Highly Programmable Haptic Decoding and Self-Adaptive Spatiotemporal Feedback Toward Embodied Intelligence

Abstract

Intelligent robots, equipped with perception, cognition, and learning capabilities, are transforming the manner by which complex tasks are approached, enhancing autonomy, efficiency, and adaptability. By contrast, conventional robotics typically struggle with precision and reliability in tasks such as grasping and recognition owing to their limited sensing and feedback mechanisms. To achieve advanced applications, robots require sophisticated spatiotemporal feedback to adjust their actions dynamically, which poses a significant challenge to the pressure-decoupling capability. Herein, a high-performance programmable and event-driven (PED) haptic interface with real-time, self-regulating spatiotemporal feedback, empowering robots with dynamic grasping adaptation and force optimization is introduced. The PED haptic interface features a gradient pyramid metasurface-like structure, which emulates the perception of the human skin to decode tactile data. Compared with conventional devices, the PED haptic interface offers significant improvements in the detection range and sensitivity by 300% and 350%, respectively. By integrating cutting-edge haptic sensing and feedback technology with artificial intelligence, a conceptualized intelligent agent is developed that autonomously understands unstructured environments to avoid self-damage or object damage without external intervention. This breakthrough not only offers new research avenues but also significantly advances the research foundation of embodied intelligence, particularly in simulating human perception and cognition.