Uber aims to position its millions of drivers not just as a network that transports passengers, but as a massive sensor network that collects real-world data for autonomous vehicle companies. The company’s long-term plan includes placing sensor kits in human drivers’ vehicles and using the data obtained from these vehicles to train autonomous driving systems.

Uber Chief Technology Officer Praveen Neppalli Naga detailed this strategy at the TechCrunch StrictlyVC event in San Francisco. Naga states that this approach is a natural extension of the AV Labs program announced in January.

Real-world data collection for autonomous vehicles

In its current AV Labs program, Uber collects data with a fleet of small, sensor-equipped vehicles it operates. The company plans to expand this structure to millions of drivers around the world in the future.

Praveen Neppalli Naga emphasizes that the biggest obstacle in the autonomous vehicle development process is no longer basic technology, but data. Autonomous vehicle companies need real-world scenarios in different cities and traffic conditions.

Uber is building an infrastructure that can provide data from requested regions using its network of drivers already on the road. This structure can be an important resource for autonomous vehicles as well as other artificial intelligence models trained with physical world scenarios.

The company had previously withdrawn from its own autonomous vehicle development efforts. The new data strategy places Uber at the data layer of the autonomous vehicle ecosystem without making it a direct vehicle manufacturer.

Uber currently has partnerships with 25 autonomous vehicle companies, and within this scope, a data library called AV cloud is being created. Partner companies can use tagged sensor data to train their own autonomous driving models.

Simulation and future strategy

Partner companies can test the models they have trained on real Uber trips using a simulation logic called shadow mode. With this method, the performance of the model is measured on real journey data before the autonomous vehicle physically sets off on the road.

Praveen Neppalli Naga states that Uber does not see making money from this data as its main goal and that they want to democratize data. But having large-scale and private real-world sensor data could give Uber serious bargaining power in the autonomous vehicle market.

There are some regulatory issues ahead of the plan, such as technical and legal preparations. Before placing sensors in human drivers’ vehicles, Uber must clarify the working principles of its sensor kits and data sharing rules.

While companies like Waymo collect data with their own fleets, Uber’s extensive network of drivers around the world offers a distinct competitive advantage. With the right hardware and permissions, this network could become a large-scale data infrastructure for autonomous driving companies. How do you think Uber’s transformation of its drivers into data collection tools will affect the development of autonomous driving technologies?