May Mobility has announced MARS, a new dataset it has jointly developed in partnership with the NYU Tandon School of Engineering. Through it, the partners aim to help OEMs and suppliers accelerate the development of new AV technologies by providing researchers with a wide pool of real-world driving data captured from multiple vehicles over repeated trips.
Standing for MultiAgent, multitraveRSal, the dataset was collected from four May Mobility autonomous Toyota Sienna vehicles that operated within a 12-mile (20-kilometer) zone encompassing residential, commercial, and university areas in a U.S. city. The AV company’s subscription service, FleetAPI, provides access to real-time and historical data from its vehicles – allowing data partners to access real-world information including sensor data (LiDAR, Camera), GPS/IMU, vehicle state, and more.
Work on MARS began in November 2022, when NYU Tandon started planning it with May Mobility. Since then, the school’s researchers worked closely with May Mobility’s engineering to access the studied fleet group’s daily operational sensor data, selecting more than 1.4 million frames of synchronized sensor data. This data included scenarios where multiple vehicles encountered one another on the road, providing valuable data on how autonomous vehicles might cooperate and communicate in the future. The multitraversal nature of MARS allowed for this data to capture a broad range of conditions, with the engineers and researchers identifying 67 locations along the route and collecting data from thousands of passes through these areas at different times of day, in varying weather conditions.
Because the dataset is collected from multiple commercial vehicles operating live in city streets, it can play a key role in training and validating the AI systems necessary to power AVs. To demonstrate its potential, the NYU Tandon team conducted initial experiments in visual place recognition and 3D scene reconstruction. The team highlighted that these tasks are fundamental to an AV’s ability to locate itself and understand its surroundings.
Through their collaboration, May Mobility and NYU Tandon are ultimately looking to create safer mobility, while improving the accuracy and effectiveness of autonomous driving algorithms.
