Sungjun Heo

Ph.D. Candidate at Graduate School of Artificial Intelligence, UNIST

image_prof.jpg

Robotics & Mobility Laboratory

UNIST, South Korea

I am a Ph.D. Candidate at the Graduate School of Artificial Intelligence in UNIST, advised by Prof. Jeong hwan Jeon. I am a member of the Robotics & Mobility Laboratory.

My research fields include:

  • Autonomous Driving - E2E Driving / Learning based Trajectory Planning
  • Robotics - Learning based Manipulation / Multi-Agent RL
  • Reinforcement Learning - Hierarchical RL / Sim-to-Real Robotics RL
  • End-to-End Learning - Vision to Control / Data Distribution


Google Scholar

Curriculum Vitae (CV)

LinkedIn

News

Aug 07, 2026 🎉 IEEE SMC 2026 Student and Young Professional Travel Grant has been accepted.
Jun 17, 2026 🎉 My 4th and 5th papers are accepted / There are two papers have been accepted in IEEE International Conference on Systems, Man, and Cybernetics (2026 SMC) - Flagship conference of the IEEE SMC Society
Jun 17, 2026 🎉 My third paper has been accepted in IEEE/RSJ International Conference on Intelligent Robots and Systems (2026 IROS)-(Major Conference of IEEE RAS)
May 06, 2026 🎉 (Top 8.5% of all Accepted Papers) Second paper has been selected for an Oral Presentation at the 2026 IEEE Intelligent Vehicles (IEEE IV 2026)
Jan 15, 2026 🎉 Second paper has been accepted in 37th IEEE Intelligent Vehicles (IEEE IV 2026)-Flagship Conference of IEEE ITSS
Dec 05, 2025 🎉 Grand Prize - Brain To Society, U-Challenge Festival (1 / 96 teams)
Sep 30, 2025 🎉 1st place in 2025 Hyundai Motor Group's Autonomous Driving Challenge !!
Jun 24, 2025 Attending 36th IEEE Intelligent Vehicles (IEEE IV 2025)
Mar 31, 2025 🎉 My first paper has been accepted in 36th IEEE Intelligent Vehicles (IEEE IV 2025)-Flagship Conference of IEEE ITSS !!
Mar 27, 2025 🎉 3rd place in Preliminary Round of 2025 Hyundai Motor Group's Autonomous Driving Challenge !!

Publications

  1. To be disclosed after the double-blind peer review process
    Sungjun Heo, S. Yang, H. Jeong, J. Ha, and J. Jeon
  2. SHIFT-RL: Sensor-driven Hierarchical Information Fusion Transformer for BEV-based Maneuvering in Dense Multi-lane Environments
    Sungjun Heo and Jeong hwan Jeon
    Accepted; to be presented at: IEEE/RSJ International Conference on Intelligent Robots and Systems (2026 IROS)
  3. Capacity-Controlled Enforcement of Sparse Traffic-Light Cues for Rule-Conformant End-to-End Driving
    Sungjun Heo, H. Jeong, E. Kim, and 4 more authors
    Accepted; to be presented at: IEEE International Conference on Systems, Man, and Cybernetics (2026 SMC) · Oral Presentation
  4. Decoupling Actor and Critic Replay for Offline Reinforcement Learning with Three-Factor Selection
    Sungjun Heo and Jeong hwan Jeon
    Accepted; to be presented at: IEEE International Conference on Systems, Man, and Cybernetics (2026 SMC) · Oral Presentation
  5. Deployment-Oriented End-to-End Autonomous Driving: Enhancing Closed-Loop Stability with a Lightweight Framework
    DOI BIB
    H. Jeong, E. Kim, Sungjun Heo, and 4 more authors
    In IEEE Intelligent Vehicles Symposium (2026 IV) · Oral Presentation (Top 8.5% of Accepted Papers)
  6. Predicted State-Based Hierarchical Reinforcement Learning for Long-Term Decision Making in Urban Dynamic Scenarios
    DOI BIB
    Sungjun Heo and Jeong hwan Jeon
    In IEEE Intelligent Vehicles Symposium (2025 IV)

Awards and Achievements

2026

  • IEEE SMC 2026 Student and Young Professional Travel Grant

    Award amount: $600

2025

  • 1st Place

    1st Place: Hyundai Motor Group's Autonomous Driving Challenge

    Award amount: $22,000

    Hyundai Motor Group Team FAST (UNIST)

    Developed and implemented robust End-to-End autonomous driving algorithms to win 1st place in the competition.

    Special privilege: Autonomous driving field trip

    News
    Project Details
    Overview

    Development of a Camera-Only, Deployment-Oriented End-to-End Autonomous Driving Model and the entire pipeline from data collection to closed-loop validation.

    Competition Progression

    1. Round 1 Qualifiers

      16 Teams → 8 Teams

    2. Round 1 Main Round

      8 Teams → 6 Teams

      3rd Place

    3. Round 2 Qualifiers

      6 Teams → 4 Teams

    4. FINAL

      1st Place among 4 Finalists

      Winner
    Technical Details
    Full-Stack E2E Development
    Developed an End-to-End autonomous driving model from scratch, covering the entire pipeline from architecture design to implementation.
    Robustness
    Applied Imitation Learning and DAgger (Dataset Aggregation) to handle covariate shift and out-of-distribution scenarios.
    Data Pipeline
    Collected and refined approximately 24 hours of driving data using multi-view cameras, IMU, and GPS.
    Consistency Enhancement
    Developed specialized modules and loss functions to ensure prediction consistency between planning and control heads, addressing a key challenge in End-to-End models.
    Labeling Policy
    Established a data labeling policy to enhance closed-loop driving stability.
  • Grand Prize

    Grand Prize: Brain To Society, U-Challenge Festival

    Award amount: $3,600

    UNIST Team leader · 1st out of 96 teams

    Project: Sensor-driven Hierarchical Information Fusion Transformer for BEV-based Maneuvering in Dense Multi-lane Environments

    News
  • 3rd Place

    3rd Place: Hyundai Motor Group's Autonomous Driving Challenge

    Award amount: $3,600

    Hyundai Motor Group Team FAST (UNIST) · Round 1 main round

    News (opens in a new tab)
  • Excellence Award

    Excellence Award: AI Tech Open Workshop

    Award amount: $2,400

    UNIST Team FAST

    Research: Development of an End-to-End Autonomous Driving Model Based on a High-Dimensional Simulator

2023

  • Undergraduate Research Award

    Research: Autonomous Driving Waypoint-generation based on Vehicle Kinematics

2021

  • Academic Achievement Award

    Undergraduate academic achievement excellence award.