Researchers developed a cross-embodiment imitation learning system for quadrupedal robots, combining human and LocoMan data through a unified teleoperation pipeline and modular architecture. They introduced the first LocoMan manipulation dataset and validated the approach on six tasks, achieving a 41.9% overall and 79.7% OOD performance improvement. Pretraining with human data further boosted success rates, enabling strong results with only half the robot data. Video Credit: Yaru Niu #robotics #

Carcass - Embodiment

I don’t believe this video got the attention it truly deserved. Honestly, the amount of tech packed in here will help you build XR experiences quickly with the powerful tools provided by Unity’s AR Foundation and the Unity OpenXR Meta Plugin. 💡But does this use Meta’s SDK? It doesn’t, OpenXR is used in this case and Unity uses Meta’s extension found within Khronos Group OpenXR Specification. 📣 Here are some of the extensions included with OpenXR and abstracted through AR Foundation components: