Projects

Implemented a 2-approximation TSP solver using Minimum Spanning Tree + DFS, then refined with 2-opt and nearest-neighbour heuristics to minimize total tour cost.

Implemented Dijkstra and A* with differential constraints on a TurtleBot. Generated wheel RPMs alongside the path and visualised real traversal in VRep.

BFS-based puzzle solver that explores all reachable configurations until the goal state is reached, with custom input support and two complexity modes.

Used the Prometheus frontier exploration package in ROS to autonomously explore unknown environments. Developed with TDD across three one-week sprints.

PRM algorithm generates a roadmap for the static scene, then A* plans a time-optimal trajectory that avoids dynamic obstacles with known constant velocities.

Classic Dijkstra and A* implementations on a static map with configurable start/goal, comparing optimality and runtime between the two approaches.

Design and simulation of a Linear Quadratic Regulator (LQR) controller for a double-pendulum gantry crane system to achieve stable, energy-optimal control.

Full design cycle of an articulated fruit-picking robot: link-length design, forward & inverse kinematics derivation, and MATLAB simulation of the complete system.

Perspective transform + histogram-based lane pixel detection with polynomial fitting for turn prediction. Runs on front-facing camera footage.

Underwater buoy segmentation using Gaussian Mixture Models + Expectation-Maximization for robust colour thresholding under dynamic lighting. Detection via Hough Circles.

MSER-based sign detection followed by HOG + SVM classification achieving 97.4% accuracy across 8 commonly-used traffic signs.

Custom AR tag detection via edge/corner detection, perspective transform for ID recognition, and virtual cube superimposition using a computed projection matrix.

Differential tracker that warps frames to align with a first-frame template. Enhanced with LAB color space and histogram equalization for illumination invariance.