a notable figure in both theoretical and applied AI research.
Research Focus
Key Achievements
Top Papers
- 1
- 2Soft Hindsight Experience Replay11 citations · 2020
- 3
- 4
- 5Quantile Regression Hindsight Experience Replay3 citations · 2020