ROS is a large project with many ancestors and contributors. The need for an open-ended collaboration framework was felt by many people in the robotics research community, and many projects have been created towards this goal.
Various efforts at Stanford University in the mid-2000s involving integrative, embodied AI, such as the STanford AI Robot (STAIR) and the Personal Robots (PR) program, created in-house prototypes of flexible, dynamic software systems intended for robotics use. In 2007, Willow Garage, a nearby visionary robotics incubator, provided significant resources to extend these concepts much further and create well-tested implementations. The effort was boosted by countless researchers who contributed their time and expertise to both the core ROS ideas and to its fundamental software packages. Throughout, the software was developed in the open using the permissive BSD open-source license, and gradually has become a widely-used platform in the robotics research community.
From the start, ROS was developed at multiple institutions and for multiple robots, including many institutions who received PR2 robots from Willow Garage. Although it would have been far simpler for all contributors to place their code on the same servers, over the years, the "federated" model has emerged as one of the great strengths of the ROS ecosystem. Any group can start their own ROS code repository on their own servers, and they maintain full ownership and control of it. They don't need anyone's permission. If they choose to make their repository publicly available, they can receive the recognition and credit they deserve for their achievements, and benefit from specific technical feedback and improvements like all open source software projects.
The ROS ecosystem now consists of tens of thousands of users worldwide, working in domains ranging from tabletop hobby projects to large industrial automation systems.