나는엔지니어다
This page was machine-translated and may differ from the original. View original

Aetina Launches Edge AI System 'AIE-KT78/68' for Robots

Google 우선 소스Published2026.09.21 10:29


 
Based on NVIDIA Jetson Thor, Maximum 2070 FP4 TFLOPS Targeting Collaborative and Humanoid Robots
 
As on-device AI computing demand in robotic systems increases, Aetina Corporation has introduced new edge AI system products based on NVIDIA Jetson Thor. Designed primarily for collaborative robots (Cobots) and humanoid robots, this product integrates high-performance AI computing, multi-sensor connectivity, and industrial real-time control capabilities into a single system.
 
On the 21st, Aetina officially launched the flagship edge AI systems 'DeviceEdge AIE-KT78' and 'DeviceEdge AIE-KT68', and announced that orders and shipments are immediately available.

Both products adopt NVIDIA Blackwell architecture and can run multimodal generative AI, large language models (LLM), vision-language models (VLM), and vision-language-action (VLA) models on-device without cloud connectivity, the company explained.
 
AIE-KT78 and AIE-KT68 Specifications Comparison
 
The AIE-KT78 features the NVIDIA Jetson T5000 module and provides 128GB 256-bit LPDDR5X memory and maximum 2,070 FP4 teraflops (TFLOPS) of AI computing performance.

The AIE-KT68 is based on the NVIDIA Jetson T4000 module and supports 64GB 256-bit LPDDR5X memory and maximum 1,200 FP4 TFLOPS.

Both products are designed with consideration for balance among performance, power consumption, and system cost, Aetina stated.
 
Sensor Connectivity and Real-Time Control Architecture
 
The AIE-KT78/68 simultaneously supports QSFP28 connectivity (maximum 4x25Gbps) and dual RJ45 10GbE ports, enabling multi-viewpoint perception through maximum 8 GMSL2 camera inputs.

It can process data from various industrial sensors at high speed, including 3D LiDAR, radar, depth sensing cameras, and inertial measurement units (IMU).

A dedicated RJ45 1GbE EtherCAT port operates as an independent EtherCAT master, controlling motors, joints, and actuators with microsecond-level synchronization.
 
On the hardware side, the compact form factor with 80mm thickness supports 9–48VDC input voltage and −25°C to +55°C operating temperature.

It features USB 3.2, isolated digital I/O, and M.2 expansion slots, and supports the complete NVIDIA AI software stack including Linux-based NVIDIA JetPack 7, CUDA, TensorRT, DeepStream, Holoscan, and Isaac ROS.
 
Richard Hung, Vice President of Product Business at Aetina, stated, "The competitiveness of collaborative robots and humanoid robots is shifting from single-model computing performance to system capabilities that can integrate perception, inference, decision-making, and action in real-time in actual environments. We provide end-to-end technical support to help developers and system integrators (SI) reduce total cost of ownership (TCO) and shorten the timeframe from proof of concept (PoC) to large-scale deployment."
To request a correction, reply or follow-up report on this article, see how to file a request. Previously published statements are collected in corrections & replies.
배종인 기자
배종인 Reporter

Comments