Embedded AI

Minimization of algorithm and system-optimized technology to apply AI technology to robots and to enable inferences to be made in embedded environments with limited resources.

Model Optimization

Current AI technology research is being conducted via learning and analysis in a cloud-based system or in a high-spec GPU environment. As the size of the AI model increases and becomes more difficult to operate in real-time using an embedded controller, optimization of developed AI models using minimization and compression technology is a necessity in order to operate robots in limited environments.

Through our latest research on model minimization, Robotics LAB is developing unique optimization technology for maintaining the accuracy of the model, minimizing power consumption, and accelerating computation.

Embedded AI System Development

Numerous AI algorithms are installed on robots and serviced. As the service becomes more advanced, operating Multi-AI algorithms with a single controller becomes increasingly difficult. To solve this, an H/W Accelerator is being used for the development of NPU, VPU, and other AI algorithms as they play a part in developing the optimal Embedded AI system. Furthermore, Agent S/W will be developed to efficiently control the Embedded AI system, and research is underway regarding distribution deduction under embedded environments and for real-time multi-tasking.