![](\"https:\/\/tctechcrunch2011.files.wordpress.com\/2017\/06\/cognata_day_time_control.png?w=738\")
<\/p>\n<\/p>\n
Rand Corporation reported in 2016 that autonomous cars would need to be tested over 11 billion miles<\/a> in order to prove that they\u2019re better drivers than humans. With a fleet of a hundred cars running 24 hours a day, that would take 500 years, according to the report. That seems both improbable and impractical, so companies like Cognata<\/a> are filling the autonomous vehicle testing void with virtual miles.<\/p>\n The need to establish the reliability of autonomous vehicles in short order has prompted Emerge Innovation Capital, Maniv Mobility, and Airbus Ventures to invest $5 million in funding in Cognata. The company\u2019s simulation system uses artificial intelligence, deep learning, and computer vision to create a virtual environment for testing cars without drivers \u2014 no roads required. It simulates the data collected by the vehicle\u2019s sensors when encountering other vehicles and a variety of road conditions.<\/p>\n Cognata recreates real cities in 3D, starting with San Francisco, with every building and tree in place. Then it layers in data, like traffic models from different cities, to understand how vehicles drive and react in each particular location. \u201cWe can see the difference between San Francisco, Bangalore, and London,\u201d said CEO Danny Atsmon in a phone interview. The third layer of Cognata\u2019s virtual test is the emulation of the 40 or so sensors that autonomous vehicles require, including cameras, lidar, and GPS. These simulations are run using the computers provided by the auto manufacturer or Tier 1 supplier in a far safer environment than, say, the street you live on.<\/p>\n