Embodied Intelligence in Wearable Robotics

Abstract Embodied intelligence refers to the capacity of an agent to acquire autonomy and adaptivity through interactions between its body, controller, and environment, a paradigm originally developed in biological and robotic contexts. In this paper, we extend this notion to the domain of wearable robots, particularly soft exosuits and hybrid soft–rigid devices that augment or restore human movement. The proposed framework integrates morphology-aware generative design, reinforcement learning–driven controllers, and strategies for sim-to-real transfer informed by simulators. Human-inspired modelling of soft–rigid interactions and ergonomic interfaces is combined with adaptive learning to create systems that are not only mechanically efficient but also user-centred and comfortable. By integrating advances from soft robotics, biomechanics, and machine learning, this work proposes a pathway toward next-generation wearable robots that embody intelligence in their morphology, materials, and control policies. The perspective opens new opportunities for designing assistive and augmentative systems that co-adapt with human users.