Mechanical Engineering undergraduate at Soochow University, building toward mobile robot navigation, embodied intelligence, deep learning, reinforcement learning, world models, and AI-assisted developer workflows.
- I work on robot navigation, multimodal perception, deep reinforcement learning, and embodied intelligence systems.
- I am strongly interested in deep learning and reinforcement learning, and I am now studying network architectures and training methods in depth.
- I am especially fascinated by world models and Dreamer-style model-based RL, where agents learn latent dynamics, imagine future trajectories, discover causal structure, and develop curiosity-driven exploration.
- My current research interests include proactive navigation in dynamic environments, social robot navigation, multi-agent coordination, and learning-based decision making.
- I am learning and experimenting with CNNs, U-Net, RNN/LSTM/GRU models, Transformers, attention mechanisms, GNN-style relational modeling, policy/value networks, PPO, and MAPPO.
- I like projects that close the loop from modeling and simulation to ROS integration, real robot debugging, and experimental analysis.
- Outside research code, I build small tools around developer automation, personal workflows, Linux desktop configuration, and technical writing.
| Direction | What I am doing |
|---|---|
| Proactive robot navigation | Temporal traversability prediction, risk-aware A*, dynamic-door navigation, and topology-aware planning experiments. |
| Social navigation | Interaction-aware SVO modeling, crowd navigation with deep reinforcement learning, and lightweight reproducible simulation environments. |
| Deep learning and RL | Studying neural network structures, representation learning, sequence prediction, attention modules, policy optimization, and multi-agent reinforcement learning. |
| World models | Exploring Dreamer-style agents, predictive latent dynamics, causal representation learning, intrinsic motivation, and curiosity-driven exploration. |
| Robot systems | ROS-based delivery robot platform with SLAM, AMCL, DWA/TEB planning, QR/OCR perception, and manipulator control. |
| Focus | Notes |
|---|---|
| Mainline | Deep learning, reinforcement learning, world models, and robot navigation. |
| Models I am studying | CNN, U-Net, RNN, LSTM, GRU, Transformer, Attention, GNN-style relational models. |
| RL direction | Model-based RL, Dreamer-style agents, PPO/MAPPO, policy/value learning, intrinsic motivation. |
| Questions I care about | How agents build predictive latent states, discover causal structure, and explore with curiosity. |
- Patent work around robot path planning, risk-aware navigation, and mobile robot systems, including first-inventor authorization for a robot path-planning method.
- Student technology association work, technical writing, and long-term robotics competition practice.
More GitHub summary cards
- Blog: saytt.cc
- Email: [email protected]