Baidu’s Apollo Seeks to Democratize US Autonomous Driving
This July, Baidu unveiled its autonomous driving platform Apollo. Billed as “the Android of the auto industry,” Apollo aims at democratizing autonomous driving with open-source data and code. Baidu sees Silicon Valley as a hub of self-driving technologies and the most important region outside China for developing local autonomous driving communities.
Apollo gives developers access to a complete set of vehicle, hardware, software, and cloud data service solutions; as well as open-source codes for environment perception, route planning, vehicle control, and operating system. By following Apollo’s instructions, a software engineer could convert a Lincoln MKZ into a self-driving vehicle in about 48 hours.
Synced recently attended a Baidu meet-up at the company’s newly-opened Silicon Valley offices, where details were released on the latest iteration of the Apollo platform, Apollo 1.5.
The upgraded Apollo adds features such as end-to-end deep learning, obstacle perception, and route planning. Developers can now enable vehicles to run autonomously within fixed lanes, and to identify obstacles accurately during both day and night.
In addition to algorithms, Apollo 1.5 also adds cloud-based HD maps and a simulation engine backed by vast amounts of relevant data such as traffic lights, road signs and lane markers, to enable developers to test their algorithms.
Baidu values US talent with autonomous driving expertise. Since Apollo’s debut on GitHub, more than 1,300 companies have downloaded the Apollo source code and nearly 100 companies have applied for open data via the Apollo website. While 60% are from China, most of the others are from Silicon Valley.
Jingao Wang, Senior Director of Baidu’s Intelligent Driving Group and Head of the Apollo Project, says Baidu plans to build an Apollo ecosystem by pumping US$1.5 billion into 100 autonomous driving projects over the next three years. About 40% of that will go to US-based projects.
“The Apollo Fund is similar to IFund (a US$200 million capital fund encouraging developers to create applications, services, and components for Apple products). Startups can get funding for project launch, marketing, and sales, which could boost the ecosystem,” says Wang.
Another Apollo ecosystem initiative is education. Wang told Synced that Baidu is working on online education courses about autonomous driving, which would be similar to Udacity’s popular nano-degrees.
Attendees, especially local developers, seemed optimistic about Apollo’s American debut. Sebastian Brannstrom, Engineering Manager at Lyft’s Palo Alto self-driving team, said he was impressed by its potential for lowering R&D costs. “We are open to any partnership, and certainly Apollo is a very competitive player in the Silicon Valley.”
Emrah Adamey is Principle Research Scientist at California-based self-driving startup Voyage, which currently uses open-source tools like ROS (Robot Operating System). Adamey told Synced he thinks Apollo’s open-source platform is a good way to move autonomous driving forward, especially if it offers superior tools.
However, Georgia Institute of Technology student Alan (last name withheld) remained skeptical:“I wonder how data collaboration will actually work between Baidu and other autonomous driving projects?”
Apollo Senior Software Architect Jinghao Miao said Baidu is willing to install data collection devices in participating autonomous driving projects. “The majority of Apollo data is visual data for training perception sourced from Baidu, though we encourage developers to share data with us so we can help with labeling.”
The autonomous driving industry is expected to gestate ever-increasing business opportunities globally, and Miao believes Apollo can make a difference locally: “Autonomous driving is still in its early stages, but Baidu hopes Apollo can help individuals and small startups in the US overcome obstacles and generate innovations.”
Author: Tony Peng | Editor: Michael Sarazen