When tough legislation can make the self-driving industry unsustainable
Autonomous technologies hold a grand vision of disrupting the transportation industry. A promise of making automotive travel safer, and allowing people to focus on more creative tasks – leaving A to B logistics to the tech sector. With level-4 autonomous cars available on the market as soon as 2021, the transportation industry is on the verge of being disrupted – but at what price, and is it really feasible?
One of the most prominent sectors where drive automation can redefine the industry is in truck-based logistics. With more than 3,5 million trucking-related workers in the US alone, the industry accounts for 2,3% of the country’s active workforce. The benefits of having autonomous trucks are quite vast, making delivery times shorter, highways substantially safer, and increasing the productivity of each vehicle. It promises to replace human-based driving, by using the latest technological advancements and making A to B services more affordable.
How many truckers can an AI replace?
Robot trucks can operate 24 hours, 365 days a year, stopping only for a quick check, maintenance, or a cargo pickup. Taking the US trucker’s standard 70-hour weekly working shift in mind, each autonomous truck can handle as much work as two human drivers, assuming that 20% of time the truck will be used for procedures that are non-driving related.
How much money can an autonomous truck save?
From a business perspective, autonomous trucks allow for a significant reduction of the human workforce and promise to potentially make truck-based logistics services more affordable. With the median salary of a private fleet truck driver being 73 000 USD a year, autonomous trucks can be a bigger investment at first, but can potentially save 146 000 USD every year. With the average lifetime of a truck varying between 10 to 12 years, autonomous technology could save around 1 500 000 USD during the life span of each heavy-duty vehicle, leaving fuel/energy as the main cost incurring variable in the equation.
Sensors, Data, Computing and Decision-making
Companies like Waymo, Tesla, Daimler, Volvo, Uber, and Embark are making large bets, investing in technologies that can potentially reinvent the trucking industry. With so many active players on the market, autonomous driving technology is not a miracle tech; it’s a result of many years of R&D, legislation lobbying, a highly competitive industry, and lots of financial investments from tech and automotive giants. From a tech perspective, all this development relies on four pillars of technology:
- Sensors – such as LIDARs and cameras that generate environmental data.
- Storage infra – an infrastructure used for storing data, on the edge and in the cloud, for future analysis and improving algorithms.
- Computing & Graphics – the hardware responsible for autonomous driving calculations and decision making.
- Autonomous software – the software stack, that in the end makes all the decisions for the car.
All these pillars incur costs for the overall architecture of the car, though many elements will gradually become more affordable. With Moore’s law still being intact, the trend of more computing per dollar will make the cost of advanced calculations more reasonable. The autonomous software stack can always redefine its business model to fit the market, as software is extremely scalable. Advanced sensor technologies like LIDARs, are still in their early days when it comes to production, and analysts are predicting better and more affordable sensors just around the corner.
How data and storage can be the bottleneck
Storage infrastructure, on the other hand, is a well-refined industry. There are no long-term expectations for significant price drops in NAND-based or Spinning Disk solutions. And so far, there are no signs of an upcoming technology that could disrupt the edge storage market. With black boxes on the edge of the car – and cloud storage maintaining the data in a hybrid environment – this infrastructure ensures reliable handling of sensor and decision-making information throughout its lifetime.
The edge and cloud infrastructure of autonomous cars both clearly need a lot of storing capacity to handle this data. According to our calculations, autonomous cars produce 1.4 Terabytes of data in just one hour, so assuming the car would drive 20 hours a day, the daily total would reach 28TB. Let’s assume that this is the amount of data that would need to be stored on the car itself; for such a scenario, the best answer would be to use an SSD solution as high data streams require fast writing. We’d require 28TB of SSD storage, putting the total cost at only 4 000 USD. This assumption takes into account consumer priced devices, and it’s important to point out that automotive and enterprise-grade solutions tend to have significantly higher prices. At the end of the shift, the data would probably move to the cloud, freeing the edge storage of the car, and moving data to cold storage HDDs. These devices are much more affordable than Flash storage, being able to store the same amount of data for less than a quarter of the price, while costing just 840 USD to store all this data.
The price of data
Generating data is not that complex; you take a sensor and put it in a dynamic environment, creating change in the sensor parameters and thus generating data. The broader the scope of the sensor, the more data it provides. Storing data, however, is more complex, as there’s always a tariff on preserving information – the more data gets preserved, the greater the cost. When it comes to autonomous cars, the limit at which data becomes irrelevant and can be scrapped is not yet defined. Most of these legislations are still up for discussion, creating certain risks and opening room for speculation.
The uncertainty of legislation
Legislation is an area that can create a lot of risk through uncertainty. We don’t really know what the final guidelines are for autonomous car data, especially when it comes to an industry like logistics, and that can be a risk. Here’s a speculative example: if all the autonomous data would need to be stored for 5 years prior to deletion, then storage costs, ignoring the entire infrastructure, would have a significant effect on the total BOM (Bill of Materials). Consider the following simple calculation: 840 USD (cost of HDD data storage per day) * 365 days* 5 years = 1 533 000 USD. If we compare this to the previously calculated savings of 1 500 000 USD that an autonomous truck can generate by replacing its drivers, it will still be more expensive than a human controlled truck operation.
Storage infrastructure for data-driven applications
To prepare for worst-case scenarios, autonomous technology providers need to emphasize the importance of storage architecture, both in the cloud and at the edge of the car. Here at Tuxera, we work directly with Tier-1s and OEMs to solve the storage challenges of autonomous and connected cars at the edge. Working closely with our customers, we help them optimize their storage stack both on the hardware and software layers – ensuring robust, rapid, and long-lasting performance for data-driven vehicles.
Car makers and Tier-1 suppliers – see how we can make data handling and storage faster, more secure, and more reliable.