Ding Ao’s Homepage

Ding Ao

🎓 Master student in Control Science and Engineering at Beijing Institute of Technology 📧 Email: 2390897650@qq.com

An Efficient Policy Gradient Algorithm with Historical Behaviors Reusing in Multi Agent System CSIS-IAC2024, EI, First Author, Accepted Proposed a BOX-COX-based method to reuse historical behaviors in multi-agent reinforcement learning for navigation in dynamic POMDP environments.
DDPG-MPC: A Safe and Efficient Hybrid Path Planning Algorithm for Robots CSIS-IAC2024, EI, First Author, Accepted Designed an adaptive switching mechanism between DDPG and MPC for robot path planning that ensures both exploration and stability.
Multi-Agent Reinforcement Learning System Framework Based On Topological Networks In Fourier Space ASOC (TOP Q1), Second Author, Accepted Developed a multi-agent reinforcement learning framework based on Fourier-space topological networks.

ICRA RoboMaster 2022 AI Challenge – International Second Prize (2022.5) Developed advanced motion planning and control strategies to avoid static and dynamic obstacles while minimizing enemy attacks.
Huawei Cup Graduate Mathematical Contest in Modeling – Second Prize (2023.12)
Miaosuan Cup Human-Machine Hybrid Competition – National Second Prize (2024.6)
MathorCup Mathematical Application Challenge (Graduate Group) – Regional Second Prize (2024.4)

End-to-End Algorithm Intern

Developed a lane-change decision classifier and improved recall using a Ray-based pipeline and tag-enhanced loss.
Designed a centerline-matching strategy to improve anchor-path labeling for lane-change learning, leading to more timely and effective decisions in road tests.

Programming: C++, Python, ROS, Matlab
Algorithms: Deep Reinforcement Learning (DDPG, SAC, PPO), MPC, LQR, PID, SLAM
Familiar with: PyTorch, Transformer architectures, mmdet3D, autonomous driving models
Strong at designing reward functions and reproducing top conference/journal algorithms

📺 Demo Video of ICRA RoboMaster 2022