原作者 Tingde Liu

本站由 GaryLee1210 维护，完整保留原站的研究内容供学习阅读。下面是原作者 Tingde Liu 的个人介绍，其经历与项目属于原作者。查看原站介绍。

I am a robotics and AI engineer with an M.Sc. in Mechatronics and Robotics from Leibniz Universität Hannover (LUH), now based in Beijing and working on embodied intelligence. My current research centers on Vision-Language Navigation (VLN), Embodied Agent Frameworks, and 3D Large Language Models (3DLLM) — and the broader question of what it would actually take for a robot to understand and act in the world the way we do. I am interested in genuine embodied intelligence, not just systems that appear to navigate, but ones that truly reason about space, language, and intention.

Experience

  • AI Engineer, Embodied Intelligence — Beijing, China (2025 – present). I work at a robotics company focused on embodied navigation. My main responsibility is the embodied navigation system framework — the harness that binds perception, memory, planning, and control into a stack a real robot can run — with Vision-Language Navigation (VLN) and agentic navigation as my research directions.
  • Research Assistant — IPH – Institut für Integrierte Produktion Hannover. Deploying robots in real industrial environments — navigation, 3D reconstruction, semantic mapping — where the gap between a working demo and a system that runs reliably on a factory floor turned out to be enormous.
  • Research Associate — Institut für Kartographie und Geoinformatik (IKG), LUH. Nearly two years on spatial intelligence, developing multimodal large language models for point clouds — teaching them to read urban LiDAR as it actually comes: sparse, noisy, colorless.

About This Blog

This blog is my working space for collecting, organizing, and sharing the ideas that shape my research: survey posts, paper notes, and long-form reflections on embodied AI. I keep it public to make my learning process visible and reusable — a place to track what I read, what I build, and how my understanding changes over time, instead of letting those notes scatter across documents and bookmarks. The goal is simple: turn ongoing research into something structured, searchable, and worth revisiting.

I also want it to be a friendly open-source space where people feel welcome to read, contribute, discuss, and help improve the ideas here together.

Research Interests

These experiences converged around a set of questions I keep returning to:

  • How can language models reason meaningfully about 3D space?
  • What does it take for a robot to navigate using natural instructions?
  • How do we bridge the gap between simulation and real-world perception?
  • How do we build a general-purpose agentic navigation framework — one that integrates perception, memory, planning, and action well enough to carry a robot through very different long-horizon navigation tasks?
  • How can agentic AI give robots something closer to genuine agency: not just executing instructions, but forming intentions, adapting plans, and acting with purpose?

In practice this means I work on Vision-Language Navigation (VLN), Embodied Agent Frameworks, 3DLLM, and the infrastructure that makes embodied intelligence real.

Technical Skills

Languages: Python, C++, MATLAB Frameworks & Systems: PyTorch, ROS2, LangChain, CUDA Vision-Language & Multimodal: CLIP, BLIP, LLaVA, Vision-Language Modeling, Multimodal LLMs 3D Vision & Perception: LiDAR processing, PCL, 3D Gaussian Splatting, YOLO Robotics: Robot Perception, Motion Planning, INS, SLAM, Imitation Learning, Reinforcement Learning, Policy Learning Simulation: Gazebo, Isaac Sim, Habitat Tools: Claude Code, Openclaw, Docker, Git

Continuously learning and exploring the infinite possibilities of AI and Robotics!