Accenture sees need to reboot autonomous driving

Accenture sees need to reboot autonomous driving
Business news |
The development of autonomous driving is not really getting off the ground. This realisation unites the entire automotive industry. However, there is no consensus on the question of why this is so. The consulting firm Accenture has now presented a comprehensive analysis - and says what needs to be done.
By Christoph Hammerschmidt

Share:

A decade ago, the development of new hardware and algorithms gave Deep Learning hope of a breakthrough. Artificial intelligence as a technology for capturing the vehicle environment, especially in combination with cameras, seemed to open up the royal road to autonomous driving (AD), write the Accenture experts in their now published white paper “Rebooting Autonomous Driving”.

The report explores AI’s technical challenges to help illustrate the market opportunity and its trajectory. AD is, according to the Accenture analysts, not about to disrupt incumbent automotive firms’ business models, but the cost of doing nothing is steep. If Original Equipment Manufacturers (OEMs) go all in and fundamentally transform their products and processes, the new configuration plays to their strength: they can use their increased assets in the field to collect an even richer dataset, the authors say.

Their recommendation: To succeed in the AD market, OEMs will need to become much more software-driven. Only a software-first model will help OEMs to leverage all the data necessary to push automation using a mechanism Tesla calls “shadow mode.” Shadow mode entails equipping every car with AD technology that continuously runs in the background—without linking to driving actuators. This process compares the driver’s behavior with system decision-making. Shadow mode is an ingenious—yet demanding—mechanism to source edge cases from human drivers.

The authors point out that the fact that the millions of vehicles out in the field represent an important asset in that these vehicles can help them to collect valuable usage data – data that are very useful for the design of new models. This asset, however, is available only to the incumbent carmakers – start-up companies have the short end of the stick, however innovative their technologies are.

Nevertheless, against the backdrop of new usage models, electronics architectures, drive technologies and connectivity options, automotive OEMs need to rethink their business models as well as their products, the study finds. In particular, AD has the potential to shake up existing business models and reshuffle the cards on the market. However, the experts believe that this has still a long way to go to become reality; as obstacles on the way to AD, the authors see complex technical and regulatory challenges as well as staggering investment needs. Artificial Intelligence (AI), often hailed as the core technology to solve these problems, seems to be more a part of the problem. Example? There are reports that AI-based ADAS are unable to recognize traffic lights with a broken glass or misinterpret a low-standing moon for a yellow traffic light.

Negligible? Solvable? Not so easy. “The evidence from unprecedented field trials has humbled even the bravest futurists,” the Accenture authors conclude. “Because the AI hurdles are much higher than experts initially assumed, today we can only apply real automated features that fully take over the driving process in highly restricted, regulated situations,” they write in their paper.

So what is to be done by OEMs and automotive electronics designers? First, jumping into the autonomous game may require an evolutionary approach rather than trying to revolutionize the entire industry, the study says. Beyond that, the authors have developed five practical recommendations for carmakers:

  • Starting (or continuing) investing in AI early, but in carefully measured stages. The focus should be put on Level 2, because this will be the biggest market for decades, and OEMs can lead the pack here, providing the most valuable real- world data
  • Get the preconditions right by becoming software-defined. Future investment should flow into software stacks and developer ecosystems – and not into powertrain technologies.
  • Use fleet analytics to develop relevant big data and edge cases. This means that they should take advantage of the data provided by the connected vehicles in the field. Something the volume manufacturers in particular can benefit from.
  • Develop a data-first mentality. This enables them to actively develop and define the AD market based on the insights from analytics, big data and edge cases.
  • Partnerships with tech players, tier-one suppliers and/or service providers may help speed time to market and reliability.

The entire White Paper can be downloaded here.

Related articles:

Carmakers, suppliers launch open mobility blockchain initiative

Qualcomm opens German office to be closer to car makers

Cariad, TomTom to jointly develop VW Group’s next-gen navigation system

KPIT shifts up a gear in race for software-defined car

“Open systems have stronger assertiveness than closed ones”

Continental, Jaguar, ZF invest in automotive OS developer

Project develops interaction of software components for future cars

Toyota bolsters its software ecosystem

 

 

Linked Articles
eeNews Automotive
10s