Waymo to test vehicles in Florida and share its data set
Waymo announced that it is bringing both its self-driving vehicles—the Chrysler Pacificas and a Jaguar I-Pace—to Florida this month to begin heavy rain testing.
Heavy rain can create noise for sensors, and wet roads may result in other road users behaving differently. Testing will also allow researchers to understand the unique driving conditions and to see how rain affects Waymo’s vehicle movements.
First, the company is spending several weeks driving on a closed course in Naples, where it will test its sensor suite, which includes LiDAR, cameras, and radar, during the rainy season. Then, the vehicles will be brought to public roads in Miami, where they will be manually operated by trained test drivers to collect data in real-world driving situations in heavy rain. Florida residents also will start seeing a few of the vehicles on highways between Orlando, Tampa, Fort Myers, and Miami, as Waymo gathers information about Florida roads.
Waymo also announced that it will share its open dataset, which has been gathered over 10 million autonomous miles in 25 cities, with the research community.
The Waymo Open Dataset, a multimodal sensor dataset for autonomous driving, is available free to researchers at waymo.com/open. The dataset is comprised of high-resolution sensor data collected by Waymo self-driving vehicles. The release contains data from 1000 driving segments, each of which captures 20 seconds of continuous driving, corresponding to 200,000 frames at 10 Hz per sensor.
It covers dense urban and suburban environments across Phoenix, Kirkland, WA, Mountain View, CA, and San Francisco, over a wide spectrum of driving conditions (day and night, dawn and dusk, and sun and rain).
Each segment contains sensor data from five high-resolution Waymo LiDARs and five front-and-side-facing cameras. The dataset includes LiDAR frames and images with vehicles, pedestrians, cyclists, and signage labeled, capturing a total of 12 million 3D labels and 1.2 million 2D labels.
Waymo also is working on 3D perception models that fuse data from multiple cameras and LiDAR. The entire self-driving system—including hardware and software—is designed to work seamlessly together. This encompasses choice of sensor placement and high-quality temporal synchronization.
Waymo says the data has the potential to help researchers make advances in 2D and 3D perception, domain adaptation, scene understanding, and behavior prediction. Waymo anticipates that the research community will generate further directions with its data that will not only help to make self-driving vehicles more capable, but also will impact other fields and applications, such as computer vision and robotics.