In 2017 the Chinese search company Baidu released Apollo, a full autonomous-driving software stack, as open source. Where most self-driving efforts kept their code locked inside one company, Baidu published Apollo on GitHub and invited carmakers and suppliers to build on it. The project describes itself as “a high performance, flexible architecture which accelerates the development, testing, and deployment of Autonomous Vehicles.”
Apollo bundles the major pieces of an autonomy stack, perception, localization, prediction, planning, control, and decision-making, into one codebase. Early versions did little more than follow GPS waypoints, but the platform grew to support urban driving features such as lane changing, traffic-light recognition, and obstacle avoidance. The repository has drawn tens of thousands of stars and hundreds of contributors, and Baidu directs partners and businesses to its companion site at apollo.auto.
The strategic bet was that autonomy is too expensive and too hard for any single firm to solve alone, so the company that supplies the shared software layer gains leverage over the whole ecosystem, much as Android did for smartphones. Apollo became the technical foundation for Baidu’s own Apollo Go robotaxi service.
For a business reader, Apollo is a clear case study in platform strategy applied to a frontier technology: rather than racing to ship a finished car, Baidu tried to own the common infrastructure that other people’s cars would run on.