Real-Time Robot Self-Modeling via Direct Sensorimotor Decoding: Overcoming the Neural Rendering Latency Bottleneck

Autonomous robot self-modeling-the ability of a robot to learn its own morphology from sensory data-is crucial for resilience in unstructured environments. While recent advances in neural rendering, particularly Neural Radiance Fields (NeRFs), have enabled high-fidelity 3D self-models, their computational latency (typically