Mingxiang Chen

About Me

I am a postgraduate student at the University of Manchester, studying Robotics. I received my BSc (Hons) from The Huazhong University of Science and Technology.
I am passionate about technology and enjoy using the latest software and hardware to solve problems.
In the future, I would like to start a startup focusing on a field that interests me, such as Autonomous Vehicles, SLAM, Computer Vision, and AI.

RESEARCH INTERESTS & OBJECTIVES

Autonomous Navigation

Preliminary development plan:

Multi-sensor fusion architecture design

Adaptive path planning algorithms

Real-time performance optimization

Dynamic obstacle prediction models

Long-term localization enhancement

Computer Vision

Core Research Directions:

Real-time multi-object tracking

Multi-view 3D reconstruction

Low-light image enhancement

Industrial defect detection

Multi-modal perception fusion

Deep Learning

Strategic Focus Areas:

Edge device model compression

Few-shot learning techniques

Vision-language fusion

Generative data augmentation

Explainable AI frameworks

WORKING PROJECTS

2020/07 ~ 2023/02 ,Image Processing Engineer, TPLINK

Responsible for the lens selection, function research and image related work of the company's various key projects (high sales volume and good user feedback) as well as the company's first and most valued full-featured high-speed dome project. Responsible for the training of new employees in the two courses ‘sensor principle and performance’ and ‘image brightness processing principle’ and the development of the ‘image brightness processing principle’ course.
Conducted several topic studies, such as the topic of humanoid optimisation.

2023/11/15~2024/07/29, Image Debugging Engineer, Beijing T&S Technology Co.

1）Complete this company's first camera image parameter debugging and shipment.
2）Write a lot of image group normative annotation technical guidance documents and test cases, etc.
3）Assist in the completion of the book laboratory and the big video division exhibition building.

UNIVERSITY PROJECTS

Workpiece Inspection

Gigabit Ethernet industrial cameras are used to acquire pictures in real time, and on the Microsoft Visual C++ platform, image processing knowledge is applied to achieve workpiece classification detection (such as screws, nuts, etc.) and automatic recognition. (such as screws, nuts, etc.), automatic identification of workpiece categories, counting the number of workpieces and area. First of all, image grey scale and binarization processing, the presence of noise on the image will affect the detection, so it is necessary to further image denoising pre-processing, and then find the connecting region, through the boundary detection and region growth and other ways to obtain features (size, shape, area, grey value of the centre and other features), and finally to achieve the goal of detecting and identifying.

Small target detection of ships in the field of navigation

In Visual Studio platform, C programming is used to achieve the detection of small targets on ships based on multi-feature analysis and morphological reconstruction. Accompanied by complex sea clutter background of different sizes, different numbers, different brightness, different shapes of the ship targets have a low false detection rate, the detection results using a rectangular box, the output of the number of detection results and the specific location.

Aviation booking and information management system

Realise the real-time aviation information management system with C language programming, provide friendly aviation passengers and administrators with basic operation interface, realise administrators‘ backstage information management (e.g. adding, deleting, modifying flights, etc.), passengers and administrators’ log-in, passengers booking and refunding, and passengers' information modification.

Remote sensing image key point detection

The neural network method Faster R-CNN is implemented on the basis of the neural network method. First of all, the algorithm comparison analysis research to determine the algorithm and optimisation strategy, and then the data set collection and slicing as well as data calibration, followed by the establishment of neural networks, the detection of key points such as the corners of the surface building to achieve accurate guidance, training of the neural network, and ultimately the neural network detection and so on.

ROBOTICS PROJECTS

Autonomous Leo Rover Robotics Project(Ongoing Optimization)

In the autonomous Leo Rover robotics project in Manchester, I was primarily responsible for navigation, mapping, final code integration, and logic design, while also assisting with the robotic arm grasping tasks. I utilized SLAM for real-time map construction, integrating LiDAR, IMU, and odometry data (EKF) to enhance localization accuracy. I implemented A for global path planning and TEB for local obstacle avoidance, optimizing the robot's autonomous movement capabilities. This experience enhanced my practical skills in Gazebo, ROS2, SLAM, and path planning, improved my simulation abilities, and deepened my understanding of autonomous navigation systems.

Adaptive A Path Planning Algorithm*

In developing the Adaptive A* path planning algorithm, I optimized search efficiency across different grid sizes by implementing three adaptive selection mechanisms using dictionaries, 2D arrays, and 1D memory-optimized structures. The algorithm supports efficient four-directional movement, leveraging the Euclidean heuristic and a six-layer error-checking mechanism to enhance path continuity and robustness. Through this project, I deepened my understanding of A* optimization strategies, strengthened my application of data structures in performance and memory management, and improved my ability to design intelligent navigation in complex environments.

Netflix Simple Robot

Using some commonly used waste items in life, we design a web robot that can be remotely controlled to walk, can walk autonomously, and can avoid obstacles when it meets obstacles. The robot's subsequent optimisation direction: automatic tracing function, voice control function, interactive function (including feedback on voice, vibration, etc.).

SLAMTEC RPLIDAR A2M12 Performance Evaluation Project

In the SLAMTEC RPLIDAR A2M12 performance evaluation project, I conducted experimental assessments of LiDAR accuracy, resolution, and detection range, analyzing its impact on the Leo Rover's SLAM and obstacle avoidance capabilities. I identified key limitations, including distance miscalculations, environmental noise interference, and vertical blind spots, and proposed multi-sensor fusion and algorithmic optimizations as improvement strategies. This experience enhanced my skills in robot perception, SLAM, sensor data analysis, and experimental evaluation, deepening my understanding of LiDAR-based mapping and autonomous navigation systems.

Drone 3D Feedback Control System Development Project (Ongoing Optimization)

As a core developer of this project, I was responsible for designing and testing the multi-axis PID-based drone position control algorithm. By incorporating differential filtering, integral clamping, and fifth-order trajectory planning, I optimized the steady-state error to 0.12m and reduced overshoot by 42%. Currently, I am proposing future optimizations and continuous improvements, including: (1) Developing a Model Predictive Control (MPC) module to enhance adaptability in dynamic environments; (2) Implementing reinforcement learning-based parameter auto-tuning to reduce manual tuning efforts; (3) Building wind disturbance and sensor noise models in a simulation environment to validate algorithm robustness. Through this project, I further enhanced my expertise in drone dynamics modeling, multivariable control optimization, and Python-based real-time system development.

Robotic Systems for Sustainable Development

Educational video about the contribution of robotic systems to addressing the challenges around sustainability and their ethical perspective.

TECHNICAL SKILLS

Robotics Systems

Manipulation Systems MoveIt! · URDF · ROS Control Designing and implementing robotic manipulation systems using ROS tools for task automation.

Autonomous Navigation ROS2 Nav2 · A* · D* Lite Developing algorithms for autonomous navigation and path planning in dynamic environments.

Sensor Fusion EKF · IMU · LiDAR Combining multiple sensor data streams to improve robot perception and localization accuracy.

Development Stack

Core Tools VSCode · Jupyter · Git · Python Using modern development tools to write, test, and version-control robotic code and algorithms.

Environment Management Conda Managing project dependencies and ensuring reproducibility across different platforms and teams.

Scientific Computing NumPy · SciPy · Matplotlib Applying scientific computing libraries to perform data analysis, modeling, and visualization in robotics.

Visualization and Simulation Tools Rviz · Gazebo · TF · Behaviour Trees Utilizing powerful simulation and visualization tools for robot modeling, testing, and debugging.

Intelligent Perception

Computer Vision OpenCV · YOLOv8 Leveraging computer vision techniques for real-time object detection and image processing.

3D Reconstruction Multi-view · SLAM Techniques for reconstructing 3D environments using stereo vision, Structure-from-Motion (SFM), and SLAM.

Deep Learning TensorFlow · PyTorch Building and training deep neural networks for tasks such as classification, regression, and segmentation.

Extended Capabilities

Hardware Integration Raspberry Pi · Mechanical Design · LeoRover Robot Assembly Designing and integrating hardware systems for robotics, including the use of Raspberry Pi and mechanical components for building robots like LeoRover

Prototyping Tools 3D Printing · Laser Cutting Utilizing advanced prototyping tools for creating custom parts and models, with expertise in 3D printing, laser cutting, and CNC milling for rapid prototyping.

Technical Documentation LaTeX · Github · Jupyter Writing comprehensive technical documentation using LaTeX for high-quality reports and papers, and managing projects on GitHub for version control. Also proficient in creating interactive Jupyter notebooks for analysis and presentation.