Toyota Selects Apex.OS To Underpin Automated Vehicle Platform

In the middle of the last decade, seemingly hundreds of companies popped up out of nowhere to chase the automated driving system (ADS) gold rush. One of the many tools that made it possible for almost anyone to start an ADS effort was ROS, the open-source robot operating system. While ROS was a great platform to develop on, it was missing many key features to make it usable for a safety critical environment like a vehicle control system. Enter Apex.OS, which has now been selected by Toyota for future ADS programs. 

Apex.AI which produces Apex.OS is the brainchild of CEO Jan Becker and CTO Dejan Pangercic, a pair of former ROS developers. In 2017, they forked ROS and began work to address the issues that made it impossible to directly take ROS to a production vehicle program. Over the past four years, they have developed Apex.OS in parallel with ROS and kept it consistent with ROS application programming interfaces (APIs. 

Thanks to the consistent APIs, a developer can start work on a project using the open source ROS and when they are ready toward a production ready program, they can take the application and put it on Apex.OS without changes. Since ROS was developed as a general purpose platform for robotic systems, it lacks some of the features needed for reliable, real-time control systems. The Apex.AI team have re-written the underlying code and libraries of ROS to achieve the desired safety levels such as making it memory static. 

“What that means is there is no dynamic memory allocation anymore during runtime,” said Becker. “That works great for a consumer device where you want to have as many apps as possible on a device, but running software in real time, doesn’t work when you allocate memory.”

In a vehicle control system, you won’t be running just any random applications as you would on a computer or mobile device. Everything is predetermined so you can allocate the needed memory at boot up. This is especially critical for real-time, safety critical control systems like ADS.  A range of other changes such as blocking calls to input-output ports are all designed to help keep everything running smoothly and on schedule. 

Apex.AI has also rewritten many of the underlining libraries to make the OS hardware platform agnostic and consistent regardless of which platform it is running on. When using ROS, developers have to recompile much of it with completely different library files depending on whether they are running on an X86 or ARM CPU, a GPU or FPGA. Apex.AI has reimplemented these so that they can run on any of the major platforms. 

As a result of this rewrite, Apex.OS has been certified as meeting the ISO 26262 automotive functional safety standard.  Within the ISO 26262 standard there are several automotive safety integrity levels (ASIL) with ASIL-D being the strongest risk management level and the one applied to Apex.OS. This means that the software has been verified as safe to use on production vehicle applications.

According to Becker, Apex.OS is already being used on several off-road vehicle applications while some on-road applications are projected to begin around 2023. Toyota has been working with Apex.AI almost since the project was launched in 2017 and the automaker’s Woven Planet division is now working to implement Apex.OS into its Arene vehicle development platform. Arene is intended to be the most programmable vehicle in the world to enable developers to quickly get new mobility and automation applications up and running for development and deployment. 

In addition to Toyota’s Woven Planet, Apex.AI is also partnering with Japanese tech startup Tier IV. Tier IV created Autoware, the first open-source automated driving software stack. Tier IV will be combining its software with Apex.OS to enable developers to work on automation with a platform that is capable of eventually moving into a production environment.