Building Embodied AI Systems: The Agents, the Architectural Principles, Challenges and Application Domains

The topic of embodied AI is experiencing tremendous growth due to recent advancements in computer vision, machine learning, and artificial intelligence. Intel Labs and Facebook AI Research (FAIR) have been leading new initiatives in the field of embodied AI. The definition of “embodied” is “giving an idea a tangible or visible form.” “Embodied AI” is just “AI for virtual robots,” to put it simply. More precisely, Embodied AI is the area of study that deals with AI challenges for virtual robots that can move, see, talk, and communicate with other virtual robots. These simulated robots are subsequently applied to real-world robots as solutions. AI subfields have been largely divided due to a variety of constraints. Embodied AI, on the other hand, unites a number of interdisciplinary disciplines, including robotics, computer vision, reinforcement learning, navigation, physics-based simulations and Natural Language Processing (NLP). Computer Vision techniques have played a major role in the evolution of Embodied AI as a research area, even if it needs to succeed in numerous AI subfields.