Today, everywhere you look, you find automotive giants plowing in resources in beefing up the self-driving technology. Tesla claims its has the hardware to support full autonomy in its cars, when the regulatory agency greelights it. Then there are the automotive biggies GM, Ford, Volkswagen, Mercedes, BMW that are releasing advanced driver assistance technology in their cars.
Besides automotive vendors, tech leaders Google, Baidu, Uber, Lyft are disrupting the space with in-house AI hardware/ autonomous systems for car manufacturers. The year 2017 saw rapid development in autonomous tech with tech giants and carmakers testing cars on the streets of Detroit, Michigan with some success. In fact, Detroit carmaker Ford was the first vendor to start testing its Level 5 vehicles on the roads and even acqui-hired an AI startup for a whopping $1 million for its virtual driver system.
Industry reports indicate that while carmakers are following a cautious, step-by-step approach in implementing autonomous features like adaptive cruise control and object avoidance, tech giants, especially Google, Uber and even Baidu are pushing the lofty goal of Level 5 cars on the road, at least by 2021 – a move that would potentially open a new revenue stream in the form of ride-hailing services.
Despite the hubbub around self-driving technology, there are too may gray areas the carmakers are grappling with. Most automotive companies have only achieved SAE level 3 capabilities —which means the cars can only drive on expressways and not in the midst of traffic wherein a driver may be required to take control.
Meanwhile, Tesla, that has been working the longest in self-driving technology and has claims they are on level 5 and won’t need human driver assistance, besides setting a destination into the computer. Independent tech analyst Richard Windsor and founder of Radio Free Mobile sounded off the alarm earlier this year when he wrote that that he was “dubious of any automaker having genuinely autonomous vehicles by 2020, mainly because the liability issue is unresolved. “This is good news for the automotive industry that is notoriously slow to adapt to and implement new technology as it will have more time to defend its position against the new entrants,” he wrote earlier this year.
Self-driving cars not feasible until 2025, yet carmakers have announced optimistic deadlines
The industry view is that self-driving isn’t yet feasible on roads in all environments. A Reuters estimate notes that it will take around six years longer to achieve the final stage of autonomous vehicle development. Germany’s top auto component supplier Bosch officially declined to reveal when an autonomous car might become publicly available.
While some automotive companies believe they are already on the point of level 5, leading autonomous car development platform suppliers such as NVIDIA, Mobileye which are scaling AI technology have a different view. NVIDIA CEO Jen-Hsun Huang revealed at Bosch Connected World conference in March this year that he expected to have chips that will allow Level 3 automated driving, available by the end of this year. It was a view seconded by Bosch’s CEO Volkmar Denner who revealed a similar timeline for fully self-driving cars for mainstream customers — not before 2025. Yet leading carmakers have rolled out tighter deadlines – Audi markets self-driving by 2020, Nissan has reportedly pledged affordable self-driving cars by 2020, the leader of all Ford will roll out cars by 2021 while Tesla has again promised to have driverless tech ready by 2018 and anticipating regulatory approval by 2021.
The SAE International, a global association committed to advancing engineering, defines the levels of autonomy as:
- Level 3 – Conditional Automation wherein human intervention is required
- Level 4 – High Automation wherein the autonomous system can handle most dynamic driving task but human intervention may be requested
- Level 5 – Full Automation is the where the autonomous system can perform all dynamic driving tasks under all roadway and environmental conditions
As far as autonomous technology is concerned, the highest level achieved is Level 3
In 2016, Elon Musk backed Tesla shocked the world by announcing that the cars were outfitted with hardware required for autonomous system (cameras, sensors, radar), so that when the software catches up the cars being produced will be “self-driving” ready. Musk even proclaimed that 2017 would see a Tesla car driving itself without any human interaction. Since his famous announcement in 2016, Autopilot 2.0/Enhanced Autopilot software updates had been few and far between and the company’s top engineering brass leading the software development has seen several exits. In 2016, Tesla also braced with a deadly crash that killed a man driving a Tesla Model S while using Autopilot.
Hype vs Reality
- Here’s the reality — autonomous systems, especially Level 4 and 5 have very specific hardware requirements and the cars have to demonstrate product readiness in order to be safe and reliable.
- Most carmakers testing self-driving cars usually do it in a restricted environment and the mass-market use will be initially confined to safe environments such as downtowns, college campuses.
- With Level-3 enabled cars, buyers are getting a taste of some self-driving features. Drivers are allowed to switch to autopilot only on single-lane highways.
- According to an industry analyst, even if the cars achieve Level-5 and become production ready, it will be sometime before the regulatory authorities catch up to make it a legal reality.
- NVIDIA CEO told attendees at a conference that fully self-driving cars that can operate under all environments require enormous computing power. “No human could write enough code to capture the vast diversity and complexity that we do so easily, called driving,” he said.
- Even though automakers have rolled out optimistic timelines, they are aware the significant time-lag in obtaining the necessary sign-off from regulatory authorities.
- The major distinction between Level 4 and Level 5 full autonomy is that Level 4 does not cover every driving scenario. And in a rush to roll out Level 5, most companies are planning to skip the semi-autonomous system like Tesla’s Autopilot and GM’s Super Cruise (pegged as Hands-Free driving tech) to deploy at scale. Here’s a catch: L3 enabled vehicles which are on road are useful for collecting data which can be used to improve L4-5. By skipping L3-4, carmakers are taking a huge risk.
- With competition heating up in the space, companies are betting top dollars on driverless tech to be ahead of the game with aggressive marketing which markets latest tech front and centre. The game is led by Tesla, self-promoting itself with unrealistic deadlines and bold pronouncements.
- According to industry analysts who are watching the space, if one parses the message in marketing speak, it implies that some autonomous car components will be available by 2020, but the mass production of first-generation autonomous vehicles will follow later.
Try deep learning using MATLAB