Research and personal projects spanning robotics, machine learning, and embedded systems.
A ZED 2i-based markerless upper-body teleoperation framework covering head, arm, elbow, and hip joints, targeting multiple robot platforms including SoftBank Pepper, NAO, Agibot X2, and Unitree, across both NAOqi and ROS2 middleware. Agibot X2 integration uses RViz2 with robot_state_publisher for simulation. Forms the basis of a conference paper in development for ICRA 2027.
My core PhD project, focused on adaptive robotic systems that perceive context and dynamically adjust their social presence in human-robot interactions. Investigates multimodal context understanding and behaviour modulation, enabling robots to self-regulate their expressiveness, responsiveness, and engagement across different situations. Includes a modular imitation learning (GAIL-based) pipeline for learning social behaviours from paired human skeletal data.
Honours research project detecting driver drowsiness through eye closure and yawning detection. Facial regions were identified using a Haar cascade classifier, with facial landmarks located via SVM and HOG. Eye closure was evaluated through eye aspect ratio, and yawning through the distance between upper and lower lips. Achieved 96 percent accuracy in field testing across diverse scenarios.
A human-faced robot designed to mimic facial expressions and mirror a person's facial cues, developed as an interactive companion for kindergarten children, with a focus on helping introverted kids engage more actively.
A smart dress that responds dynamically to sound and voice for live performances. Built with Arduino and a sound sensor, the dress detects sound frequencies and triggers embedded LED lights, illuminating during high-frequency or vocal segments and dimming during quieter moments, to create a synchronised light show throughout a performance.