The journey to fully autonomous vehicles is taking longer than some people initially thought. Why is it proving so complex?
Mitra Shah: There are two main areas of complexity: dynamic driving tasks and the operational design domain. Dynamic driving tasks are the real-time functions required to operate a vehicle safely – signaling, maneuvering, braking, avoiding traffic and so on. It’s a huge amount for a machine to process and react to. The operational design domain refers to the environment the vehicle operates in – things like different types of weather, surrounding infrastructure, times of day. In certain reduced operational domains, like campus shuttles, there are already solutions that provide level four autonomy with no human involvement. But in most situations, we’re only at level two, with systems such as advanced cruise control and self-parking. Reaching higher levels of autonomy essentially requires the industry to create systems with the same, or better, knowledge, reasoning, and responses that human drivers can provide.
Presumably there are a lot of different companies trying to solve these problems?
Modar Horani: Indeed. Cars are going from being mechanical products with some electric and software components to becoming a software-driven platform with different pieces of hardware. OEMs aren’t going to do that by themselves. The open-source community is helping to accelerate innovation and create a lot of good products to build autonomous vehicles – from simulation and testing through to components for the vehicles. But ultimately, OEMs need to develop an overarching operating system for these cars – they need to get all these different bits of hardware and software to speak the same language. Creating that cross-compatibility is tough. It means the development, testing, evaluation, and approval phases can take a very long time.
So to improve things, umlaut has created the Automated Driving Engineering Platform (ADEP). Tell us more.
Colin Goldsmith: Essentially, ADEP is an ecosystem made up of different software and hardware elements that helps OEMs and suppliers to speed up the development of components for connected autonomous vehicles.
Mitra Shah: There are two different platforms within ADEP. One is the development platform. This enables us to rapidly prototype use cases consisting of variety of sensors, software, compute from simulation to test vehicle deployments. Allowing early-stage iterative feedback loops to find out the most appropriate solution for each client. The second main function is the evaluation platform Once the autonomous vehicle software is developed, the client can leverage ADEP to do performance and safety evaluation leading to compliance to standards and regulations in different countries. So, if a client has developed a vehicle in Germany and wants to deploy it in the US, ADEP will evaluate what, if any, adjustments need to be made and how best to do them. Standards and regulations for autonomous vehicles are evolving all the time and can vary from region to region, so this evaluation platform really helps companies save a considerable amount of time and resources.
Could you give us any recent examples?
Mitra Shah: Sure. We recently worked with a Japanese OEM in North America on a blind spot monitoring system for its auto lane change assist feature. The feature was already in production, but they had numerous failure rates and the OEM wanted to figure out how to improve the feature. We developed a prototype platform for them with a 360-degree camera system, three front-facing cameras, corner radars and a long-range radar at the front, which meant they didn’t have to go out and interact with multiple sensor manufacturers. We also carried out the data collection, analysis, filtration and even visualization. Thanks to ADEP, we were able to finish the project around two weeks quicker than normal and with a 30% increase in data collection frequency.
Is ADEP aimed purely at OEMs or is it also for automotive suppliers?
Modar Horani: Mostly OEMs, but suppliers can also use it to validate products. Say a supplier wants to evaluate an electronic control until that they are developing. Our team here would use a bespoke combination of software and hardware to simulate a wider operating environment to see how it would perform against scenarios in different regions.
What solutions are being enabled by ADEP and how will this accelerate the road to fully autonomous vehicles?
Mitra Shah: One solution developed through ADEP’s development accelerator platform is MDAPT, which is used to prototype perception and computer vision use cases and further in ADEP’s Evaluation Platform.
Learn more about MDAPT in an upcoming interview.