Hidden Biases of End-to-End Driving Models

End-to-end driving systems have recently made rapid progress, in particular on CARLA. Independent of their major contribution, they introduce changes to minor system components. Consequently, the source of improvements is unclear. We identify two biases that recur in nearly all state-of-the-art methods and are critical for the observed progress on CARLA: (1) lateral recovery via a strong inductive bias towards target point following, and (2) longitudinal averaging of multimodal waypoint predictions for slowing down. We investigate the drawbacks of these biases and identify principled alternatives. By incorporating our insights, we develop TF++, a simple end-to-end method that ranks first on the Longest6 and LAV benchmarks, gaining 11 driving score over the best prior work on Longest6.

PDF Abstract ICCV 2023 PDF ICCV 2023 Abstract


Results from the Paper

Task Dataset Model Metric Name Metric Value Global Rank Benchmark
CARLA MAP Leaderboard CARLA Map TF++ Driving score 61.17 # 1
Route completion 81.81 # 2
Infraction penalty 0.70 # 3
CARLA longest6 CARLA TransFuser++ WP (TF++WP) Driving Score 73 # 1
Route Completion 97 # 1
Infraction Score 0.56 # 12
CARLA longest6 CARLA TransFuser++ (TF++) Driving Score 69 # 3
Route Completion 94 # 3
Infraction Score 0.72 # 4
Autonomous Driving CARLA Leaderboard TF++ WP Driving Score 66.32 # 4
Route Completion 78.57 # 7
Infraction penalty 0.84 # 3