From Silicon To Software

 

Top 5 Challenges to Achieve High-Level Automated Driving

autonomous driving technology

By Editorial Team

The development of assisted and automated driving technologies continues to advance at a rapid pace. Automated driving takes a number of forms, and thanks to extraordinary advancements in the automotive industry, several of them, including predictive cruise control, lane centering, and automated parking, are already within our grasp. However, to successfully navigate on the road of widespread adoption, there remain several obstacles that we need to overcome.

In a keynote presentation at this year’s Synopsys ARC Processor Virtual Summit 2021, Dr. Jürgen Bortolazzi, head of Advanced Driver Assistance Systems (ADAS) and Automated Driving at Porsche AG, said realizing the full potential of automated driving is largely a question of partnership.

To begin with, automotive manufacturers still need to establish full alignment with legal authorities on the technology’s implementation. At the same time, there is the need for extensive cooperation between carmakers themselves, as well as strategic collaboration with a wide range of specialists in fields such as data, hardware, software, and artificial intelligence (AI). Volkswagen Group, which owns Porsche among brands including Audi, Bentley, and Lamborghini, is already relatively advanced on its automation journey. Even if at first glance, automation doesn’t appear to sit naturally with what are known to be “drivers’ cars” — cars that people buy largely for the enjoyment of their performance.

The thrill of the open road is one element of driving. But as Dr. Bortolazzi points out, “a lot of driving conditions are not fun” — such as traffic jams, searching for parking spaces, and then actually parking. German drivers spend an average of 41 hours each year looking for somewhere to park, while in New York the figure rises to an astonishing 107 hours! The Volkswagen Group’s efforts currently center on alleviating these stressors while preserving the excitement of driving. “Automated driving will accompany, not replace, the ability to manually drive a Porsche,” affirms Dr. Bortolazzi.

Read on to learn about the top five challenges for autonomous vehicles to achieve higher levels of automated mobility and the need for enhanced design methodologies to deploy Level 3 and 4 automated systems.

1. A New Traffic System Means Entirely New Legal and Social Frameworks

Dr. Bortolazzi cited “acceptance” as a key challenge in introducing autonomous vehicles to market, calling it a “societal issue” that still requires trust-building. Considerations run from the obvious safety concerns to areas such as energy consumption and data ownership. Currently, there is little sense in introducing autonomous technology as an isolated feature; rather, it must be an integral part of the traffic system of the future.

Infrastructural acceptance follows the establishment of legal frameworks. In Germany, authorities recently passed a law allowing Level 4 autonomous vehicles to operate on public roads with an official permit. Dr. Bortolazzi said he expects the first autonomous passenger cars to make their German road debut next year. But he added that globally, each region and market are at very different stages, and establishing a framework for unmanned systems remains a quantum leap in legal terms.

The ruling in Germany is a major advancement to the industry, and other countries are expected to follow suit. China will be one of the key players of the anticipated autonomous revolution due to close cooperation between central legal authorities, individual regions, industrial partners, and startups — promoting dynamic development.

2. Automotive OEMs Can’t Expect to Succeed Alone

Alliances such as that between Volkswagen and Ford are a financial and sometimes ideological challenge that players in the autonomous space must reckon with. Dr. Bortolazzi sees the market as divided into two distinct categories — traditional carmakers aiming to improve the experience for the driver, and newer mobility companies looking to take the driver out of the equation altogether.

In both fields, strategic alliances will be necessary to advance the technology toward the mainstream. “Alliances are a dynamic, huge investment effort,” said Dr. Bortolazzi. It’s expected that ongoing consolidation will be driven by the need to set clear standards and build scalable platforms that a wide range of manufacturers and service providers can use.

In the China market, the need for tie-ups with local entities will be essential for international carmakers to develop autonomous services in the market as foreign companies are barred from holding China-based data independently. BMW, for instance, has already partnered with both Baidu and Tencent for system development and data computing and storage, respectively.

3. Vehicle Data Collection Requires an Overhaul

There is a technological gulf between Level 2 (partial driving automation) and Level 3 (conditional driving automation) and above. Broadly speaking, the focus of the next few years will be the optimization of Level 2 capabilities, while the shift to Levels 3 and 4 will happen in stages, with restricted usability and availability.

Levels of Autonomous Driving

However, system limits will be tight due to the risk reductions that OEMs have to perform, and there will be a stepwise extension of system links. Developing the necessary algorithms will be a job for large teams as it will require a new, much more wide-reaching approach to data collection. AI training sets will need to draw on a vast repository of information collected from a combination of prototype and existing customer cars. Collecting, processing, and storing datasets at this scale is a huge effort that requires new architecture of data management systems.

Safety and security are further key concerns in terms of data as well as on the road, which adds to an already challenging development process. A vehicle’s automation system needs to be secure against attack from hackers seeking to harm either the manufacturer or the customer directly.

4. New Processing Demands Call for a Quantum Leap

Fully automated driving depends on functional redundancy — the ability of the vehicle to respond to a wide range of complex scenarios. A combination of cameras, radar, ultrasonic sensors, night vision, and high-definition mapping, which provides high redundancy at the front of a vehicle and coverage at the back and sides, is adequate for Level 2 and Level 2 Plus functionality. For Level 3 and above, LiDAR (light detection and ranging) technology becomes necessary, as well as a two- to three-fold increase in redundancy on the back and sides.

Current processing performance requirements are already high. Volkswagen’s existing system runs more than 35 applications from 10 suppliers, with real-time customer safety functions spread across multiple hosts. Dr. Bortolazzi says the next generation of ADAS calls for on-chip solutions that provide a higher level of integration while offering enhanced energy efficiency and scalability.

The need for a processing platform 5 to 10x as powerful as the current one will be necessary to fully enable Level 3 and 4 functionality, what he refers to as “a quantum leap in board processing performance.”

5. Closed-Loop System Is Critical for Development Speed

Lastly, effective verification and validation (V&V) calls for a significant improvement in vehicle measurement equipment, data management, and investment in proving grounds, methods, and tools. The automotive industry is already using simulation to good effect but must extend its use dramatically. This means establishing a closed loop that encompasses data collection from test and customer vehicles, analytics, and processing.

The creation of such a loop stands to reduce development cycles from a period of several months to weeks – an essential factor for automated mobility to advance, reinforcing the need for strategic partnerships. In Volkswagen’s case, this is an alliance with Microsoft that makes use of the company’s Azure cloud platform to build a layered architecture comprising data management and processing and analysis tools.

Summary

The stakes are high, and as software and electrical/electronic (E/E) architecture become a central feature of mobility, an additional challenge is sure to be talent and the task of building new technological skillsets from within. The next stage of the autonomous journey is set to be one of continuous learning for everyone, and it is a truly an exciting time to witness breakthroughs the likes of which the industry has never seen.

To watch Dr. Jürgen Bortolazzi’s full keynote presentation and much more from this year’s ARC Summit on demand, register here.

Learn More