Real autonomous driving with limited or no human interaction is still many years away, according to analysis by Semicast Research. Almost every vehicle maker and seemingly every tech company has now announced plans to introduce semi autonomous and fully autonomous vehicles.
To not do so now renders a company - and its stock price - as a throwback to the wagon and horse and thus heading for extinction; witness Ford’s dumping of CEO Mark Fields earlier this year on the basis that his plans for self-driving cars were not ambitious enough. Everyone is now racing to catch up with Tesla, the most technologically experimental vehicle maker - and also one of the newest but lowest volume producers - in a battle for supremacy.
The journey from today to tomorrow is set out by the Society of Automotive Engineers (SAE), a US-based, globally association and standards developing organisation. The SAE has defined a series of automation levels, mapping a path from entirely human operation (Level 0) to machine-only operation (Level 5). Semicast sees Level 4 light vehicles accounting for less than one percent of global production in 2024, with production of Level 5 light vehicles not commencing until the second half of the next decade at the earliest. For now this leaves humans in control, with Artificial Intelligence (AI) providing driver assistance (Levels 1-3) and a substantial - but steadily falling - portion of the market with no machine intelligence at all (Level 0).
The long term trend to semi and fully autonomous light vehicles looks set to mostly take place from 2025 to 2040, with machine learning algorithms rapidly becoming more sophisticated and capable of operating reliably in a growing range of driving situations over this period. The scale of the financial investment and R&D effort now flooding into autonomous driving is extraordinary and the pace of change from human drivers to AI is subject to much discussion.
This is a debate which encompasses not just vehicle makers and the tech industry, but also lawmakers, legislators, regulators (NHTSA) and investigators (NTSB). The market’s current situation is reminiscent of the early days of the internet and its expected impact on commerce, with incumbents and new entrants vying for supremacy in a rapidly changing environment. As then, it will be many years before the winners and losers of the autonomous driving race are known.
What is known is that humans are an inherently weak link; through a combination of boredom, distraction and fatigue, humans are not best suited to the repetitive and typically uneventful nature of driving. Just witness any freeway, with some drivers passing countless miles looking at their cellphone, entertainment system, tablet, or reading a book, all the while in control of two tons of metal traveling at 60 miles or more an hour. On the long road ahead to fully autonomous driving, Semicast sees a short term missing link: driver monitoring systems.
Driver Monitoring Systems (DMS) are camera based and monitor both the head and eye gaze of the driver, using technology functionally similar to that seen on the recently launched iPhone X. In Level 0/1/2 vehicles, if the driver’s attention is distracted from the road ahead for more than a few seconds, or if early signs of fatigue are detected (changes to blink speed and duration), DMS can alert the driver audibly or visually; in Level 3/4 vehicles, DMS will also determine the state of alertness of the driver in the event the AI needs to hand back control.
Semicast estimates the global installation rate for DMS to have been less than one percent in 2016 but sees this rising rapidly, for example in the 2018 Cadillac CT6 using GM Super Cruise. Considering legislation for the mandatory introduction of other safety systems - airbags, ESC braking and tire pressure warning - in many countries over the last ten years, similar mandatory legislation for the installation of DMS in at least one of the major light vehicle production regions is highly likely long before Level 3/4 vehicles become common.
The first person to be killed in a Tesla operating in Autopilot mode was in Florida in May 2016. The vehicle’s vision system is reported to have misidentified as sky a white-sided semi-trailer towed by a truck crossing an intersection; with the vision sensors indicating no obstruction and the driver distracted, neither the AI autopilot nor driver braked before impact. The vehicle lost its roof as the car passed beneath the trailer and the driver his life.
DMS did not feature at all at the Frankfurt Motor Show (IAA) in Germany last week, but the NTSB did - on the first press day of the show - publish its report into the Tesla fatality. In a report titled 'Driver Errors, Overreliance on Automation, Lack of Safeguards, Led to Fatal Tesla Crash' the NTSB findings include:
Colin Barnden, Principal Analyst at Semicast Research commented: “Had the limitations of Autopilot been acknowledged and the vehicle developed with the same DMS technology as the Cadillac CT6, then in my opinion this death could have been prevented.”
Coincidentally - or conspiratorially?
Euro NCAP launched 'Road Map 2025 - In Pursuit of Vision Zero' on the same day as the NTSB report, which presents a timeline for the installation of DMS from 2020. Barnden added: “This timeline is likely to jolt a number of OEMs, Tier 1s and semiconductor manufacturers into action and to catapult relatively unknown eye tracking companies, such as EDGE3 Technologies, Seeing Machines and Tobii, into the mainstream.”
As the 2017 Frankfurt Motor Show demonstrated, the auto industry is mostly dreaming of a far-off utopian autonomous future, while distracted drivers are killing themselves and others today. Barnden summed up “The message from the NTSB and from Euro NCAP is unambiguous: Quit dreaming, the time for action on DMS is now.”