The importance of data analysis in autonomous vehicle development
The road to Level 5 autonomy through autonomous vehicle development
You may know that auto makers are already running tests on autonomous cars, literally racing to be leaders with this new technology. Global investors have already poured billions of dollars into research and development of self-driving car technology over the last four years, with funding of autonomous vehicle technology outpacing the rest of automotive tech1. Clearly this is big business.
In the race to reach what the Society of Automotive Engineers (SAE) calls Level 5 autonomy—fully autonomous vehicles— automakers must contend with the challenges of training vehicles to drive better than a human. This drives cost and time in an already costly journey. While critical to both safety and legal certification, there is no opportunity for failure here. However, reducing the time and cost of autonomous vehicle development is critical to effectively compete in the race to bring self-driving vehicles to market. Manufacturers must look at new options if they are going to keep up the pace.
Using autonomous vehicle data
Autonomous vehicles combine a variety of sensors to perceive their surroundings, including radar, lidar, computer vision, sonar, and GPS, among others. These sensors interpret sensory information to identity navigation paths, avoid obstacles and read relevant markers, like road signs. In multiple locations around the world autonomous vehicle development teams run tests that take thousands of hours of test drive data. One eight-hour shift can create more than 100 terabytes of data. This massive amount of data must be collected, offloaded, stored, and interpreted for algorithmic training to build vehicle decision-making.
The big challenge: How to efficiently manage all the data that gets generated during the tests and teach the vehicle how to make decisions faster in very diverse conditions …even a moral dilemma. And how do you teach the vehicle to make an adjustment it’s not been trained to make when an unexpected issue in the real world becomes an event that should change the vehicles behavior?
Self driving car development must effectively use a staggering amount of data
The development challenge for teams seeking SAE Level 5 autonomy is to collect and store a staggering amount of sensory data, and to ultimately analyze and interpret this data to produce control systems that perceive information and accurately navigate the roads it encounters. Modern self-driving car technology uses algorithms to fuse data from sensors and other sources to drive accurately and without incident. Incorporating machine learning and artificial intelligence (AI) into building vehicle autonomy requires ongoing application and evolving expertise.
Who will win the autonomous driving race?
The pace of autonomous vehicle development depends on research and development and technical ability to collect, store and analyze massive amounts of sensor data to not only convert data into algorithms that improve autonomous driving accuracy and performance, but to develop intelligent autonomy. Automakers who can accelerate this process will gain a winning edge in the race to SAE Level 5 autonomy.
Learn about our autonomous vehicle development offering, DXC Robotic Drive.
