developereconomics.com
There is growing consensus that fully autonomous cars will become a reality by 2020. Google self-driving cars have driven over 1.2 million miles. Elon Musk, Tesla CEO, predicted in September 2015 that Tesla cars will have fully autonomous capability in 3 years. Zvi Aviram, CEO of MobileEye, a supplier of self-driving systems to many car makers, expects their technology will support fully autonomous driving by 2019.
Most traditional car makers still see autonomous driving as a feature of the car, rather than a market shift that will open the path to the creation of a completely new winner-takes-all industry. It’s just like PC makers focusing on adding connectivity to their products and missing the transition to the Internet platforms (Google Search, Amazon, Facebook). Or telecom operators focusing on adding always-on fast data connectivity to their networks and missing the transition to the mobile platforms (Google Android, Apple iOS).
Is the same about to happen in the car industry? Are car makers about to miss the transition to transportation platforms in the same way as PC makers missed the transition to Internet platforms and telecom operators missed the transition to mobile platforms?
The future transportation value stack will be very different from the existing automotive industry. It quite remarkable that only two companies, Google and Uber, are present in all layers of the stack that are necessary for creating a dominant transportation-as-a-service platform.
The car hardware (the body, the power train, the wheels) increasingly becomes a commodity. Modern cars are good-enough for typical everyday use offering little opportunity for differentiation. Car commoditisation will only accelerate with the transition to electric vehicles. Electric vehicles are much simpler mechanically and easier to make, which opens the gates for new players, including such electronics and Internet services players like Apple, Google, LeTV and even Acer. It’s also notable that Tesla ‘open-sourced” their electric vehicle patents in 2014 pledging not initiate patent lawsuits against anyone who, in good faith, uses Tesla’s technology.
Autonomous driving is about guiding the car along the road, following the rules while avoiding obstacles and crashes. It involves lots of sensors, computing power and sophisticated software, but the most important part here is the ‘data’. Self-driving systems are machine learning systems that are trained to evaluate the environment and make fast decisions on how to react.
The ‘data’ represents all the collective experience learned by multiple cars driving in test and real-world conditions. The more cars you have on the road and the more miles these cars have driven in all possible conditions, the more experienced, safe and precise the self-driving system becomes. Google is undisputed leader here having its fleet of test cars driven over 1 million miles. Tesla’s Autopilot feature introduced in October 2015 on Model S cars will allow Tesla to start training its self-driving system in real-life conditions on tens of thousands of cars.
Uber seem to be behind in terms of putting real self-driving cars on the roads. The company poached 40 researchers and engineers from the Carnegie Mellon’s robotics lab in March 2015 and partnered with University of Arizona on optics research for self-driving cars.
Navigation is about figuring out which roads and streets the car should drive on in order to get from point A to point B. Google is again is a clear leader here with Google Maps and Waze. A consortium of German carmakers (Audi, BMW and Daimler) is trying to uphold an alternative acquiring the Here Maps business from Nokia in August 2015 for $3.1 Billion. Uber also works to create a proprietary mapping platform winning independence from Google and Here Maps. The company acquired San Jose-based deCarta in March 2015, absorbed part of Microsoft Bing mapping assets in June 2015 and has partnered with TomTomin November 2015 to use its mapping and traffic data. (Is Microsoft about to miss the huge opportunity in the future automotive and transportation markets?)
Fleet routing this is where it gets much more interesting. Self-driving cars combined with Uber-style on-demand services make individual car ownership less and less attractive. Some people even claim that hardware-as-a-service is the end game for Tesla. The shared usage models will turn car market into something that looks like a public transport platform, where operators will match in real-time the demand for transportation with the location and the capacity of self-driving vehicles. In other words, fleet guidance is about deciding in real-time where every car needs to go. Which car needs go to a specific pick up point? Shall the car drive to where the demand is expected in the coming 15 minutes? What is the optimal time to recharge or refuel? When and where to go to do the service and maintenance? Where to park, and more.
This is a very complex computational problem to solve at the scale required to support fleets of thousands of self-driving cars. Bill Gurley, one of Uber’s early investors, gives a glimpse into how difficult it is in his blog explaining why UberPool is the new Uber’s “Big Hairy Audacious Goal.” (BHAG). UberPool helps the company to build capabilities that will be directly relevant for the optimal routing of large autonomous fleets.
I’m sure Google is not standing still here as well. Being a machine learning company, it has the scale and the technical depth to become the leader in this space. Add to that real-time bidding capabilities with extremely complex optimisations that Google has mastered for its online ad business. One can even argue that building such transportation platform is the reason for Google’s interest in self-driving cars.
It’s very difficult to see how traditional car makers will be able to compete with software-centric companies in this space.
Finally, the transportation platform is the most intriguing part of the value stack. Moving people around Uber-style is not the only use for self-driving cars. What else can we do with the fully autonomous fleet of robotic vehicles, given that they don’t not have to look as Uber or Google cars of today? These robotic vehicles can be specialized delivery vehicles (see this Domino’s Pizza car as a hint for how they may look like), small delivery drones like Transwheel or StarShip or even autonomous motorbikes, like Motobot by Yamaha.
The number of possibilities and applications for autonomous transportation is mind boggling. No single company, even as nimble and well-funded as Google or Uber, will be able to address all possible needs and use cases by themselves. The recipe for addressing these yet to be known needs and use cases is in plain sight. It is a platform connecting vehicle manufacturers, vehicle operators, service providers and application developers with users (much like Google did with Android).
The platform will harvest permissionless innovation by startups and developers to discover and deploy newservices and applications we cannot even imagine today – in the same way that no one could predict Instagram, Snapchat or WeChat on smartphones. Uber already works with developers extending its service into a platform. Google also has a long history of relying on permissionless innovation by developers to win its competitive battles, from Google Maps to Android. It’s only natural that Google will use the same approach to dominate self-driving cars.
It’s still too early in the game to say which companies will dominate the future transportation market. One thing is a safe bet: The future transportation ecosystem will look very different from the existing automotive industry. It will resemble modern technology ecosystems with their platform business models, permissionless innovation by developers, and domination of software-centric companies.
No comments:
Post a Comment