Self Model

Self-modeling, the ability of an agent (robot or biological system) to create an internal representation of its own physical structure and dynamics, is a burgeoning area of research aiming to enhance autonomy and adaptability. Current work focuses on developing self-models using various neural network architectures, including deep reinforcement learning and neural fields, often trained on proprioceptive or visual data to predict robot configurations or full-body movements. These self-models are proving valuable for tasks such as motion planning, control, and damage detection in robotics, while also offering insights into the biological mechanisms underlying self-awareness and agency. The development of accurate and efficient self-models is crucial for creating more robust and adaptable intelligent systems.