Anand Bhattad, an assistant professor of computer science at the Johns Hopkins University—along with collaborators at the Toyota Technological Institute at Chicago (TTIC) and the Toyota Research Institute—took home the Best Paper Award in Robotics Learning at the 2026 IEEE International Conference on Robotics and Automation (ICRA), held June 1–5 in Vienna, Austria.
ICRA is the Institute of Electrical and Electronics Engineers Robotics and Automation Society’s flagship conference, gathering the world’s top researchers and industry leaders to share ideas, exchange knowledge, and advance the field of robotics for the benefit of humanity.
The team’s paper, “Do You Know Where Your Camera Is? View-Invariant Policy Learning with Camera Conditioning,” which was initiated while Bhattad served as a research assistant professor at TTIC, was recognized for proposing a novel approach to imitation learning by conditioning policies on camera extrinisics.
Robot policies are typically trained from a fixed camera viewpoint, causing them to fail when the camera moves or is repositioned. The team’s work improves viewpoint generalization by conditioning policies on camera position and orientation. To evaluate viewpoint robustness, the researchers also introduced six new robotic manipulation tasks that they released alongside their code and demonstrations.
“This paper is timely as there is a lot of debate on whether 3D vision is going to be useful for robotics,” says Bhattad. “Our research shows that it is.”
Bhattad’s research focuses on building physical intelligence in embodied agents, sitting at the intersection of computer vision, computer graphics, generative modeling, and physical reasoning. A main thread in his work is understanding what generative visual models understand (and don’t) about the physical world. He is further interested in how these models compare to human perception, and how that comparison can guide the design of intelligent systems grounded in physical principles.