R&D Engineer specializing in algorithm development for robotics, automation, and high-precision systems. My work focuses on the intersection of Machine Learning, Image/Signal Processing, and Control Systems to create intelligent industrial solutions. I have a passion for the full development lifecycle, from hardware prototyping and data acquisition to model implementation and performance validation.
This repository contains the work for my M.Sc. research on classifying deformable objects by "touch" where vision alone is insufficient.
- Objective: To design a system that can differentiate objects (e.g., ripe vs. unripe fruit) based on their physical properties.
- Method:
- Designed and built a testbed with a linear motor and force-controlled data acquisition system (C++, Sensoray DAQ).
- Collected time-series data (force, position, current) by deforming various objects.
- Engineered features from raw sensor signals and applied Time-Series Forest Classifiers (Python,
sktime).
- Result: Achieved >94% classification accuracy, demonstrating a robust method for sensorless force estimation and haptic-based object identification.
Haptic Testbed in Action | Data acquisition
Repository Link | Data Processing Scripts
- Objective: To improve the accuracy of a high-precision gantry system by compensating for thermal expansion and mapping 2D errors.
- Tech: Python, Halcon, OpenCV, PySide6, ACS SPiiPlus.
- LinkedIn: linkedin.com/in/malith-jayawardhana
- Email: malithjayawardhana@gmail.com