Library

Curriculum Vitae

Last updated: Jul 2026

Profile

Design Engineering student at Imperial College London’s Dyson School, working across robotics, computer vision, and embodied AI. My research bridges learning-based perception with verifiable, safe planning for reliable real-world autonomy—from certified UAV navigation in dynamic crowds to open-vocabulary 6D pose estimation and vision–language–action policies for humanoid manipulation.

Education

  • June 2023 — June 2027 · MEng Design Engineering
    Imperial College London · Dyson School of Design Engineering · First Class (Expected)
    Main courses: Machine Learning, Robotics, Computing, Data Science, Industrial Design Engineering, Finite Element Analysis, Electronics, Physical Computing, Materials and Manufacturing.

Publications

  • Ye, B., & Boyle, D. Certified Safe Go-Around for UAV Navigation in Dynamic Crowds. First-author manuscript in preparation; target venue: IEEE RA-L / ICRA 2027.
  • Hou, P., Zhang, Y., Ye, B., & Wu, Y. (2026). Bridging the Depth Gap: Adaptive Scene-Instance Alignment for Training-Free Depth Refinement in Robotic Manipulation Scenes. IEEE Robotics and Automation Letters, vol. 11, no. 5, pp. 5454–5461. doi:10.1109/LRA.2026.3673899
  • Hou, P., Zhang, Y., Zhou, W., Ye, B., & Wu, Y. (2025). A lightweight network for category-level open-vocabulary object pose estimation with enhanced cross implicit space transformation. Engineering Applications of Artificial Intelligence, Vol. 155, 110890. doi:10.1016/j.engappai.2025.110890

Research Experience

  • April 2026 — Present · Certified Safe Go-Around for UAV Navigation in Dynamic Crowds
    Individual research supervised by Dr David Boyle, Systems and Algorithms Laboratory, Imperial College London
    Built a certified safety layer that lets a quadrotor fly the fastest safe go-around near moving pedestrians instead of only halting: constant-acceleration Kalman prediction feeding an EGO-Planner, verified by a Bernstein / de Casteljau exact continuous-time collision certificate with conformal-prediction margins. Zero collisions across 304 certified flights on a 38-scenario MetaUrban benchmark (vs. 9.2% for the native EGO-Planner), with a proven ≤5% collision-probability bound.

  • July 2025 — October 2025 · Research Assistant, Central Research Institute, Midea Group (Foshan, China)
    Developed a Vision Pro–based humanoid teleoperation system with spatial anchoring and low-latency streaming; deployed Physical Intelligence’s π₀.₅ vision–language–action policy for autonomous pouring and picking. Designed an open-vocabulary 6D pose estimation pipeline (GroundingDINO, SAM-HQ, DINOv3, FoundationPose) and a Blender synthetic-data pipeline with domain randomization.

  • July 2024 — September 2024 · Research Assistant, Embodied Intelligence, KUKA Robotics (Shanghai, China)
    Implemented oven-door manipulation and screw-tightening on a household humanoid using stereo disparity; built reinforcement-learning pipelines in RoboCasa, Isaac Gym, and MuJoCo with sim-to-real transfer; developed ROS status-monitoring and facial-expression modules for observability and human–robot interaction.

  • July 2024 — September 2024 · Vision-Guided Robotic-Arm Grasping (Self-directed)
    6-DoF arm grasping system in Python over UART, using YOLOv8 on Intel RealSense RGB-D input, hand–eye calibration, and IKPy inverse-kinematics planning.

Teaching

  • April 2026 — June 2026 · Undergraduate Teaching Assistant, Electronics 1
    Imperial College London — ran first-year lab sessions, helped students debug analogue and digital circuits, and marked lab work with individual feedback.

Selected Course Projects

  • Nature Kit — Early-childhood interactive product reducing passive screen time, with manufacturing-ready CAD and a generative storytelling module (Jan–Jun 2025)
  • Statistical ML for Successfulness Prediction — 81.7%-accuracy GDP classification with Random Forest / XGBoost, custom Performance Index, and UMAP diagnostics (Apr–Jun 2024)

Technical Skills

Programming

  • Python, C++, MATLAB, JavaScript

Machine Learning & Robotics

  • PyTorch, NumPy, OpenCV, Open3D
  • ROS / ROS2, MuJoCo, Isaac Gym, RoboCasa
  • Blender Python API

Tools

  • Git, Linux, Fusion 360, LaTeX

Languages

  • Mandarin Chinese — Native
  • English — Fluent (TOEFL 112)

Contact