Core Technology

Utilize compact optical path optimized design to significantly reduce sensor volume, enabling enhanced compatibility, allowing seamless integration into dexterous hands and complex mechanical structures.

Employ deep learning and multi-sensor fusion technology to achieve high-precision depth calibration without fiducial marker dependency, equipped with multimodal capabilities including 3D shape perception, 6D distributed force detection, and slip perception, while optimize algorithm architecture to reduce computational load for efficient tactile feedback.

Implement self-developed high-precision calibration algorithm, which can automatically capture sensor images, encode data into high-dimensional features via embedded encoders, and compute calibration parameters based on feature extraction, achieving out-of-factory consistency control, and enabling real-time error compensation to ensure long-term operational stability.

Leverage advanced optical and mechanical simulation models to generate high-fidelity tactile data, supporting AI training for robots and enhancing generalization capability of tactile models.

Integrate vision and tactile information through a physics-informed transformer framework, decoupling high-level scene reasoning from fine-grained interaction control. By leveraging imitation learning and reinforcement learning, establish a perception-to-control closed loop,enabling robots to generalize precise operational skills to various environments.