GitHub Profile →
research.hachile.org →
National Undergraduate Innovation and Entrepreneurship Training Program · Lead Planner
Southwest Building 1 Digital Twin Platform
Built a future-college community platform for Southwest Building 1 using the infrastructure digital base developed by Academician Zhu Hehua's team and Tongji University's iS3 Lab. I led proposal development, team recruitment, laboratory and faculty coordination, plan writing, API applications, Blender 3D modeling, cloud-server evaluation and deployment, and UI design. The project was later approved as a national-level undergraduate innovation project.
Independent Development
IASR System: Integrated Air Situation Recognition System
Processes video for object detection and tracking, calculates traffic-situation indicators and congestion indices, estimates visibility from video clarity, emits events in a unified format, and routes them to operations, security, emergency response, or traffic-management recipients according to configuration. The project received third prize at Guohao College's 3rd “Yanxue Jihao” Poster Exhibition.
National Undergraduate Innovation and Entrepreneurship Training Program · Team Member
AegisDrive: World-Model-Assisted Minimum-Risk Decision System for Autonomous-Driving Failures
Uses a dual-track architecture combining in-cabin monitoring and external perception. Eye and mouth geometry rules determine driver state, while a lightweight YOLOv8n model performs vehicle detection and monocular ranging. A world-model-based pipeline processes two video streams concurrently with Python multiprocessing and issues real-time warnings. I developed the in-cabin facial-recognition component and contributed to project documentation. The project won second prize in the freshman division of the 25th Tongluren Transportation Science and Technology Competition, was upgraded to a national-level undergraduate innovation project, and received a silver award in the Tongji University round of the 9th China International College Students' Innovation Competition.
SITP Team Member · Embodied Intelligence
QDepth-VLA: Lightweight Vision Models for Orchard Harvesting
Generates pseudo-depth with a monocular depth model, discretizes it into depth tokens through a VQ-VAE, and injects geometric information into the VLA backbone through a Depth Expert and Hybrid Attention. Using only a standard RGB camera, the project validates a closed-loop recognition–localization–harvesting prototype in simulation and on a desktop robotic arm. I handled early literature research, faculty and laboratory coordination, harvesting-method analysis, and robotic-arm development. The work combines two SITP undergraduate innovation projects and is competing in the 15th Challenge Cup Chinese College Students' Entrepreneurship Competition.
Independent Development · Time-Series Analysis
Multimodal Cockpit Seat-Pressure Analysis System
Analyzes pressure trends, stability indicators, and temporal features in cockpit seat-pressure sensor data across different body-weight groups, providing data support for seat-design optimization and ergonomics research. Uses pandas, NumPy, and Matplotlib for data processing, numerical computation, and visualization.
Development Team Member · Data Systems
Global University Faculty Information System
Proposed and supported with data by the APJIE journal at the Shanghai International College of Intellectual Property. Built with Python web scraping and an SQL database, the system supports queries by age, institution, research output, and other dimensions.
First Author · Manuscript Under Review
AI-Based Analysis and Prediction of Lithium-Dendrite Growth in Lithium-Metal Batteries
Constructed a multimodal dataset containing scanning-electron-microscope images, electrochemical time series, and electrolyte and interface descriptors. Used CNNs for dendrite morphology recognition and grading, convolutional and recurrent neural networks to predict polarization voltage and overpotential fluctuations, and regression models to analyze relationships among electrolyte composition, transport properties, and dendrite growth, providing an interpretable screening approach for safer lithium-metal battery design.