On-board computing cluster for complex measurements
The product is in development process
Applications
Designed to solve distributed computing tasks in the field of artificial vision in the robotic complex motion control circuit.
- Motion parameters calculation according to the data obtained from artificial vision
- Local and relative navigation management
- Global navigation management, allocation using a known digital vector terrain map
- Objects recognition and classification in the video data using neural network algorithms
Major application properties
- Scalable software and hardware architecture allowing to increase an accuracy and frequency of real-time solution
- Use of built-in graphics coprocessors for optimization of neural network algorithms and processing of artificial vision data
- 2 to 16 computing modules
- Network cluster replication tools based on the PCI-Express interface
Key features and properties
- Modular pipeline software architecture development of technical vision algorithms
- Migh-speed data exchange between modules due to PCI-Express interface in non-transparent bridge mode (NTB mode)
Specifications
Computing machine data |
|
Base module type |
ARM Cortex-A57 + NVIDIA Maxwell (Jetson TX-1) ARM Cortex-A57 + NVIDIA Pascal (Jetson TX-2) |
Number of computing units per module |
2 - 14 |
RAM |
Up to 8 GB/1866 MHz per module |
Interfaces supported |
CAN RS-485 Ethernet 10/100/1000 PCI-Express (внутриприборный)
|
Power Specifications |
|
Power supply voltage range |
9–36 V |
Per module power consumption, up to |
120 W |
Environmental resistance |
|
Operating temperatures range |
- 40 to +50 °C |
Relative humidity |
6 to 98% |
Pressure |
0 to 820 mm Hg |
Weight and dimensions |
|
Weight |
4500* g |
Dimensions |
270х202х140* mm |
Reliability specifications |
|
MTBF |
10000 hours |
Life time |
10 years |
*8 modules configuration