Raviteja Upadrashta
Papers
3
Total Citations
17
H-Index
3
About
Raviteja Upadrashta is a robotics researcher focused on bridging the gap between natural human communication and autonomous robot manipulation. His work centers on enabling robots to understand and execute tasks from natural language instructions, a critical step toward intuitive human-robot collaboration. His most influential paper, "Translating Natural Language Instructions to Computer Programs for Robot Manipulation" (2021, 9 citations), proposes a novel approach where language-conditioned models predict actuator commands by first translating instructions into executable computer programs, rather than directly mapping text to actions. This work addresses the long-standing challenge of grounding abstract language in physical robot behavior. In "Teaching Robots Novel Objects by Pointing at Them" (2020, 5 citations), Upadrashta introduces an end-to-end neural network that allows robots to learn about new objects simply through human pointing gestures, enabling rapid knowledge acquisition in unstructured environments. His research on "Spatial Reasoning from Natural Language Instructions for Robot Manipulation" (2021, 3 citations) further advances the field by developing a pipelined architecture that localizes objects in a scene and performs spatial reasoning from text, allowing robots to understand relational commands like "place the cup to the left of the plate." Upadrashta’s contributions are paving the way for more adaptable, teachable robots that can work alongside humans in dynamic settings.
Research Focus
Key Achievements
Top Papers
- 1
- 2Teaching Robots Novel Objects by Pointing at Them5 citations · 2020
- 3