Skip to main content
Back to all posts

March 31, 2025

Announcing the 2025 Waymo Open Dataset Challenges

  • Technology
2025 WOD Challenges

For years, the Waymo Open Dataset and associated Challenges have advanced research in autonomous driving, robotics, computer vision, and machine learning. Today, we are thrilled to announce our 6th round of Challenges, continuing to guide the community toward impactful innovations.

Autonomous vehicles represent the most mature application of AI in the physical world. Since the launch of the Waymo Open Dataset in 2019, it has become a key resource for over 53,000 researchers in machine learning and robotics around the world. To date, more than 4,000 academic papers have been published either citing the dataset or using it as a benchmark.

Today, we are excited to kick off the 2025 Waymo Open Dataset Challenges, which will run from March 31 to May 22, 2025. The 2025 Challenges will focus on four key areas of research:

  • Vision-based End-to-End Driving: Use camera inputs to predict optimal driving behavior in challenging situations.

  • Scenario Generation: Develop generative models to add agents to a scene in order to create realistic traffic scenarios.

  • Interaction Prediction: Forecast complex agent interaction outcomes. 

  • Sim Agents: Generate realistic future behaviors for all agents in a scene using historical trajectory data, with the option of leveraging our Waymax simulator.

This year, we are particularly excited to host the Vision-based End-to-End Driving Challenge for the first time, offering a unique opportunity for participants to explore innovative AI techniques in this rapidly evolving research area. It’s important to note that, while this research Challenge focuses on a singular sensor modality, Waymo remains committed to utilizing a diverse sensor suite as we firmly believe it’s essential for autonomous driving in real-world conditions.  

Alongside the Vision-based End-to-End Driving Challenge, we are expanding the dataset with 5,000 new segments. This camera dataset extension is focused on fostering research around long-tail driving scenarios – some of the most challenging or rare driving situations that drivers encounter in different environments, such as construction zones during marathons, pedestrians falling off scooters, and unexpected obstacles on freeways. 

The remaining Challenges feature updated metrics based on our learnings from past Challenges. We’re excited to see how the research community pushes the boundaries of what’s possible this year. 

Examples of long-tail scenarios included in the End-to-End Driving dataset

The 2025 Challenges come with a total of $72,000 in cash prizes. First-place winners in each Challenge will receive $10,000, while second and third place winners will be awarded $5,000 and $3,000, respectively. In addition, standout submissions may be invited to present at the Workshop on Autonomous Driving at CVPR in June 2025 that Waymo is co-organizing. 

The Challenges will close at 11:59 PM Pacific on May 22, 2025, but leaderboards will remain open for continued research contributions. Full competition rules and eligibility criteria are available here.

Good luck to all participants, and as always, we welcome your feedback!