AMD Demonstrates an End-to-End Embodied-AI Stack
An AMD-accelerated stack spans data-center training, simulation and Ryzen AI edge deployment for VLA manipulation.
An AMD-accelerated stack spans data-center training, simulation and Ryzen AI edge deployment for VLA manipulation.
A controlled study found prediction objectives determine whether robot world models retain mass, stiffness, drag, and force information.
Symmetry-based trajectory augmentation improved training speed and success rates on real robot insertion, routing, and relocation tasks.
Researchers introduced HERO, a hierarchical embodied agent that autonomously gathers experience and consolidates reusable manipulation policies.
A benchmark evaluates decision-level safety and reliability across multi-robot and aerial scenarios.
Researchers propose more granular evaluation of spatial reasoning needed for physical manipulation and navigation.
A systems framework links robust task execution, hardware safeguards, evidence and deployment governance.
Researchers map attacks across sensing, world modelling, planning, execution, feedback and long-term adaptation.
A lightweight episodic memory module enables a humanoid robot head to retain personalized conversational context across sessions.
An exploratory mobile-robot study reports spatial memory and action consequence reasoning from a general multimodal model.