◷ ⁽¹⁰⁶⁾

Vision for action is the use of visual information specifically to guide movement, interaction, and physical behavior in embodied AI systems. Unlike traditional computer vision, which focuses mainly on identifying or labeling objects, vision for action focuses on helping robots decide: In physical AI, vision…

Multi-agent environments involve multiple embodied AI systems interacting with each other while sharing the same physical or social space. These environments may include: Instead of operating alone, embodied systems in multi-agent environments must continuously: Think about how humans naturally move through crowded spaces. People avoid…

Navigation and mapping allow embodied AI systems to move purposefully through environments while building and updating internal spatial representations of the world around them. These capabilities help robots: In physical AI, navigation is far more than simply moving from point A to point B. Embodied…

Object manipulation is the ability of a physical AI system to detect, grasp, move, rotate, use, and interact with objects in the real world. It is one of the most important capabilities in embodied robotics because many useful real-world tasks depend on handling physical objects…

Physics interaction refers to how embodied AI systems perceive, predict, and respond to the physical properties and dynamics of the real world. This includes understanding and interacting with: Physical AI systems must constantly deal with these dynamics while moving, manipulating objects, and interacting safely with…