Hi, I'm Truong Thien An — you can call me Andy.
A fourth-year student in the Honors Program in Computer Engineering at HCMUT – VNU-HCM, working at the intersection of Artificial Intelligence and resource-constrained hardware.
Currently in UAV/Robotics R&D, my focus spans Deep Reinforcement Learning, Edge AI, and real-time perception systems. I'm drawn to the challenge of bridging theoretical AI with the strict latency and memory constraints of embedded and autonomous platforms — translating raw sensor data into reliable, real-world autonomous actions.
A fast adapter in high-paced research environments, I gravitate toward problems that demand both rigorous engineering and creative problem-solving.
| Project | What I did | Tech | Link |
|---|---|---|---|
| Autonomous UAV Navigation & Obstacle Avoidance (DRL) April 2026 – Present · Internship Project |
• Designed and trained a PX4-based quadrotor (x500_depth) to navigate cluttered environments and avoid randomized pillar obstacles in Gazebo SITL. • Built a 5-stage curriculum that progressively introduces obstacle density, domain noise (sensor drift, actuator jitter, wind, mass variation), and sim-to-real transfer constraints — enabling the agent to generalize from a clean simulator to noisy real-world dynamics. • Engineered a potential-based reward shaping (PBRS) system with a hybrid progress signal, stabilizing policy convergence and eliminating reward hacking artifacts. • Developed a proprietary real-time perception pipeline converting stereo depth and LiDAR into structured spatial representations for the RL policy — designed for sub-50 ms end-to-end latency on edge hardware. • Architected a multi-environment parallel training setup with full ROS 2 / Gazebo isolation per rank, enabling reproducible multi-instance SITL at scale. |
     |
🔗 GitHub |
| RobotDog Lite3 (DeepRobotics) – Autonomous Navigation & Hazard Detection Oct 2025 – Present · URA Research Group, HCMUT |
• Developed RL components for autonomous movement and decision-making on a physical quadruped platform. • Owned end-to-end Motion API: state feedback, safety constraints, UI–backend command integration, and documentation. • Codebase proprietary; details available upon request. |
  |
🔒 Private (Coteccons/URA) |
| Environmental Sound Recognition (Edge ML on Embedded Device) Oct 2025 – Dec 2025 · Team of 4 |
• Trained an environmental sound classification model — dataset preparation, augmentation, and evaluation pipeline. • Integrated on-device inference and benchmarked latency vs. accuracy trade-offs across model compression strategies. |
   |
🔗 GitHub |
| STM32 Multi-Mode Traffic Light Controller Nov 2025 – Dec 2025 · Team of 2 |
• Designed system-wide FSM for multi-mode transitions and deterministic timing. • Implemented traffic behaviors in Embedded C with a lightweight cooperative scheduler. |
   |
🔗 GitHub |
 |
antruong.andy.work@gmail.com |
 |
linkedin.com/in/antruong24112005 |
 |
Trương Thiên Ân |
 |
@trth_an2411 |