The Bosch Automated Driving division faces among many the challenge that the toolchain for developing automated driving solutions becomes more complex by every level of automation. Well established tools are not well integrated for the use cases needed and it is easy to detect gaps in the overall toolchain for dedicated tasks. Instead of solving these challenges alone, wasting lots of money on the path, the Bosch division together with partners from the industry but also from other domains builds up a community to solve the toolchain challenge together in the OpenADx initiative. With this the partners expect a substantial saving on toolchain costs but also a better integration within organization and especially at the interfaces between cooperating organizations. This talk will present the approach and the current state of the community.
Automated driving solutions introduce a new complexity into the development of embedded systems in a car. This complexity rises with each level of control and autonomy of the automated driving systems. New tool categories have to be added like machine learning, but also existing technologies, like simulation, are stretched to their current limits. E.g., it is expected that the validation of a fully automated driving solution requires to do test drives in the amount of millions of kilometers. This results in the need for very complex simulation as part of the validation as well as the handling of extensive amounts of data in order to ensure quality at a realizable effort.
The toolchain for such challenges is complex and the integration of all the tools coming from different domains cost a lot of effort without a real competitive advantage towards the automated driving solution. Therefore, the Bosch Automated Driving division, together with Microsoft, is currently building up an ecosystem of companies within the industry including OEMs, tier 1 suppliers, tool vendors, research organisations but also partners from other industries like the IT industry. The intiative called OpenADx is supported by the Eclipse Foundation as a host for the activities.
The goal of this endeavor is to stop wasting money on the introduction of a proprietary toolchain in each of the companies and to share the development costs for the toolchain with partners from the industry. Besides the benefit of sharing the costs, the expected result is a better integration of the toolchain within the organizations, but especially also at the interface between cooperating organizations. For tool vendors and research organizations, the advantage of the approach is in the existence of an integration backbone which allows the provider to easily integrate new technology or tools into a working environment that runs in a multitude of customer organizations, instead of providing proprietary solutions for single customers.
The goal of the initiative is not to replace existing tools. There are many tools and technologies, be they commercial or open source, that solve perfectly their job and should do so in the future. The goal is to define development workflows and to support the integration of tools along those workflows as well as to fill gaps identified and not solved by existing tooling. This integration glue will be provided as open source software under the umbrella of the Eclipse Foundation.
The current state of the initiative is that it is currently searching for interested parties throughout the world. The goal is to make this an industry effort not driven by single partners that is big enough to propose and realize an industry standard for automated driving solution development. To get the ecosystem running with only a limited commitment necessary, the initiative starts small. Through a series of so-called Hackfests, i.e., 5 to 10 day Hackathon events, the interested parties get the chance to make first hand experience in the collaborative approach. Additionally, the members get to a common ground on the technologies available and the development approaches at hand in order to identify common integration goals which build the basis for first open source projects spanned from the ecosystem. Based on the experiences of these Hackfests, the members of the initiative will furthermore decide on topics like how to go public officially or how the intiative will be structured.
At the time of the talk, the first Hackfests have taken place in which we will have provided prototypical solutions for the area of closed-loop simulation of automated driving functions with a perfect perception and the handling of massive amounts of data in the toolchain. The results of these Hackfests will be shown in the presentation as well as an outlook on the further activities.
Speakers: Lars Geyer-Blaumeiser