• B.S. Candidate in Information & Communication Engineering, Chungbuk National University
  _{Mar. 2023 – Present}
• Research Student, Multimedia Information Processing (MIP) Laboratory
  _{Mar. 2025 – Jun. 2026}
• Advanced Division Leader, HyperCore AI & Data Analytics Club
  _{Jan. 2026 – Present}
• 🔭 Preparing for graduate study in AI/NLP

• Faculty-supervised research project on robust reinforcement learning-based autonomous driving with multimodal sensor fusion.
• Implemented a Shapley-guided multimodal fusion pipeline using RGB, LiDAR, route, and ego-state observations.

• Designed and implemented the multimodal fusion structure.
• Built the reinforcement learning pipeline and experiment workflow for contribution-guided sensor fusion.

• Built a sensor fusion pipeline for RGB, LiDAR, route, and ego-state inputs.
• Organized the project into prototype, experiment, and final implementation repositories.
• Related manuscript submitted to KIISE Transactions on Computing Practices (KTCP), 2026.

• Main Implementation
• Experiments
• Prototype

Jun. 2025 – Jun. 2026

Python · PyTorch · Stable-Baselines3 · CARLA · Reinforcement Learning

• Solar Pro3-based AI agent service for tutor lesson reports, payment reminders, schedule coordination, and parent communication.
• Built an educational AI service that helps tutors manage lesson records and parent-facing communication through LLM agents.

• Implemented AI-agent workflows and backend infrastructure.
• Connected structured agent outputs to the service flow for tutoring operations.

• Designed agents for lesson reports, parent communication, payment reminders, and schedule coordination.
• Used structured JSON-style outputs to connect LLM reasoning with application features.
• Developed as part of the Upstage MixUp Agent Hackathon.

• solar-tutorboard

May. 2026

LLM Agent · Solar Pro3 · Python · TypeScript · Supabase

• End-to-end trajectory forecasting project on the Argoverse 2 Motion Forecasting dataset.
• Implemented a forecasting pipeline that predicts future vehicle and pedestrian trajectories from past motion histories.

• Built data processing, model training, evaluation, and visualization workflows.
• Compared multiple forecasting approaches under a consistent experimental setup.

• Compared Linear, LSTM, Transformer, Direct Diffusion, and PCA Latent Diffusion models.
• Evaluated models with common trajectory forecasting metrics and visual analysis.
• Focused on reproducible end-to-end implementation rather than leaderboard-only tuning.

• vehicle-trajectory-prediction-av2

Mar. 2026 – Jun. 2026

Python · PyTorch · Argoverse 2 · LSTM · Transformer · Diffusion Models

• WiFi CSI-based patient monitoring MVP for bed-state classification, fall-risk event detection, and respiration monitoring.
• Designed a camera-free monitoring system that uses wireless channel changes around a hospital bed to detect patient state and risk events.

• Built the MVP pipeline for CSI window features, model comparison, and monitoring-oriented classification.
• Organized the system around practical privacy-preserving patient monitoring scenarios.

• Classified bed-related states such as lying, standing, moving, leaving bed, and fall-like risk events.
• Used WiFi CSI signals as a privacy-preserving alternative to camera-based monitoring.
• Explored lightweight deep learning models for practical monitoring scenarios.

• hospital-fall-detection-wifi-csi

Apr. 2026 – Present

Python · WiFi CSI · Deep Learning · TinyTransformer · Signal Processing

Programming

Machine Learning Frameworks

AI / Machine Learning

Autonomous Driving / Mobility

Development / Tools

Data / Signals

Published

• Hyeban Namgung, Kuk Seorin, Park Geonwoo, Bae Seohyun, Song Yonghwi, Lee Hyesong, Han Jian, Hyojoong Kim, "A Comparative Study on the Prevalence of Forensic Flies Using Chicken Corpse", Journal of Science Education for the Gifted, 2020.

Manuscripts

• Under Review: Song, YongHwi, [Co-authors]. "Shapley-style Contribution-Guided Sensor Fusion for Robust Reinforcement Learning-Based Autonomous Driving." Submitted to KIISE Transactions on Computing Practices (KTCP), 2026. First author.
• In Preparation: Song, YongHwi. "A Comparative Survey of Domestic and International Reinforcement Learning Research for Autonomous Driving." Manuscript in preparation, 2026.

• Email: thddydgnl1937@gmail.com
• GitHub: github.com/thddydgnl