It may be a surprise but the roots of autonomous vehicle testing date back to 1986 with the commencement of the Eureka PROMETHEUS (Programme for a European Traffic of Highest Efficiency and Unprecedented Safety) Project, a unique collaboration between many major European automotive manufacturers, automotive suppliers, universities and institutes. The project “established the foundations for the networked mobility of tomorrow” through reaching breakthrough milestones early on in the autonomous vehicle industry. This was demonstrated for the first time in October 1994 when their research vehicle covered more than 1,000 kilometers of autonomous driving at speeds of 130km/h.
This breakthrough was a precursor to the ultimate vision of the mobility industry, which is to become Connected, Autonomous, Shared & Electric. In other words, this could be called ubiquitous mobility. Sven Beiker expressed this as “whenever, wherever, whatever the weather” in the recent “Innovations in Mobility & Transportation” event hosted by Caliber AI and SAE International. The convergence of these trends paves the way forward, transforming the whole mobility industry and more disruptively, how humans travel in their daily lives — just as the internet and smartphones disrupted how humans communicated and absorbed information in their daily lives.
By 2030, the automotive industry could experience a $1.5 trillion USD or 30% increase in its market value because of recurring revenue streams made possible through shared mobility, in contrast to the traditional automotive market which will be valued at $5.2 trillion USD by that time. This marks the evolution of the ownership economy to a passenger-focused economy, as the Boston Consulting Group estimates that “23% to 26% of miles driven in the United States, or about 800 billion to 925 billion miles, could be traveled in Shared Autonomous Electric Vehicles by 2030”.
This may all seem amazing in the limelight, but by delving further into the industry it is evident that there are still many barriers withstanding before ubiquitous mobility can be achieved. Some of the biggest challenges of the Autonomous Vehicle industry as outlined by Yaser Khalighi, Founder & CEO of Caliber AI in his talk regarding “Data Challenges and Solutions for Autonomous Vehicles” are currently regulations, liability, engineering, ownership, social acceptance, and data.
The main challenges we are faced within the data journey can be divided into four steps:
Data Collection & Storage, Data Management, Data Labelling, and Data Validation.
In this post, we address the first three in more depth and then look at how data is reshaping the value chain.
Data collection presents two problems. One is the sheer volume of data which must be collected in order to validate the reliability of AVs. Second is the veracity of the data and ensuring the data we are collecting is up to the standard which it needs to be in order to be used to train computer vision models.
Data storage presents a huge problem as AVs generate between 4–10 terabytes of data per day. This means that AV companies must have an infrastructure in place both inside and outside their vehicles which has the capacity to handle large volumes of data and maintains scalability. Also has the ability to be scaled.
Data labeling presents an interesting labor-intensive challenge for any AI system, AVs are no different. Each hour of AV data collected requires 1000 human hours to be labeled and the precision of the annotations must be 100% precise because even a 0.1% error can cause an accident when in the real world.
Data management is where it all comes together — a unified platform that facilitates the flow of data from collection to storage to labeling and finally validation. The platform needs to record all types of metadata regarding the data that is collected by the AV and track the journey of the data in order to maintain efficiency and usability for engineering teams. Although many other aspects of the AV industry are more admired, without the Data Management Platform it is almost impossible to support the whole Information Architecture which is needed to operate Test AVs.
Where Data fits in the Autonomous Vehicle Value chain
There aren’t many companies in this day and age which possess all the necessary expertise and resources to build an autonomous vehicle from the ground up. From Tier 1 to Tier 2 suppliers and even sub-tiers, each field is filled with its own complexities.
This suggests that the AV value chain will differ from traditional ones in which the value chain is already mature. As a result, data can be regarded as its own separate sub-tier, where it is made up of solutions that can all be individually provided to many OEMs, Tier 1 & Tier 2 Suppliers directly.
The current state of the AV Value Chain is depicted in the graph above, where OEMs and Tier 1 suppliers are the same as the traditional automotive value chain but Tier 2 Suppliers are mainly AV Hardware & Software suppliers. The Data sub-tier is then broken down into 4 sections, which can all feed into any individual OEM, Tier 1 or Tier 2 Supplier.
Due to the complex and evolving landscape of the AV Industry at present, we must look deeper at each company in order to understand how the Data sub-tier specifically feeds into the AV Value Chain. This can be done by starting at the top of the value chain and drilling-down to draw a more detailed picture of the company’s alliance strategy.
We do this in the next three posts.