Baidu Open Sources Its Autonomous Driving Technology

Baidu has announced that its new platform for building autonomous driving systems will be open sourced. Apart from driving progress in this segment, the move could improve defenses against interference from outside the car network.

Apollo, the new project from the Chinese web services company, integrates all the components partners need to create their own driving systems, from software and hardware platforms to cloud data services. Testing tools and knowledge of the different functions of the car (obstacle perception, trajectory planning, vehicle control and operating systems) are also freely available.

This collaborative ecosystem envisioned for the Apollo project would expose Baidu’s software and hardware contribution, as well as assets from partners (sensors, compatible vehicles), to assessment from entities with varying levels of experience with intelligent driving and its safety. The endgame is to introduce self-driving cars on highways and city roads by 2020. Such vehicles have been tested successfully in traffic by Google’s Waymo, Uber, and Baidu. The next step before this technology becomes mainstream is to eliminate cyber security concerns.

Researchers have exposed the vulnerability of computer systems in multiple car brands, either through responsible disclosure of the bugs or by demonstrating the efficiency of their exploits. Undoubtedly, hacking a car and taking control over it today requires some technical understanding and significant research but, as details about the cars’ digital innards become available to a wider public, security holes can be spotted and fixed faster.

Security of systems controlling autonomous cars is even more critical since everything is computer-assisted and needs to connect to an outside service, at least to get map updates or real-time traffic information. All this extends the attack surface. A person cannot always be counted on to intervene manually in stopping an attack because passengers may completely trust the car’s decisions and act late, or they can be distracted and miss the signs.

A benefit of the open source code model is the contribution and free inspection from multiple parties. This helps improve the project on all fronts, from functionality to security and hardware compatibility, improvements that benefit the entire community. Baidu is not the first player to share its code and hardware with others in the autonomous driving sector: George Hotz, the hacker famed for unlocking the iPhone, has released to the public his company’s self-driving software; Udacity does the same with its fully autonomous vehicle project.

Qi Lu, Group President and Chief Operating Officer at Baidu, said in a press release announcing Apollo that the company “took the initiative to open our autonomous driving technology to the industry in order to encourage greater innovation and opportunities, making better use of our technology to drive the evolution of the entire industry. An open, innovative industry ecosystem initiated by Baidu will accelerate the development of autonomous driving in the US and other developed automotive markets.”