Yecheng (Geoffrey) WANG

Chengdu, CHINA | SWJTU-Leeds Joint School

Independently developed an enhanced YOLOv8n algorithm integrating a P2 detection layer, Non-Local Block attention, and BiFPN for complex underwater environments to improve marine benthic organism detection for superior feature extraction and robustness, outperforming baselines with minimal complexity.

Hangzhou, CHINA | KNEWBOTS

Conducted production-level calibration and large-scale regression case testing (600+ metrics) of multi-sensor SLAM systems on Autonomous Mobile Robots. Diagnosed hardware perception failures and addressed deployment-stage edge cases in multi-sensor fusion and loop closure.

Leeds, UK | MECH2636 Assignment 3 | Developer & Modeler

Developed a MATLAB Inverse Kinematics solver and coordinate transformation framework for a 2-DOF robotic arm in a colonoscopy simulator. Engineered a full-stack simulation to validate reachability and automate data pipelines for C++ embedded integration.

Leeds, UK | MECH2636 Assignment 2 | Control Systems Developer & Embodied Systems Engineer

Engineered a C++ embedded control architecture featuring a Finite State Machine and PI velocity controller for an autonomous delivery buggy. Validated on ReLOAD platform, optimizing PID gains and sensor placement to mitigate wheel slip and diverage and ensure precise payload delivery.

Chengdu, China | MECH1010 Coursework | Control Hardware & Software Developer

Developed an autonomous UAV altitude control system using Arduino and C++, achieving stable hovering within ±0.02m via optimized PID parameters, integrating full-stack hardware and telemetry log analysis.

Chengdu, China | MECH1206 | MATLAB Modeler & Analyst

Developed a 2D truss bridge mathematical model using engineering statics and solved a 24-Equation Matrix via MATLAB for load simulation. Created a MATLAB GUI to visualize empirical correlation, successfully predicting vulnerable components and maximum load capacity.

Chengdu, China | Student Research Training Program | Designer & Hardware & Embodied Developer

Developed a Raspberry-Pi driven smart cart integrating vision and weight sensors, using an MSDA-enhanced YOLOv8n backbone to achieve 92.4% mAP. Designed a 2-DOF pan-tilt laser pointer feedback mechanism to physically indicate non-recognized items.