Mohit Deshpande

Papers

Total Citations

H-Index

About

Mohit Deshpande is a rising researcher at the intersection of robotics, computer vision, and active perception, with a focus on enabling intelligent, resource-constrained robots to understand and navigate their environments. His work challenges the traditional decoupling of perception and planning, instead developing systems where a robot’s movement is deliberately chosen to improve its own understanding. In his highly-cited 2023 paper, "Learning to View: Decision Transformers for Active Object Detection," Deshpande introduces a novel framework that treats active perception as a sequential decision-making problem, using transformer architectures to learn optimal camera viewpoints for object detection. This work, garnering 16 citations, represents a significant step toward more efficient and autonomous robotic exploration. He further addresses the critical challenge of map stability in low-compute, narrow-field-of-view robots with his work on "Lighthouses and Global Graph Stabilization: Active SLAM," proposing a method to prevent catastrophic drift in simultaneous localization and mapping systems. By tackling the practical constraints of real-world hardware, Deshpande’s contributions are paving the way for more robust and perceptive autonomous systems, from search-and-rescue drones to household robots.