Robot learning, from data to deployment机器人学习：从数据到真机部署
VLA fine-tuning and multimodal control lead this portfolio, supported by published tactile research, industrial inspection and efficient neural networks. Each case explains my contribution, implementation and current evidence.重点展示 VLA 微调与多模态机器人控制，并通过已发表的触觉研究、工业检测和高效神经网络研究说明技术积累。每个案例交代我的贡献、实现过程与验证进展。
Each case separates the problem, my contribution, the technical path and the evidence. Open one for the complete story.从 VLA 示教、微调、部署与评测出发，延伸到触觉探索、工业机器人集成和脉冲模型压缩。各项目分别标明已完成工作与持续研发内容。
DEMONSTRATE遥操示教2× FRANKA
→
FINE-TUNE模型微调π0.5-DROID · 8× H100
→
DEPLOY本地部署RTX 5090 · TACTILE POLICY触觉策略
MODEL / VTLA PIPELINE模型 / VTLA 链路Concept model概念图
01Robotics机器人
Active development正在开发
Tactile VLA fine-tuning and real-robot control with π0.5-DROIDπ0.5-DROID 触觉 VLA 微调与真机控制
Adapting π0.5-DROID to visual–tactile demonstrations: reproducible data preparation, shared image adapters and three completed LoRA fine-tunes.基于视觉触觉示教适配 π0.5-DROID，建立可追溯的数据处理与共享图像接口，完成三组 LoRA 微调。
I lead the Bristol-VLA project and its VLA workstream, working with colleagues on hardware and demonstration collection. My focus is turning multimodal observations into auditable training inputs and deployable policy interfaces for Franka manipulation.我负责 Bristol-VLA 项目及其 VLA 路线，与团队协作开展硬件集成和示教采集，重点将多模态观测转为可核验的训练输入，以及面向 Franka 操作的策略接口。
Role项目角色
Project Lead · VLA Development项目负责人 · VLA 研发
Support项目资助
NVIDIA Developer Program · 8× H100 training accessNVIDIA Developer Program · 8 张 H100 训练资源
Autonomous composite quality assurance with neuromorphic touch用神经形态触觉检测复合材料缺陷
Developing event-based tactile sensing to detect surface defects in composite manufacturing.针对复合材料制造中的表面缺陷，开发事件驱动触觉传感器与机器人检测系统。
Independently wrote the successful proposal and lead hardware development for low-latency composite inspection.担任项目联合负责人，独立完成申请并获批，目前负责触觉传感器研发与硬件集成，面向复合材料表面检测。
Preprint · workshop version accepted预印本 · 研讨会版本已接收
Efficient spiking models for neuromorphic touch面向边缘设备的脉冲神经网络压缩
Compressing spiking networks for low-resource deployment.这项研究压缩训练好的脉冲神经网络，使其能部署在资源有限的设备上。
Develops one-shot post-training pruning and quantisation methods that preserve useful spike dynamics while reducing model cost.我参与训练后单步剪枝与量化研究：模型训练完成后直接压缩，无需重新训练。
How do touch speed and exploratory movements change what people and robots learn about texture?通过人类与机器人实验，比较触摸速度、接触力和探索动作对纹理感知的影响。
Studies velocity modulation and compound exploratory motions to improve texture discrimination under realistic contact variation.我通过人类触摸实验和神经形态机器人实验，分析速度与复合动作对纹理识别的影响。
Role项目角色
First author · Equal-contribution researcher第一作者 · 共同一作
Neuromorphic tactile framework for composite surfaces复合材料表面的触觉检测
A seed project connecting event-based touch with real-time surface analysis.这项种子项目研究用事件驱动触觉实时分析复合材料表面的方法。
Led an early-career award to establish the sensing, data and experimental framework for tactile composite inspection.我以 PI 身份主持这项种子项目，确定复合材料触觉检测的传感方案、数据处理方法和实验设计。
Robotics engineering and research communication work that shaped how I build and explain technical systems.除研究项目外，我还在企业承担过机械臂测试、技术文档和应用演示，并为大学制作科研视频、运营官方社交媒体。
A01
Communication传播
Research communication at Bristol布里斯托大学科研传播
Directed, filmed and edited university research videos, managed social content and supported international students.我策划、拍摄并剪辑大学科研视频，运营社交媒体，并为国际学生答疑。
Tested Magician, M1 and M1 Pro robots, developed application demonstrations, and worked on PLC communication, hand–eye calibration and palletising during an engineering internship.实习期间为 Magician、M1 和 M1 Pro 开展真机测试、技术文档与应用演示，并参与 PLC 通信、手眼标定和码垛流程开发。