At the Bosch ConnectedWorld conference in Berlin, a new open source autonomous driving accelerator was introduced. OpenADx focuses on the software development toolchain for autonomous driving, an enabling component in the landscape of highly autonomous driving.
The goals here are the creation of a toolchain specification that is accepted industry-wide, a reference architecture, and the interoperability between tools based on open interfaces. The initiative strives to become the leading autonomous driving ecosystem. The first OpenADx hackathon took place at Bosch ConnectedExperience, bringing together the hardware and software stacks of Appstacle, the German Aerospace Center, the Eclipse Foundation, Driverless Formula Student, Mathworks, Microsoft, Red Hat, and Bosch.
Autonomous driving is a complex challenge. We are dealing with an ever-increasing complexity of in-vehicle hardware and software as well as cloud technologies in the backend. Autonomous driving applications will be constantly enriched and updated, even after being rolled out in the field. This will require new approaches in testing and simulation. The players who master these complex development processes best will be the ones to find success out on the market. In the push toward more efficiency, the interplay between development, test and validation tools is growing in importance. Hence, the autonomous driving toolchain – the third technology cluster of autonomous driving – is emerging from the shadows and into the light.
Fragmented autonomous driving toolchain
If you look at the tools needed to cover the entire autonomous driving development chain, you can see it’s a jungle out there. Hundreds of specialized tools from many different vendors need to be integrated to create a complete end-to-end toolchain. Few of these tools were ever designed to be compatible with one another. This costs the industry time and money.
What can we do to address this? One answer is the OpenADx initiative, which Bosch and Microsoft have initiated together with many further players in this space. OpenADx aims to bring transparency and better integration capabilities into this highly heterogeneous tool space. It is combining open source and closed source with a focus on interfaces between tools. OpenADx is an initiative that is open to all OEMs, Tier 1s, and tool and technology vendors. Open standards represent one of the fundamental requirements in order to be able to use the opportunities presented by autonomous driving.
Benefits for the autonomous driving ecosystem
Advantages for OEMs and Tier 1s: The autonomous driving toolchain enables developers to collaborate better and faster. They use a series of highly integrated tools. For example, they can seamlessly transfer data and code for each step in autonomous driving application development. OpenADx envisions a generic toolchain solution that can be used industry-wide with the objective of significantly reducing provision costs.
Advantages for the manufacturers of autonomous driving tools and technology: Seamless integration of their software tools and technologies in the autonomous driving toolchain will increase the likelihood of their tools being implemented by industry players. The respective tools and technologies will then be used by more developers, yielding higher market penetration and revenue.
Simulation: The challenge in the simulation of autonomous driving functions consists of the high number of test cases quickly generated by variation in parameters. Moreover, the distance between the Earth and the sun has to be driven 44 times in order to validate functionalities such as a highway pilot through suitable tests. These tests can be run correspondingly faster in virtual, simulated environments in which the tools deliver additional efficiency through seamless interaction.
Massive Data Ingest and Management: Developing and validating autonomous driving functions requires the collection of data from various sensors, actuators, and other vehicle sources. The work being done in this testbed deals with the question of how the large and heterogeneous data volume coming from an autonomous vehicle’s system can be transferred into corresponding development systems.