Ariel Rodriguez

Multimodal behavioral cloning fails in surprising ways: latent-variable policies can't balance regularization, while action-space generative policies are fundamentally limited by the smoothness of their action mappings.

LAR-MoE unlocks structured expert specialization in robotic imitation learning from unlabeled demonstrations, rivaling supervised methods without needing manual phase annotations.

Ditch the da Vinci: this open-source surgical robotics platform brings precision and flexibility to autonomous laparoscopic procedures using standard industrial robots.