Peiyang(Ben) Li
I am a student passionate about computer science, engineering, and physics. I am currently studying reinforcement learning, computer vision, and robotics with IsaacLab. I am open to all challenges and opportunities in the field.
Education
Bachelor's Degree, Northwestern University
2025 - 2029
IBDP Candidate, Shanghai Pinghe School
2023 - 2025
Graduated on the Principal's List (top 5% of class)
Work Experience
Research Intern (Algorithm), Differential Robotics, Hangzhou
June 2025 - August 2025
Worked on two autonomous drone projects:
- Optimized sparse reward RL for obstacle avoidance navigation
- Implemented real-time computer vision for high-speed stunt maneuvers, achieved 90% success rate in extreme conditions using MobileNetV3 architecture
Projects
End-to-End RL Drone Navigation System
2025
- Built an end-to-end navigation pipeline in IsaacLab using deep reinforcement learning (PPO), mapping high-dimensional sensor inputs (depth images + proprioceptive states) directly to low-level motor actions.
- Implemented curriculum learning & domain randomization (obstacle density, dynamic disturbances, sensor noise) to improve generalization and enable sim-to-real transfer.
- Designed hybrid reward shaping: integrated Dijkstra-based path priors, directional distance rewards, ESDF collision penalties, and smoothness constraints to stabilize long-horizon flight.
End-to-End High-Speed Drone Frame Navigation System 2025
2025
- Developed an End-to-End vision-based navigation system enabling drones to traverse 85° tilted frames at 4–6 m/s with near-100% success under standard conditions.
- Developed a custom deflicker algorithm to mitigate indoor lighting flicker, ensuring stable frame detection.
- Designed and trained CNN-based segmentation models with multi-color frame detection, incorporating extensive data augmentation (noise, brightness shifts, geometric transforms) to enhance robustness.
- Optimized MobilenetV3 + DeeplabV3+ backbone by reducing network depth; achieved ~3ms inference latency on Jetson Xavier NX using FP16 TensorRT quantization. Demonstrated superior speed–accuracy tradeoff compared to UNet, SegFormer, MobileViTs, and EfficientNet.
- Trained control policy in simulation with RL, leveraging shaped rewards (center alignment, collision penalties, smoothness constraints, Distillation-awareness Regularization); deployed as a compact network for real-time execution.
- Deployed segmentation and policy jointly with combined latency <16ms, enabling >60Hz control frequency for onboard flight.
- Conducted extensive real-world tests, validating segmentation robustness and system stability in high-speed traversal.
- Acquired strong manual piloting skills, supporting iterative testing and system verification.
Modeling Soccer Ball Trajectory
2023
- Developed computational simulation using Python, hydrodynamics, and ML
- Incorporated Magnus effect for accurate modeling
- Achieved <40cm trajectory error in windless conditions
Educational Inequality & Technological Development in China
2023
- Conducted quantitative provincial analysis using OLS Regression
- Created Educational Equality Index (EEI) with over 20 sub0indices to measure disparities thoroughly
- Revealed counterintuitive phenomena that certain developed provinces have greater inequality due to social segregation.
Decoding Online Dialogue: Message Context & Response Patterns
2024
- Analyzed response patterns, personality, and generation using mixed-method approach
- conducted 200+ participants email experiment combined with semmi-structured interviews (N=5)
- Found blank email subjects had higher read rates, but not necessarily response rates.
- Also found out that email with a more official tone are more likely to have higher response rate.
Synchronous vs. Asynchronous Communication in Virtual Student Interaction
2024 UC Santa Barbara Summer Research Academy
- Conducted mixed-method analysis (N=123) on group cohesion at UC Santa Barbara Summer Research Academy
- Found asynchronous tools increased efficiency and flexibility even though they seem less effecitve
- Gained academic writing skills
Skills
LaTeX
Python
Data Analysis
Machine Learning
Reinforcement Learning
Computer Vision
Web Development
Drone Operation
Photography
Video Editing (DaVinci Resolve, Premiere, After Effects)
3D modeling (Blender)
Awards
Mathematics & Modeling
- High School Mathematical Contest in Modeling (HiMCM): Outstanding (Global Top 1%) Link to paper (***)
- American Regions Mathematics League (ARML): Global Top 10
Physics
- British Physics Olympiad (BPhO): Gold (Global Top 10%)
- PhysicsBowl: Gold (Global Top 2%)
Chemistry
- UK Chemistry Olympiad (UKChO): Gold (Top 8%)
Media & Arts
- Original Barrier-Free Movie Creation Competition (Regional): Silver Award (Top 2)
Languages
Chinese (Native), English (Native), Japanese (Basic)
Interests
Photography, Filmmaking, Travel, Soccer, Badminton.****(create a page on twisonl using cf images (cargo style)) View my photography portfolio for creative work samples.
References
CV template inspired by Jon Barron