Expand and prune improves candidate options for safer self driving
Expand-and-Prune: Test-Time Candidate-Space Adaptation for End-to-End Autonomous Driving
RoboticsComputer Vision and Pattern Recognition
Summary
Self-driving cars plan their paths by considering a limited number of possible routes, but when the set of options is too small, the car might miss important maneuvers. The authors created a method called Expand-and-Prune that adds more candidate routes based on driving intent and past predictions, then removes less promising ones based on uncertainty. This approach improves the car’s decision-making without retraining the system. Tests show it works better than previous methods, especially in new or unexpected driving situations.
What this means in practice
- •For autonomous vehicle engineers: Enhance autonomous vehicle planners by dynamically expanding route options and pruning candidates to improve driving decisions under uncertainty.
- •For robotic navigation developers: Incorporate adaptive candidate management to improve trajectory planning in robotic systems that operate in changing or uncertain environments.
Authors
Fabian Schmidt, Karol Fedurko, Markus Enzweiler, Abhinav Valada
Abstract
Candidate-based end-to-end driving planners reason over a finite set of trajectory hypotheses. However, for sparse planners, a limited candidate set may provide insufficient support for scene-relevant maneuvers, while fixed downstream candidate budgets do not account for prediction uncertainty. To address these challenges, we introduce Expand-and-Prune, a test-time framework that jointly adapts candidate support and capacity without updating learned parameters. The framework expands the native candidate set using semantic driving intent and temporally propagated predictions and then uses uncertainty in intermediate trajectory scores to adaptively prune the resulting candidates. Across four heterogeneous candidate-based planners and four benchmarks, Expand-and-Prune consistently improves planning performance. It achieves state-of-the-art results on LangAuto, increasing the Driving Score from 68.2 to 75.9, while providing further improvement on Bench2Drive, Fail2Drive, and NAVSIMv2, with particularly strong improvements in out-of-distribution scenarios. Moreover, in its full instantiation, 72.5% of selected trajectories originate from the newly introduced intent-guided or temporal candidates, confirming that candidate-space expansion provides useful hypotheses beyond those generated natively by the planner. The code is available at https://github.com/iis-esslingen/Expand-and-Prune.