Aleksei Staroverov

Moscow Institute of Physics and Technology, Russian Academy of Sciences

Papers

Total Citations

H-Index

About

Aleksei Staroverov is a rising researcher in Embodied AI and robotics, focusing on how agents perceive, navigate, and interact with human-centric environments. His work bridges computer vision, reinforcement learning, and multimodal language models to create more robust robotic systems. Staroverov’s most cited paper, “HPointLoc” (11 citations), introduces a point-based indoor place recognition method using synthetic RGB-D images, advancing visual localization. He also developed “Hierarchical Landmark Policy Optimization” (9 citations), a novel approach to visual indoor navigation that improves an agent’s ability to find objects by semantic category through hierarchical reinforcement learning. In “Fine-Tuning Multimodal Transformer Models for Generating Actions” (8 citations), Staroverov proposed RozumFormer, a bimodal transformer that translates language instructions into robotic manipulation actions in both virtual and real environments. His work “Skill Fusion in Hybrid Robotic Framework” (8 citations) tackles visual object goal navigation by fusing exploration and navigation skills, enabling agents to efficiently locate and interact with target objects. With over 36 citations across his top papers, Staroverov is making meaningful contributions to making robots more capable and intuitive in everyday spaces.