D. A. Sasi Kiran

Indian Institute of Technology Hyderabad

Papers

Total Citations

H-Index

About

D. A. Sasi Kiran is a researcher advancing the field of embodied AI and robotic navigation, with a primary focus on **Object Goal Navigation (ObjectNav)**—a challenging task requiring a robot to locate and move to a specific object in an unseen environment. His work addresses the core challenge of enabling autonomous agents to build and leverage semantic understanding of their surroundings for efficient goal-directed exploration. In his 2022 paper on "Object Goal Navigation using Data Regularized Q-Learning," Kiran introduced a framework that incrementally constructs a semantic map of the environment, using a regularized reinforcement learning approach to select long-term goals. This work has garnered **8 citations** for its practical approach to improving navigation policy stability. In a complementary study on "Spatial Relation Graph and Graph Convolutional Network for Object Goal Navigation," he pioneered the use of graph neural networks to encode spatial relationships from robot trajectory history, allowing agents to learn structural patterns of object co-occurrence. This second paper, also with **8 citations**, demonstrates his ability to integrate geometric reasoning with deep learning. Together, these contributions provide a foundation for more intelligent, map-building navigation systems that move beyond reactive policies toward structured, relational reasoning in complex environments.