Leaderboard Overview

As autonomous-driving technology advances rapidly, an efficient and fair Leaderboard system can encourage learners to exchange ideas and engage with leading models. A Leaderboard is not only a showcase of technical capability, but also a powerful driver of progress—from theoretical exploration to practical application—bringing together insights from around the world.

  • Perception Leaderboard:The computer-vision community already provides comprehensive Perception leaderboards, such as Waymo, NuScenes, and KITTI, so the Learning Center does not maintain a separate Perception leaderboard.

  • Prediction Leaderboard:Autonomous-driving trajectory prediction is a classic prediction task with several existing leaderboards for different data types, such as INTERACTION for strongly interactive data. The OnSite Learning Center focuses on vulnerable road users (VRUs) through theOnSiteVRU Trajectory Prediction Leaderboard, opening a new direction for trajectory prediction.

  • Decision-Making and Planning Leaderboard:The OnSite Learning Center Decision-Making and Planning Leaderboard is based on Track 1 of the Autonomous Driving Algorithm Challenge and provides urban and highway test cases for off-season use.

  • Control Leaderboard:Because of limitations in test environments and test cases, the OnSite Learning Center does not yet provide a Control leaderboard. We welcome your contribution if you are interested.

  • End-to-End Leaderboard:Because of limitations in test environments and test cases, the OnSite Learning Center does not yet provide an End-to-End leaderboard. We welcome your contribution if you are interested.