Teaching autonomous vehicles to drive is incredibly data and compute intensive

Dr Florian Baumann, EMEA CTO Automotive and AI, Dell Technologies.

Future vehicles are expected to be heavily reliant on data and connectivity. Teaching autonomous vehicles to drive is incredibly data and compute intensive and demands secure and cost-effective data management. It also requires an agile, integrated, software-defined infrastructure that spans everything from edge to core, and across multiple clouds. This is where Dell Technologies offers a wide range of enterprise architecture solutions from storage to cloud. The Dell EMC automotive data storage solutions can help organisations achieve speed and differentiation with simplified data management and predictable performance at the massive scale required for Advanced Driver Assistance Systems and autonomous driving development, testing and AI training.

There are five levels of autonomous driving, as outlined by the US National Highway Traffic Safety Administration, with increasing use of automated system per level. Most of the current technology limitations are found in the advanced levels. The AI limitations and ability of the algorithms to respond to unfamiliar scenarios is one current drawback. Sensor responses to environmental changes, like weather and terrain, is another one. 

Currently, fully autonomous vehicles are only operating in completely controlled environments. Often not considered is also the approval and inclusion of ADAS AD functionalities and in-vehicle cybersecurity topics.

It is predicted that by 2030 there will be 250 million electric vehicles, 90 million autonomous vehicles and 1 ZB of data generated by the automotive industry, according to research conducted by Frost & Sullivan in partnership with Dell Technologies. Revolutionised software and application models are truly required to unlock the wealth of data that would be coming in from advanced autonomous vehicles. 

Those that succeed in the coming years will have a long-term vision, clear roadmap, and set of expert technology partners who can help them create an enterprise data-centric platform that intelligently manages the complexity and vastness of the data required to operate fleets of connected and autonomous vehicles and services.

Two key considerations are legislation on safety and security. Policy makers and stakeholders need to work together to identify these requirements and having harmonised industry standards for this is vital. Governments and independent certification authorities also need to play a role in providing laws that will guide ethical standards, since decisions are made by pre-programmed algorithms. 

The first binding agreement between more than 50 countries on common regulations for vehicles to take over the control was made in June 2020. That agreement also includes an in-vehicle black box, a data storage system that records when the automated driving functionality is activated.

Additionally, mobility leaders need to assembly teams with sufficient expertise to create the differentiating algorithms needed to run the workloads and manage the data. Public perception and acceptance are also aspects where the opinion is divided due to perceived safety and cybersecurity concerns, and more needs to be done to address this openly. 

Generally, autonomous vehicles need to sense their environment, process input and make decisions, and subsequently take actions – and research is being conducted in all these areas. Progression in communication structures, or vehicle-to everything communications V2X is essential, enabling vehicles to share information with infrastructure, cloud, networks, persons and other vehicles. 

Research on vehicle cybersecurity is also taking place to guard against attacks or breaches. Additionally, improvement of performance and increased efficiency of sensors and radars are crucial research areas for detection of surrounding environment. In summary, enhancements are being made in intelligence infrastructure, smart city integration, processing units, as well as data storage and management.

Dr Florian Baumann, EMEA CTO Automotive and AI, Dell Technologies.
Dr Florian Baumann, EMEA CTO Automotive and AI, Dell Technologies.

Key takeaways 

  • Teaching autonomous vehicles to drive is incredibly data and compute intensive.
  • Teaching autonomous vehicles to drive demands secure and cost-effective data management.
  • There are five levels of autonomous driving, as outlined by the US National Highway Traffic Safety Administration.
  • AI limitations and ability of the algorithms to respond to unfamiliar scenarios is one current drawback. 
  • Sensor responses to environmental changes, like weather and terrain is a drawback. 
  • By 2030 there will be 250 million electric vehicles, 90 million autonomous vehicles and 1 ZB of data generated.
  • The first agreement between more than 50 countries on regulations for vehicles to take over control was made in June 2020. 
  • The agreement includes in-vehicle black box that records when the automated driving functionality is activated.
  • Mobility leaders need to assemble teams with sufficient expertise to create the differentiating algorithms.
  • Improvement of performance and increased efficiency of sensors and radars are crucial research areas.