ROBOTCORE® RPU

ROBOTCORE®

The Robotic Processing Unit
specialized in ROS computations

ROBOTCORE® is a robot-specific processing unit (RPU) that helps map Robot Operating System (ROS) computational graphs to its CPUs, GPU and FPGA efficiently to obtain best performance. It empowers robots with the ability to react faster, consume less power, and deliver additional real-time capabilities.

Get ROBOTCORE® Build your own

A processing unit for the robotics architect

ROBOTCORE® features multiple CPUs, a GPU and an FPGA interconnected in a common Ethernet databus which allow combining the traditional control-driven approach used in robotics with a data-driven one. This means that when architected appropriately, through acceleration kernels (IP cores), ROBOTCORE® provides robotics architects with ROS 2 accelerated libraries that deliver faster computations (getting tasks done quickly once started), additional determinism (task happens in exactly the same timeframe, each time) and real time (meeting the time deadlines set for each task).

CPUs

Scalar Von-Neumann processors. Deliver conventional full computing control wherein a token of control indicates when a statement should be executed. Great at dealing with complex data and implementing custom control structures.

GPUs

A vector Von-Neumann processor that uses a token of control for executing a vectorized single instruction delivering domain-specific parallelism (e.g. image, video or math).

FPGAs

Flexible programmable processing and memory structures that provide eager evaluation: statements are executed as soon as data is available. Deliver best potential for parallelism, with high throughput and determinism.

Tech specs

Targeted
robot types

ROBOTCORE® is perfect to accelerate autonomous mobility, industrial manipulation and healthcare applications.

Autonomous Mobile Robots (AMRs)

Collaborative Robots (cobots)

Industrial Arms

Healthcare

Agriculture

Construction

Mining

Shuttle Vehicles

Automated Trains

Automated Bus

Automated Trucks

Automated Cars

Robotaxis

Autonomous Logistics

Packing best-of-class accelerators
from top silicon vendors

powered by the technologies from:

Developer-ready Tools
and IP Cores for robots

ROS 2 API-compatible hardware acceleration tools and robot Intellectual Property (IP) cores (IP cores). Improve your robot's latency, throughput and/or power efficiency by simply dropping them into your ROS workspace.

ROBOTCORE®
ROS 2

Accelerated ROS 2
network communications.
ROS 2 on FPGA.

Learn more

ROBOTCORE®
RTPS

Accelerated DDS
network communications.
DDS-RTPS on FPGA.

Learn more

ROBOTCORE®
UDP/IP

Accelerated UDP/IP
networking stack.
UDP/IP on FPGA. Learn more

ROBOTCORE®
Perception

Accelerated ROS 2
robotics perception
stack.

Learn more

ROBOTCORE®
Transform

Accelerated ROS 2
coordinate transformations
(tf2).

Learn more

ROBOTCORE®
Framework

Hardware acceleration
framework for ROS
and ROS 2.

Learn more

ROBOTCORE®
Cloud

Tools to speed-up ROS 2
graphs with the cloud,
and in the cloud.

Learn more

Why?

Robot brains and robot behaviors take the form of computational graphs, with data flowing between computation Nodes, across physical networks (communication buses) and while mapping to underlying sensors and actuators. The popular choice to build computational graphs for robots these days is the Robot Operating System (ROS), a framework for robot application development. Most companies building real robots today use ROS (or similar event-driven software frameworks, often in concert with ROS). ROS is thereby the common language in robotics with hundreds of companies and thousands of developers using it everyday. Since it's likely your team is already into ROS, so that you don't spend time reinventing the wheel and re-developing what already works, ROBOTCORE® focuses on accelerating ROS computations.

Which companies are using ROS? More about ROS

Tech specs

TARGETED
ROBOTS

Autonomous Mobile Robots (AMRs)

Industrial Arms

Agriculture

Mining

Automated Trains

Automated Trucks

Robotaxis

Collaborative Robots (cobots)

Healthcare

Construction

Shuttle Vehicles

Automated Bus

Automated Cars

Autonomous Logistics

HARDWARE
SPECS

CPUs


(group 1) 12-Core 64-bit Arm® Cortex®-A78 CPU 3MB L2 + 6MB L3 (CPU Max Freq 2.2 GHz)

(group 2) 4-Core 64-bit Arm® Cortex®-A53 (CPU Max Freq 1.3 GHz)

(group 2) 2-Core 32-bit Arm® Cortex-R5F real-time processor (CPU Max Freq 600MHz)

GPU

NVIDIA Ampere architecture with 2048 NVIDIA® CUDA® cores and 64 Tensor Cores (GPU Max Freq 1.3 GHz)

FPGA

256K System Logic Cells, 1248 DSPs, 26.6Mb on-chip memory (LUT: 117K, FF: 256K, DSP: 1248, BRAM: 144, URAM: 64)

Machine Learning throughput

275 TOPS

Memory

(group 1) 64GB 256-bit LPDDR5 (204.8 GB/s)

(group 2) 4GB 64-bit DDR4

Disk storage

(group 1) 64GB eMMC

(groups 1 and 2) SDHC card (external storage)

Thermal cooling

Active (Fan + Heatsink)

I/O

USB 2.0, SD/SDIO, UART, CAN 2.0B, I2C, SPI, GPIO, EtherCAT

High-speed I/O

PCIe® Gen2, USB3.0, SATA 3.1, DisplayPort, Gigabit Ethernet, 2x Time Sensitive Networking (TSN) Ethernet

Hardware synchronization (PTP)

sub-microsecond precision (<1 us)

Interconnect between group 1 and group 2

Ethernet-based

CPUs

(group 1) 12-Core 64-bit Arm® Cortex®-A78 CPU 3MB L2 + 6MB L3 (CPU Max Freq 2.2 GHz)

(group 2) 4-Core 64-bit Arm® Cortex®-A53 (CPU Max Freq 1.5 GHz)

(group 2) 2-Core 32-bit Arm® Cortex-R5F real-time processor (CPU Max Freq 600MHz)


GPU

NVIDIA Ampere architecture with 2048 NVIDIA® CUDA® cores and 64 Tensor Cores (GPU Max Freq 1.3 GHz)


FPGA

256K System Logic Cells, 1248 DSPs, 26.6Mb on-chip memory (LUT: 117K, FF: 256K, DSP: 1248, BRAM: 144, URAM: 64)


Machine Learning throughput

275 TOPS


Memory

(group 1) 64GB 256-bit LPDDR5 (204.8 GB/s)

(group 2) 4GB 64-bit DDR4


Disk storage

(group 1) 64GB eMMC

(groups 1 and 2) SDHC card (external storage)


Thermal cooling

Active (Fan + Heatsink)


I/O

USB 2.0, SD/SDIO, UART, CAN 2.0B, I2C, SPI, GPIO, EtherCAT


High-speed I/O

PCIe® Gen2, USB3.0, SATA 3.1, DisplayPort, Gigabit Ethernet, 2x Time Sensitive Networking (TSN) Ethernet


Hardware synchronization (PTP)

sub-microsecond precision (<1 us)


Interconnect between group 1 and group 2

Ethernet-based


SOFTWARE
SPECS

Operating System

Ubuntu Linux, Yocto 3.4 Honister (on demand)

Robot Operating System (ROS)

ROS 2 Humble Hawksbill

Communication middleware

Data Distribution Service (DDS)

Operating System

Ubuntu Linux, Yocto 3.4 Honister (on demand)


Robot Operating System (ROS)

ROS 2 Humble Hawksbill


Communication middleware

Data Distribution Service (DDS)


MECHANICAL
SPECS

Weight

Materials

Colour

Dimensions

2.2 kg

Aluminium

Silver

145 x 145 x 147 mm



Weight

2.2 kg

Materials

Aluminium

Colour

Orange or Silver

Dimensions

145 x 145 x 147 mm

POWER
SPECS

Power input

Power

Min. power

Max. power

9-20 V (DC power jack)

20 W

5 W (only group 2 enabled)

75 W

Power input

9-20 V (DC power jack)


Power

20 W


Min. power

5 W (only group 2 enabled)


Max. power

75 W


TOOLS AND
IP CORES

(offered separately)

ROBOTCORE® Framework (tool)

Hardware acceleration framework for ROS and ROS 2.

ROBOTCORE® Cloud (tool)

Tools to speed-up ROS 2 graphs with the cloud, and in the cloud.

ROBOTCORE® Perception (IP core)

Accelerated ROS 2 robotics perception stack.

ROBOTCORE® Transform (IP core)

Accelerated ROS 2 coordinate transformations (tf2).

ROBOTCORE® Framework (tool)

Hardware acceleration framework for ROS and ROS 2.

ROBOTCORE® Cloud (tool)

Tools to speed-up ROS 2 graphs with the cloud, and in the cloud.

ROBOTCORE® Perception (IP core)

Accelerated ROS 2 robotics perception stack.

ROBOTCORE® Transform (IP core)

Accelerated ROS 2 coordinate transformations (tf2).

Do you have any questions?

Get in touch with our team.

Let's talk Case studies