Research

🌟 What I Do

I aim to build intelligent navigation systems that enable robots to operate autonomously in complex, unstructured environments.

🔬 Research Themes

• Foundation Models for Intelligent Decision-Making
  Developing LLM and VLM-based systems for adaptive reasoning and planning. Focusing on prompt engineering, fine-tuning with domain-specific data, and real-time inference optimization for sequential decision-making tasks.

• Deep Reinforcement Learning for Adaptive Control
  Designing model-free and hierarchical RL algorithms for continuous control problems. Investigating policy learning, reward shaping, and sim-to-real transfer methods for deployment in complex, dynamic environments.

• Learning-based Motion Planning
  Creating neural planning methods that leverage learned representations for efficient path generation. Developing learned heuristics, memory-augmented frameworks, and hybrid approaches combining classical and learning-based techniques.

🚀 Future Directions

My future research aims to advance foundation models and reinforcement learning for more complex decision-making scenarios.

Key directions:

• Scaling LLM/VLM reasoning to longer horizons and multi-agent systems
• Sample-efficient RL for high-dimensional continuous control tasks
• Bridging the gap between learned policies and real-world deployment

🤖 Hardware Platforms

I work with diverse robot platforms to validate algorithms in both simulation and real-world environments.

Robot	Image	Type	Use Case
Unitree Go1		Quadruped	Visual–LiDAR fusion, RL locomotion
Unitree Go2		Quadruped	VLM navigation, cross-modal perception
Unitree G1		Humanoid	LLM-guided policy learning
Clearpath Jackal		Wheeled UGV	Real-world navigation testing
Clearpath Husky		Wheeled UGV	Outdoor mapping, multi-sensor fusion

🛠️ Simulation Environments

I design and use multiple simulation platforms for both classical planning and learning-based navigation.

Environment	Image	Description
Gazebo		Classic ROS-based simulator for wheeled robots, supporting costmaps, sensor fusion, and realistic physics.
Isaac Gym		GPU-accelerated simulation for large-scale reinforcement learning and policy optimization.
Isaac Sim		High-fidelity NVIDIA Omniverse simulator for perception, dynamics, and multi-robot coordination.
Custom Terrains		Procedurally generated terrains for testing locomotion, stability, and adaptive control.
Self-design Simulation		self-design simulation using C++ for motion planning or multi-goal motion planning.