Likith Kumar Reddy Yammanuru

Background

I am a PhD Researcher in AI and Robotics at Cranfield University. My doctoral research focuses on the strategic integration and interpretability of Vision-Language-Action (VLA) models for high-mix industrial warehouse automation. My work investigates the critical trade-off between semantic generalization and operational reliability, building concrete evaluation frameworks to analyze model latency, task understanding, and hardware precision. Prior to starting my PhD, I earned my MSc in Robotics with a First-Class equivalent (Distinction) from Cranfield University. This academic background provided me with deep expertise in advanced motion planning, reinforcement learning, adaptive control, and machine vision. During my master's studies, I collaborated on complex robotics engineering projects, including a simulation initiative with AIRBUS and the European Space Agency (ESA) focused on multi-arm robotic collaboration for in-orbit assembly. I also successfully showcased this co-developed research as a technical poster presentation at the IEEE RAS Conference. Before transitioning into full-time doctoral research, my professional career included working as a Robotics Engineer in autonomous agricultural systems, where I developed ROS2-based computer vision and sensor fusion pipelines (LiDAR, cameras, IMUs) for real-world field robots. I also have professional experience as a Data Scientist engineering machine learning recommendation algorithms, and as a Software Engineer architecting automated OCR data pipelines to optimize industrial operations.

Research opportunities

Vision-Language-Action (VLA) models, Robot Learning, Explainable AI (XAI) in robotics, autonomous warehouse automation, semantic task understanding, and real-world robotic manipulation tracking.

Current activities

Developing an interpretability framework for Vision-Language-Action models. Instead of treating these AI models like a "black box," my work is dedicated to developing frameworks that show exactly how the models make decisions during tasks. My day-to-day work involves running continuous computer simulations to collect metric data, testing these models in the lab at the Centre for Robotics and Assembly, and validating their real-world pick-and-place accuracy on physical hardware. I am structuring my PhD by writing up my results chapter-by-chapter as the data drops, starting with a systematic literature review in my first year.