Specifications

Manufacturer

Advantech
Taiwan

Module

NVIDIA Jetson Thor
T5000 / T4000

AI Performance

Up to 2070 TFLOPS FP4
inference

CPU

14-core Arm Neoverse
(T5000)

Memory

128GB LPDDR5X (T5000)
64GB LPDDR5X (T4000)

Networking

6× GbE
+ 1× SFP28 10GbE

USB

4× USB

Serial

2× COM

I/O

DIO · CAN FD

Sensor I/O

Cameras (GMSL) · LiDAR · IMU

Software

JetPack 7.1 · Ubuntu 24.04
Advantech Robotic Suite · Isaac ROS

Power

20–28V DC
ATX / AT mode

Operating Temp

-10°C ~ 50°C
(0.7 m/s airflow)

Dimensions

177 × 174 × 83 mm
1.93 kg (AFE-A702)

Mounting

Wall mounting

Certifications

IEC 61000-6-4
Heavy Industry

Form Factor Variants

ASR-A702

NVIDIA Jetson Thor AMR Control Board — board-level platform for deep custom integration into robot chassis

AFE-A702

NVIDIA Jetson Thor AMR Control System — enclosed, wall-mountable system for ready deployment

Overview

Advantech has expanded its Jetson Thor portfolio with the ASR-A702 and AFE-A702 — purpose-built robotic controllers for autonomous mobile robots (AMRs), humanoids, and unmanned vehicles. Available as a board-level platform (ASR-A702) or a fully enclosed system (AFE-A702), both are accelerated by the NVIDIA Jetson Thor T5000 or T4000 module, delivering up to 2070 TFLOPS of FP4 inference performance — more than 7.5× the AI compute of Jetson AGX Orin.

Dedicated robotics I/O includes six Gigabit Ethernet ports, a 10GbE SFP28 uplink for high-bandwidth sensor fusion, four USB ports, dual COM, digital I/O, and CAN FD — the interfaces required to connect cameras (GMSL), LiDAR, IMUs, and motor controllers in a single deterministic system. The controllers support hardware time synchronization, ESD protection, and anti-vibration construction for the demanding physical environment of mobile robotics.

Software support pairs JetPack 7.1 on Ubuntu 24.04 LTS with Advantech Robotic Suite and NVIDIA Isaac ROS, enabling GPU-accelerated SLAM, object detection, distance estimation, pose tracking, and VSLAM (visual simultaneous localization and mapping) out of the box. IEC 61000-6-4 Heavy Industry certification qualifies the platform for factory and logistics deployments.

Key Benefits

Thor-Class Edge Compute: Up to 2070 TFLOPS FP4 with 128GB LPDDR5X enables real-time AI reasoning, multi-model inference, and GPU-accelerated SLAM on the robot itself. Robot-Native I/O: 6× GbE + 10GbE SFP28, CAN FD, DIO, and multi-camera GMSL inputs consolidate perception, control, and networking in one unit. Two Form Factors: Board (ASR-A702) for OEM integration or enclosed system (AFE-A702) for immediate deployment. Hardware Time Sync: Precise sensor fusion across cameras, LiDAR, and IMU for reliable autonomous navigation. Production-Ready Software: JetPack 7.1 + Advantech Robotic Suite + Isaac ROS shorten humanoid and AMR development cycles.

Request a Quote — Advantech ASR-A702 / AFE-A702

QS Compute — global B2B supply for Advantech Jetson Thor robotic controllers. T5000 and T4000 configurations.

Get Your Quote →