Ria Doshi
Papers
2
Total Citations
19
H-Index
2
About
Ria Doshi is a rising star in robotics, whose work is redefining how machines learn to interact with the physical world. Her research centers on two ambitious frontiers: achieving dexterous, real-world manipulation and creating truly generalist robot policies. In her highly cited 2023 paper, "Dexterous Manipulation from Images," Doshi pioneered a reinforcement learning approach that uses "substep guidance" to teach multi-fingered hands complex, contact-rich tasks like manipulating underactuated objects. This work, garnering 13 citations, directly tackles one of robotics' hardest challenges—moving from simulation to robust real-world performance. Expanding her vision, her 2024 paper "Scaling Cross-Embodied Learning" introduces a single policy architecture that unifies control across radically different robot morphologies, from manipulators to drones. With 6 citations and growing, this work suggests a future where a single AI model can pilot a quadcopter, navigate a legged robot, and grasp with a robotic arm. Doshi’s contributions are not just incremental; they are foundational steps toward a future of adaptable, general-purpose robotic intelligence.
Research Focus
Key Achievements
Top Papers
- 1
- 2