Tharindu Gimras

Integrates computer vision, AR, and cloud-based communication to create a real-time 3D map of a robot's surroundings for an immersive indoor robot control experience, with 1 cm accuracy depth estimation using Microsoft Kinect V2 and semantic segmentation.

A wheel-configurable modular research platform for electric wheelchair caster wheel motion dynamics analysis, supporting 212 wheel configurations across three major wheelchair types, recognized with the 🏆 Best Poster Award at ICIPRoB 2026.

A adaptive navigation framework for transformer robots in warehouse environments, integrating overhead camera perception, binary segmentation, and a shape-aware A* algorithm to determine optimal robot configurations through narrow passages.

Migrated personal portfolio from ViteJS to NextJS with Three.js 3D elements, improving speed index from 2.737s to 1.260s and increasing SEO score from 49% to 53%.

Developed a 3D portfolio website showcasing web development skills using React, Three.js, and ViteJS.

Testing a 6-DOF Kinova Mico robot arm mounted on a Jazzy Air wheelchair platform, running a predefined looping motion that mimics bringing food to the mouth of a seated patient. No one was actually sitting in the chair. It's just the arm cycling through the motion on repeat. Built and run purely for fun, not a formal research project.

A 6-DOF robot arm simulated in MuJoCo. Forward and inverse kinematics are written from scratch, not pulled from a library. Forward kinematics uses Rodrigues' rotation formula chained into homogeneous transforms. Inverse kinematics uses a damped least-squares Jacobian solver. Both are checked against MuJoCo's own physics engine and match to floating-point precision. The same math runs live below, in the browser. Drag the target and watch the joint angles solve in real time.