Neurorobotics studies how robots can be controlled using biological neural systems or computational models inspired by them. RNN-based approaches have played a central role in modeling sensorimotor prediction, imitation, language–action integration, and object manipulation. Extensions such as RNNPB enable the learning and recognition of multiple behavioral patterns through parametric representations, while MTRNN introduces temporal hierarchies that self-organize functional action primitives and higher-level structures. These models support adaptive robot behavior, generalization, and real-time control without explicit physical modeling. Beyond robotics, neurorobotic frameworks have also been applied to computational psychiatry, offering mechanistic accounts of disorders such as schizophrenia through disruptions in hierarchical neural dynamics.