Astute System’s new computer designed for size, weight and power sensitive applications

The machine is designed for defence and aerospace applications, but can also be used in mining and earthmoving tasks

Nvidia Jetson powers Astute System's new GXA-1’s high-performance artificial intelligence (AI) computer design has been made for defense and aerospace but is equally as applicable for mining and earthmoving autonomous applications. The machine is optimised for deep learning tasks, as it is equipped with a 2048-core Nvidia Ampere architecture GPU.

The GXA-1 can deliver up to 248 Tflops, due to its 64 tensor core and dual deep learning accelerators (DLA). In a market where size, weight and power (SWaP) are critical factors, the device is an ideal solution.

Although being designed for defence and aerospace markets, the GXA-1 is suitable for advanced autonomous mining and earth-moving operations due to its SWaP-sensitive design.

Real-time processing for obstacle detection, path planning and equipment coordination can be achieved in haul trucks with the GXA-1’s installation. This is due to its ability to handle large-scale 360° situational awareness along with sensor fusion. These functions are essential for earth-moving equipment, creating a more efficient operation in complex and dynamic environments.

Equipped with an M.2 M key SSD ports, the GXA-1 can handle massive data sets. This is because its onboard hardware can provide ample bulk storage for high-resolution imaging and sensor data. These technologies are required for real-time terrain mapping and ore body analysis.

A performance level of 274 Tflops is provided by a 12-core Arm Cortex -A78AE v8.2 64-bit CPU with a 2.2GHz maximum frequency. This is paired with a memory of 64GB of 256-bit LPDDR5 memory, capable of 204.6GB/s bandwidth along with 32GB of eMMC5.1 storage. This grants the device a rapid data processing and transfer ability.

 Autonomous vehicle systems are supported by high-resolution cameras used in vehicle systems, along with robust 10 Gigabit Ethernet data plane ports for aggregating these cameras. These are accompanied by 1 Gigabit Ethernet control/data plane port, single SL-DVI output, dual CAN bus ports, triple USB 3.2 Gen2 (10Gbps) interfaces and quad analogue SD video inputs. These combine to allow for seamless integration across sensors and control systems.

The GXA-1 is supported by Nvidia Jetpack SDK and Linux for Tegra, along with Astute Systems MediaX for real-time video distribution.

Share This Article
Leave a Comment