Vedant Dave
Papers
3
Total Citations
32
H-Index
2
About
Vedant Dave is a rising researcher at the forefront of multimodal learning for robotics, with a focus on integrating vision and touch to create more intelligent, sample-efficient autonomous systems. His work addresses a critical bottleneck in robotic manipulation: how to fuse visual and tactile sensory data to enable robust interaction with complex, tangible objects. Dave’s major contributions include pioneering self-supervised contrastive pre-training for multimodal visual-tactile representation learning, a method that allows robots to build richer, more generalizable object models without extensive labeled data. His paper on this topic has already garnered 22 citations since 2024, underscoring its immediate impact. He further advanced the field with M2CURL, a framework that leverages self-supervised representations to dramatically improve the sample efficiency of multimodal reinforcement learning for robotic manipulation, earning 8 citations. Most recently, Dave introduced EnvoDat, a large-scale multisensory dataset designed to benchmark robotic spatial awareness and semantic reasoning in heterogeneous environments—a crucial step beyond the urban-centric datasets that dominate the field. Through these contributions, Dave is shaping how robots perceive and act in the real world, making him a key voice in the next generation of embodied AI research.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3