CV

Contact Information

Name Junfeng Ren
Professional Title M.S. Student in Electronic Information
Email junfengren3253@gmail.com
Location Shenzhen, China

Professional Summary

M.S. student at the Southern University of Science and Technology (SUSTech), working on 3D perception and scene understanding for autonomous driving and embodied perception. My research focuses on semantic occupancy prediction, collaborative perception, efficient multi-agent communication, and occupancy world models.

Experience

  • 2023 - 2024

    Qingdao, China

    Research Assistant
    IoT Engineering Laboratory, Shandong University of Science and Technology
    • Conducted research on real-time scheduling optimization for embedded IoT systems
    • Implemented simulation experiments and co-authored an IEEE conference paper on improved RMS-based scheduling

Education

  • 2024 - Present

    Shenzhen, China

    M.S.
    Southern University of Science and Technology
    Electronic Information
    • Research focus: 3D scene understanding, collaborative perception, and Occupancy World Models
  • 2019 - 2023

    Qingdao, China

    B.S.
    Shandong University of Science and Technology
    Internet of Things Engineering
    • Background in embedded systems, computer vision, machine learning, and IoT applications

Publications

  • 2026
    Learning to Merge Tokens for Communication-Efficient Collaborative Occupancy Prediction
    Submitted to NeurIPS 2026
  • 2026
    Bandwidth-Aware Adaptive Token Communication for Collaborative Occupancy Prediction
    Submitted to IEEE ITSC 2026
  • 2026
    Collaborative 4D Occupancy World Models with Motion-Aware Token Memory
    Manuscript in preparation
  • 2023
    Scheduling Optimization Design of IoT Embedded System Based on Improved RMS Algorithm
    IEEE AIKIIE2023

Skills

Programming & Systems: Python, PyTorch, CUDA, Linux, C/C++
Machine Learning: Deep Learning, CNNs, Transformers, Representation Learning, Multi-Modal Learning
Computer Vision: 3D Perception, Semantic Occupancy Prediction, Occupancy World Models

Languages

Chinese : Native
English : IELTS 6.0

Interests

Research Interests: Computer Vision, 3D Perception, Semantic Occupancy Prediction, Collaborative Perception, Multi-Agent Communication, Occupancy World Models

Certificates

  • Software Copyright - IoT Data Monitoring and Analysis System - National Copyright Administration of China (2023)

Projects

  • Learning to Merge Tokens for Collaborative Occupancy Prediction

    Research on communication-efficient collaborative 3D semantic occupancy prediction for autonomous driving, focusing on tokenized scene representations, spatio-temporal memory, receiver-driven communication, and content-aware token merging under limited communication bandwidth.

    • Developed a token-based collaborative occupancy prediction framework for multi-agent autonomous driving scenarios
    • Designed a spatio-temporal memory module to aggregate temporally consistent BEV representations across frames
    • Built a receiver-driven communication mechanism that allows the ego agent to request informative regions from neighboring agents
    • Proposed content-aware token merging to preserve request-relevant tokens while merging redundant low-priority tokens before transmission
    • Evaluated the framework on Semantic-OPV2V, achieving a favorable accuracy–communication trade-off for collaborative occupancy prediction
    • Submitted this work to NeurIPS 2026
  • Bandwidth-Aware Adaptive Token Communication for Collaborative Occupancy Prediction

    A research project on adaptive token communication for collaborative 3D occupancy prediction. Instead of using fixed request sizes and fixed transmission budgets, this project studies how autonomous agents can dynamically determine how many tokens to request, protect, and transmit according to scene complexity, token importance, and available bandwidth.

    • Proposed a bandwidth-aware adaptive communication strategy for collaborative occupancy prediction
    • Replaced fixed Top-K request and protection settings with adaptive token selection mechanisms
    • Studied dynamic transmitted-token allocation to balance perception accuracy and communication cost
    • Used scene complexity, token importance, and uncertainty cues to guide adaptive communication decisions
    • Built upon token-based collaborative perception with spatio-temporal memory and communication-aware fusion
    • Submitted to IEEE ITSC 2026
  • Collaborative 4D Occupancy World Models with Motion-Aware Token Memory

    A research project that explores how collaborative perception can support occupancy-based world modeling. Instead of only predicting the current 3D scene, this project aims to model how the occupancy state of a dynamic environment evolves over time by combining multi-agent observations, motion-aware token memory, and future scene prediction.

    • Formulated collaborative 3D occupancy prediction as a step toward 4D occupancy world modeling
    • Designed motion-aware token memory to capture temporal dynamics in compact occupancy representations
    • Used multi-agent observations to improve future prediction in occluded, uncertain, and dynamic regions
    • Studied future occupancy forecasting as a bridge between perception, temporal reasoning, and world models
    • Explored token-based representations for scalable 4D scene understanding
    • Manuscript in preparation for CVPR 2027
  • Autonomous Driving Perception and Decision System on NVIDIA Jetson

    An undergraduate embedded autonomous driving project integrating perception, decision-making, and control on NVIDIA Jetson.

    • Built a real-time perception pipeline with YOLOv5 object detection and OpenCV-based lane detection
    • Designed rule-based decision logic and a finite-state machine for lane following, turning, and obstacle avoidance
    • Deployed and optimized the system on NVIDIA Jetson for embedded inference and control
  • Embodied AI Soccer Task System on NAO Robot

    An undergraduate robotics project for vision-based ball detection, navigation, and kicking on a NAO humanoid robot.

    • Implemented real-time ball detection and distance estimation using visual and geometric cues
    • Designed a finite-state machine for search, approach, alignment, and kicking behaviors
    • Integrated visual feedback with NAOqi-based motion control for closed-loop task execution
  • Embedded Edge AI Face Recognition Access Control System

    Undergraduate thesis project on real-time face recognition and access control using lightweight edge AI on resource-constrained embedded devices.

    • Built a lightweight face recognition pipeline using MobileFaceNet and similarity matching
    • Applied model compression and quantization for embedded deployment
    • Integrated on-device inference, wireless communication, and mobile-based user management