Rajasimman Madhivanan

Papers

Total Citations

H-Index

About

Rajasimman Madhivanan is a rising researcher in embodied AI and robotic perception, whose work bridges the critical gap between how robots see and how they move. His primary research areas include active perception, object navigation, and vision-language models for robotics. Madhivanan’s most cited work, “Learning to View: Decision Transformers for Active Object Detection” (2023, 16 citations), introduces a novel framework that unifies planning and perception, enabling robots to autonomously reposition themselves to gather more informative visual data—a departure from traditional decoupled systems. He further advances the field with “VLPG-Nav” (2024), which tackles the practical challenge of not just reaching an object but centering it within the robot’s camera view, and “LOC-ZSON” (2024), which pioneers a language-driven, object-centric representation for zero-shot navigation in complex, cluttered scenes. By fine-tuning visual-language models with specialized losses, his work enables robots to handle ambiguous, object-level instructions without prior training. With a growing citation footprint and a focus on making robotic systems more adaptive and context-aware, Madhivanan is contributing to the next generation of intelligent, perception-driven autonomous agents.