May 04, 2017      

Just as PCs and smartphones have become more powerful at a lower cost, next-generation robots are destined for the same path. Increased processing power will be a main driver in mass adoption of robotics, and chip giant Intel Corp. is hoping to be at the center of that movement with Intel Joule.

The Intel Joule compute module is the company’s most advanced development kit designed to simplify prototyping and commercialization of drones and robots. It does this by shortening the go-to-market time and lowering development costs, according to the company.

Intel Joule can handle everything from computer vision to machine learning, and it supports Intel RealSense cameras, which enable robots to collect and analyze data from their environment.

There are two Joule development kits (see comparison chart below) to choose from: the 550x ($319) and 570x ($369). Both models are very similar, but the 570x has more RAM and a faster CPU. Intel Joule supports multiple operating systems — including Reference Linux OS (default), Ubuntu Core, and Microsoft Windows 10 IoT Core — to ensure a flexible platform.

Intel Joule comes with an expansion board that packs the following impressive specs to make prototyping drones and robots even more plug-and-play ready:

  • USB 3.0 (1 Type-A and 1 Type-C port)
  • Micro HDMI
  • Micro-SD
  • 2x 40-pin I/O connectors at 3.3V
  • On-board port for Serial-over-USB debugging
  • 2 WiFi antennas
  • RTC battery connector
  • Fan header

A glance at Joule’s price might make you think a less expensive platform, such as Beaglebone or Raspberry Pi, is a better option. But don’t be fooled: Joule is in a different class than the aforementioned boards that target simpler applications for DIYers. Intel Joule targets developers commercializing products that require serious computing power.

Intel Joule 550x vs. Intel Joule 570x

Model Intel Joule 550x Developer Kit Intel Joule 570x Developer Kit
Price $319 $369
Board Form Factor 24mm x 48mm 24mm x 48mm
Socket 2-100 pin Hirose 2-100 pin Hirose
Embedded Storage 8 GB 16 GB
DC Input Voltage Supported 3.6-5.25V 3.6-5.25V
Pre-Installed OS Linux Kernel 4.4 optimized for IOT Linux Kernel 4.4 optimized for IOT
Max Memory Size 3 GB 4 GB
Memory Type LPDDR4 LPDDR4
Max # of Memory Channels 4 4
Max Memory Bandwidth 25.6 GB/s 25.6 GB/s
Integrated Graphics Yes Yes
Graphics Output Intel HD Graphics: HDMI 1.4B Intel HD Graphics: HDMI 1.4B
USB Ports 2 2
USB 2.0 Configuration (External + Internal) 2 2
USB 3.0 Configuration (External + Internal) 1+1 1+1
Integrated WiFi 802.11ac 2x MIMO 802.11ac 2x MIMO

Intel Joule and Intel RealSense simplify computer vision

Computer vision has been a major challenge when developing autonomous robots. Laser, radar, and IR sensors provide enough accuracy, but they are cost-prohibitive for most applications.

Enter Intel’s RealSense technology, a low-cost, high-performance computer vision system that enables robots to detect and avoid objects as they navigate an environment. Intel Joule supports Intel RealSense cameras, of course, and it can process the real-time data. This makes adding computer vision to Joule-powered drones and robots more seamless than systems using components from various companies.

Savioke, for example, has used RealSense cameras in its Relay delivery robots since 2015, thanks to a wrapper for the open-source Robot Operating System (ROS).

“Intel RealSense cameras bring great low-cost depth sensing to robotics, in a platform that is widely available and easy to integrate using ROS,” said Steve Cousins, CEO of Savioke.

Intel Joule can help lower development costs

Intel Joule, about as wide as a quarter, is loaded with specs found on PCs: 64-bit quad-core Atom processor, 4K graphics, 802.11ac connectivity, and DDR4 memory features.

Joule is pre-certified for distribution in more than 80 countries, saving developers’ time by allowing them to focus more on product development rather than pursuing various certifications for their boards.

Intel Joule 570x

Intel Joule 570x with expansion board. (Credit: Intel)

Andrew Dresner, principal engineer at Interbotix Labs, built the new Joule-powered TurtleBot 2i Research Robot with the Open Source Robotics Foundation. Two previous TurtleBot models relied on netbooks or laptops for computing power. But those options didn’t suffice for the advanced performance and low cost the TurtleBot 2i’s developer had in mind.

Choosing Intel Joule shaved $500 off the TurtleBot 2i’s computing cost.

“Before Intel Joule came into play, we were looking at multiple high power NUCs, or even multiple ARM SBCs — one to run a 3D camera and one to run the physical hardware,” Dresner said. “Intel joule gave us enough computational power to not have to split the load over multiple boards.”

The Joule also enabled the TurtleBot 2i designers to easily integrate two Intel RealSense cameras. A RealSense ZR300 camera handles navigation and mapping, while the short-range SR300 RealSense camera is dedicated to the TurtleBot 2i’s MK3 robot arm workspace.

“We watched Raspberry Pi come into market, Beagleboard, DragonBoard and many others. [With Intel Joule], for the first time, we’ve found a board that really delivers the advanced robotics needs we’ve been looking for,” said Dresner. “We’re seeing for the first time a high-end manufacturer stepping into the prototyping world and delivering an affordable, high-performance, ROS-ready robotics solutions. With the combination of Intel Joule and Intel RealSense, the computer vision is handled, and there’s enough computational power to create and control a swarm of robots right out of the box.”

“The TurtleBot 2i is one of the most affordable mobile manipulator robots to ever hit the market,” Dresner said. “We rebuilt the hardware on the TurtleBot 2i from the mobile base up, added a robotic arm as a standard option, with Intel Joule and Intel RealSense. This is an industry first at the price level.”

Dresner said Intel also made Joule easy to get up and running, which wasn’t the case with Intel Edison.

“Joule is about as easy to get up and running as Raspberry Pi is, and that’s where others fall short because they require a lot of assumed knowledge,” he explained. “I had the Joule up and running in a single evening. I’ve been bogged down with two weeks of development trying to do the same thing on other platforms. The user experience between the Joule and Edison is night and day.”

Editor’s note: This article first appeared on our sibling site Robotics Trends. Stay tuned for more coverage of Intel Joule in an in-depth whitepaper and webcast to be released in May 2017.

More on Intel and Robotics Developers: