HetroD Challenge

Traffic Simulation in Heterogeneous Traffic

A benchmark for realistic future trajectory simulation in complex multi-agent urban scenes.

Back to Home

Description

This challenge focuses on multi-agent traffic simulation in highly heterogeneous urban environments, based on HetroD dataset. The goal is to evaluate models' ability to predict and simulate realistic future trajectories in complex traffic scenes involving vehicles, scooters, cyclists, and pedestrians.

Target

Given 1.1 seconds of historical observations for all agents in a scene, participants are required to predict 8 seconds of future multi-agent trajectories.

Evaluation Metrics

1. Kinematic Realism

  • Linear Speed
  • Linear Acceleration
  • Angular Speed
  • Angular Acceleration

2. Safety and Rule Compliance

  • Collision Rate - percentage of trajectories resulting in inter-agent collisions
  • Off-road Rate - percentage of trajectories leaving the drivable area
  • Traffic Violation Rate - violations such as red-light running or illegal maneuvers

3. Trajectory Accuracy

  • Average ADE (Average Displacement Error)
  • Min ADE

Leaderboard (Template)

Placeholder

Add official submissions and scores when the benchmark server is ready.

Rank Team Method Kinematic Score Safety Score Trajectory Score Overall
1 Baseline Team ExampleNet-v1 -- -- -- --
2 Pending Pending -- -- -- --
3 Pending Pending -- -- -- --