Automakers Challenged by Open Source ADAS

Very often, the hot ADAS solution of the moment is one created in-house by a talented group of developers.

That stands to reason, because some of the best software in tech history (and hardware, while we’re at it) has been devised, tested and brought to life by well-paid coders, scientists, engineers, etc. That’s why the future of automobile technology will require great masses of capital.

In fact, it’s going to be prohibitively expensive for even the most deep-pocketed automakers and solutions providers. This is the core reason why free and open-source software (FOSS) will be critical to getting us to full Level 5 autonomy. For many in the 21^st century auto space, it’s the priming that will start the pump. “To overcome the initial expenses, free and open-source software is the best way to start prototyping and development,” said Ramin Shirani, CEO and co-founder of next-generation mobility company Ethernova.

Happily, developers of ADAS and self-driving solutions have a wealth of solutions across a score of automotive and auto-adjacent disciplines to draw from. Xiaodi Hou, co-founder and onetime CEO of autonomous truck technology developer TuSimple, said: “Many software components are available from the autonomous driving industry and other neighboring industries, such as deep learning infrastructure, cloud computing and robotics. These components can greatly reduce the development cost and headcount needed for an autonomous driving company. It is more capital efficient and allows you to move much faster than the traditional way of developing.”

However, as we all know, there are drawbacks to free, versus custom-designed and built products that do exactly what we want them to. Shirani’s company specializes in ethernet connectivity between vehicles and, in order to help build its solutions, the company has availed itself of several off-the-shelf products.

This illustrates one major drawback of FOSS, though. It is, and likely always will be, a hodge-podge of software that can often perform certain individual (or limited set of) tasks well. At the risk of stating the painfully obvious, ADAS systems are already sprawling and complex, a situation that will only snowball as the race to autonomy intensifies. No FOSS solution is going to tick every box a developer needs to fill out to produce a good system.

That is the art of using FOSS; determining which solution or solutions is right for a developer or carmaker’s work and figuring out how to most effectively blend it (them) with existing proprietary tech (which an early-stage developer might not have, of course). This in itself consumes financial and operational resources, to say nothing of brainpower. Hou said: “The challenge today for companies is choosing the right path from a large number of options. There could be a hundred open-source tools you can choose from, but there is no textbook way to connect or orchestrate them.”

Another challenge is licensing; at times, the usage terms for FOSS solutions can produce roadblocks at awkward moments, in uncomfortable places. Shirani said that frequently, “restrictive licensing stops us from modifying and distribution – in many cases, it becomes mandatory for us to contribute and maintain changes back in the community. Though OEMs and technology startups like us, agree to the intent of contributing back to the community, resource constraints and testing overheads do not allows us to do this for every open source software component.”

Although swashbuckling entrepreneurs with dollar signs in their eyes might not want to hear it, getting the most out of FOSS also requires collaboration. According to Shirani, “a collective effort including corporates (OEMs, tech providers) and community is essential for full autonomy. Like the Linux operating system, unless everyone contributes, we will never be able to enable full autonomy. One OEM or one community will not be able to achieve this.”

Regardless, it’s clear that more than a few companies in the auto industry, of various shapes and scales, are relying at least somewhat on FOSS. Meanwhile all of us, at one time or another, have despaired about a free product or service that falls well short of a paid version. That begs a question – are most FOSS offerings reliable enough for important ADAS development work?

Hou strongly believes so. He said that “my overall experience is that the reputation pressure of publicizing an engineer’s code base very often drives the engineering quality of the open source projects to a much higher standard than an average proprietary project of large companies, not to mention those hacky projects from start-ups.”

It isn’t easy to use FOSS but given the savings in costs and other resources, it’s unavoidable in a vast number of cases. It will be one of the motors powering ADAS development; the trick will be to rev and steer it the right way without damaging the vehicle too much.