Uber on Monday detailed its visualization software for self-driving vehicles which the company uses to train its autonomous vehicles even when the thereof aren’t driving on actual roads. The solution allows Uber’s engineers to use the data the firm’s driverless cars have collected over the many miles they already spent on the road and either recreate some scenarios which they believe warrant further testing or tweak it and create entirely new conditions for challenging their creations. The approach allows the company to put its artificial intelligence (AI) algorithms used for navigating self-driving cars on actual roads to their very limits without risking any kind of incidents since the tests are entirely virtual. Uber didn’t go into many details regarding how the backend of its service works save for revealing that it’s based in the cloud and has instead chosen to disclose more information about the visualization aspect of the service.
The Data Visualization Team of the firm’s Advanced Technologies Group (ATG) designed its software with the goal of catering to a broad range of Uber’s engineering units, all of which have different data-related demands and hence cannot be maximally efficient by using an identical tool. Due to that state of affairs, the creators opted for a highly customizable service which can present, hide, and summarize vast amounts of data depending on its type and many other factors which are selected by users themselves, allowing them to not only run simulations with ease but also analyze them in a swift manner, consequently being able to quickly see what went wrong in any particular test and making sure such a scenario doesn’t happen in the real world. The software’s use cases are essentially limitless, though Uber’s engineers are largely employing it for creating extremely demanding driving environments for the company’s self-driving AI; even if such simulations break the algorithms and cause them to malfunction or make mistakes, these outcomes still generate extremely useful data that can be used for improving the AI in the future.
Uber’s detailed preview of its self-driving visualization software that can be seen below comes shortly after Alphabet’s Waymo provided the general public with a similarly in-depth look at its own visualization solutions which the Mountain View, California-based tech giant uses for essentially identical purposes. The two rivals are currently also embroiled in a legal battle over alleged trade secret theft which one of their former employees is said to have committed, with that case still being far from resolved.