The systems are geared towards robotics applications, including collaborative robots (cobots), humanoid robots, and autonomous machines. Their design allows for the local processing of data from various sensors and the execution of artificial intelligence models for tasks such as perception, environmental analysis, and decision-making.
The AIE-KT78 incorporates the NVIDIA Jetson T5000 module, with 128 GB of 256-bit LPDDR5X memory and AI performance of up to 2,070 FP4 TFLOPS. Meanwhile, the AIE-KT68 uses the NVIDIA Jetson T4000 module, equipped with 64 GB of 256-bit LPDDR5X memory and performance of up to 1,200 FP4 TFLOPS.
Both models can locally run multimodal generative AI applications, large language models (LLM), vision and language models (VLM), and vision-language-action (VLA) models. This on-device processing reduces reliance on cloud services for applications requiring real-time responses or local data processing.
The platform is also designed to simultaneously manage different data streams from high-resolution sensors and coordinate AI processing with the control of motors, joints, and other actuators. These features meet the requirements of robotic systems that combine environmental perception, multimodal models, and real-time control.
