Research Interests
Robotic Perception, Mobile Manipulation, Human-Robot Interaction, Vision-Language-Action (VLA), Deep Learning
As a student researcher, Md Yasin Arafat’s research focuses on autonomous robotic systems that integrate visual perception, language grounding, and manipulation in dynamic environments.
Research Areas
His work investigates vision-based multi-object tracking and target re-identification using learning-based frameworks such as DeepSORT and transformer-based trackers, optimized with data association and contrastive loss functions to maintain identity under occlusions.
He studies language-grounded robotics through vision-language models (VLM), formulating semantic grounding as a cross-modal alignment problem optimized via contrastive objectives.
His research further addresses mobile manipulation by integrating object detection, grasp pose estimation, and motion planning within a state-space control framework where task execution is optimized by minimizing trajectory costs subject to kinematic and collision constraints.
His broader interests include perception-action coupling, multimodal sensor fusion, and real-time optimization for resource-constrained robotic platforms.