1: Mapping & Autonomous cars

The age of autonomy has dawned. It all started with the DARPA once again. The organisation has been known to create next frontiers for the United states and the world right from the INTERNET. The next next frontier seem to be autonomous cars. It all started in 2005 when Stanford's Stanley, a car that won the DARPA grand challenge navigating all by itself just with GPS coordinates in the middle of nowhere. In 2007, the BOSS won the grand challenge for autonomous city driving. Driving around in controlled city like conditions, the car followed all parameters to victory. 

Enter the dragon:Google. Google recruited all the team members in the victorious teams mainly, Sebastian Thurn & Chris Urmson. Having mapped the whole world and beyond(literally: Google earth) google was running out of ambition. The self Driving car seemed to be the perfect fodder for the dragon. This concept has the potential to transform the world just like the internet did. It can reshape current transportation, insurance, healthcare, automobiles, energy, urban landscape  to name a few industry segments changing the whole economy in the process. Business Insider estimates 10 million self driving cars by 2020. 

Source: Forbes

The first step in this radical shift will be mapping. Mapping involves using huge amount of sensor data to recreate a virtual world for the car to understand its surrounding and take meaningful decisions every millisecond. The maps along with GPS gives the car a sense of where it is in the world and what its next action should be. 

Monocular camera footage is matched to prior map data by Robot Car (Image: University of Oxford)

We are going to discuss various strategies used by Google, Tesla, Uber, Mercedes, BMW, Audi and others to accomplish this task of mapping the world in High Definition and detail.

2: European consortium & its war chest

With the advent of Google into the auto industry and possible advent of Apple, both with 100s of billions of dollars in cash, the German auto makers are scrambling to counter the threat to its competitiveness.

The main step in that direction is the buying of Nokia's here maps division by a consortium of car makers like Audi, Daimler & BMW. This gives the makers access to high quality maps of nokia and also the ability to own the data. This has been widely viewed as a key enabler in self driving vehicles.

Nokia Here maps

University of Oxford is also among the many groups involved in mapping research for self driving cars. Thus mapping will be a key area of importance for automakers.

Source: University of Oxford

A demo of the university's research efforts show a demo of the visual guided navigation capabilities of its car.

It is essential to note that German automakers have made the hardware ready by means of the car and working on solving other issues like mapping data. Audi and BMW have also committed to making self driving a reality and so is Toyota.

3: Mapping & Ridesharing

Ride sharing services have spread across the globe and is changing the way people commute short distances mostly within urban areas. Services like Uber, Lyft among others have based their business models on drivers. Recent developments in self driving technology can be disruptive with the involvement of these ride sharing services. Uber recently announced the opening of research facility at Carnegie Mellon, home to "The boss" which won the DARPA urban challenge in 2007.

Uber also drives around 2+ million miles(uber is private company,so a guesstimate) which means simply creating a hardware that can easily be retrofitted onto cars will make the mapping of the world's roads a relatively easier task. This theory is also reinforced by the fact that Microsoft just sold its data acquisition task of maps to uber, while also investing in Uber. Thus mapping using crowd-sourcing can be a very powerful way to collect data for processing as shown by Tesla.

Uber now has arguably one of the top research groups in self driving cars, and massive funding to leverage that potential to market. With this infrastructure to collect data, uber and other ride sharing firms have an enormous advantage to collect data quickly.

TomTom another mapping service provider announced their capabilities in mapping for autonomous cars. A 3D border to border representation of a road by tomtom is shown below.

Carlos Ghosn recently about autonomous cars how auto industry will strive to avoid becoming a commodity much like OS driven smartphone industry took the established hardware makers by storm. Thus Renault-Nissan, Daimler and others are working at collecting data and becoming less reliant on established players like google to provide service to self driving cars mainly mapping data.

4: Google - the 800 pound gorilla

Enter the dragon. Right from the beginning of the DARPA grand challenge days Google was around trying out the self-driving cars as one of the many  moonshots it indulges in. After recruiting Sebastian thurn, Chris Urmson among others it became increasingly feasible. It announced its intention to bring the product to market, changing the moonshot into a feasible business proposition.

Already google is a leader in mapping with its Google maps, earlier known as keyhole. It is the 800 pound gorilla not just because of its enormous engineering talent and universal reach but because it can potentially give its product for free or ridiculously cheap. The $64.4 billion dollars in cash reserve certainly can help in this regard.

The google self driving prototypes have been around quite a while but the amount of data and machine learning that went into this is staggering. Chris Urmson explains this in his TED talk above.

The google car senses the surroundings and is able to identify cars from pedestrians, bikes, and road signs. It is an amazing engineering feat whereby the car examines the road and surroundings and decides when to move, and in what direction.The car needs to be driven around first so that the car maps the environment in detail and notice how a driver has performed in the following conditions. Then it tries to emulate that. The pedestrians, bikes, cars, incoming & ongoing, lane marking, stop signs, and pedestrian crossing data is fed to the car and it learns them eventually. 

There are many who doubt the gradual evolution of autonomous vehicles to self driving and some who declare that a step towards autonomy should be slow paced due to regulatory and liability concerns. The following site by Dr. Alexander Hars proves to be a good resource for people interested in learning more about self driving cars. 

5: Tesla & its Silent task master

Ever since the task master stepped into the auto industry its time for some massive disruption. Elon musk has disrupted the banking industry with Paypal, aerospace industry with SpaceX, and the auto industry with Tesla. He has already created quite a stir in by pushing every notable car company to introduce an electric car. Now he has set to disrupt many other with autonomous cars.

This is his announcement about the autopilot feature in his flagship model S in October 2014. What no one except may be this post guessed was what he was actually announcing. It was a crowd-sourced data collection mechanism for autonomous driving.

Source: Tesla motors

Approximately a year later Tesla motors have announced the result of such an action. 

Detailed map of San Francisco Bay area

Source: Tesla motors

Since all the cars are connected to the internet all the time, the data can be collected and uploaded to the central server. The engineers can then analyze the data and simulate driving conditions to test the viability of the system in place. The data collected by one car can then be distributed to all the cars. This approach alone can transform the data collection tasks for autonomous driving quickly and easily. 

Tesla can now add 1 million miles of new data every day and uses machine learning to help the car choose what action it must perform. 

6: End of mapping & era of machine learning

Thus mapping has been identified as of one of the key enablers of self driving cars. Data collection and Data analytics is a crucial part of the process. Google and Tesla have already started collecting data by the hundreds of thousands of miles. Nokia's here maps is said to have 1000s of miles of data. Thus crowd-sourcing of data collection operation can also be an effective strategy to accomplish the goal apart from the traditional method.

Tesla has announced that it is open to sell its collected data to any automaker. While it is unclear who else will be the provider of valuable data now, possessing such information is critical.

Sebastian Thurn of Google has explained the machine learning activities done by google here.

The data collected is then analysed to discern value out of it. Google has already been vocal of its methods used in  machine learning. Given the availability of hardware at albeit a high price & the research activity in machine learning, all car makers are attempting to enter the self driving race in a substantial manner.

It is clear by now that self driving cars are a reality and increasingly the value delivered in the automobile industry is going to be software driven. This trend enables entrepreneurs to enter into the field of automobile just like personal computer & internet did.

Thus self driving cars are going to be a reality within 10 years and it is going to create a profound impact on the economics of the entire world. Mapping will be one of the key differentiators in the service offered, the safety & reliability provided by automakers and other innovative providers of mobility.