Chinese web giant Baidu has accelerated development of its Project Apollo autonomous car platform, planning to freely share it with the world as the self-driving car race steps up a gear.
Self-driving cars have already found their way from the test track to the roads, but the jury is still out as to exactly when they’ll find mainstream acceptance around the world. Tesla founder Elon Musk predicts autonomous cars will dominate the roads within ten years, but China’s Baidu is a late starter which aims to race ahead of Telstra, Google’s Waymo platform and traditional car makers like Honda and Ford.
Baidu plans to have self-driving cars on the road next year, using machine learning to fast-track the maturity of its new Project Apollo self-driving car platform. It’s aiming for a wide-scale international rollout by 2020, reducing costs by improving accuracy to the point where smart cars can eventually rely on a single consumer-grade camera for navigation.
Project Apollo is designed around modular hardware to make it easier for any car manufacturer to adopt the platform, with plans to make the software and hardware reference designs freely available next month. It’s already begun to strike deals with car makers around the world.
The actions of Baidu’s street mapping car drivers were analysed to help train Project Apollo, plus it has spent time in driving simulators to accelerate its learning. This training, combined with machine learning, has helped bring Project Apollo up to speed quickly, says Gu Weihao, Baidu general manager of intelligent vehicles.
“We need to be tough on ourselves to achieve a better solution with cheaper sensors, in order to get Project Apollo on the road quickly and cost-effectively,” says Weihao – speaking at this week’s CES Asia technology show in Shanghai.
“The treasure trove of data we collected from our map collection drivers, combined with deep learning algorithms, has deliver a complete ecosystem which any company is welcome to download as use as a reference design.”
Baidu is also running autonomous car hackathons, inviting competitors to improve on Baidu’s algorithms and incorporating their advancements into the Project Apollo platform. The system’s on board processing also looks back over the last few minutes to predict what it is likely to encounter next, which has helped improve performance.
Baidu’s mapping cars, which criss-crossed China while tracking their driver’s actions to help train Project Apollo. Photo: Adam Turner
“The self-driving cars is going to be a huge market and we want to accelerate its progress,” Weihao says. “Project Apollo is improving every day and we’re on track to deliver a safe and secure self-driving ecosystem.”
By Adam Turner
Sydney Morning Herald