RECTOR: Priority-Aware Rule-Based Reranking for Compliance-Aware Autonomous Driving Trajectory Selection
arXiv:2605.25095v1 Announce Type: cross Abstract: Autonomous driving stacks must pick one trajectory from a multi-modal candidate set; choosing by model confidence ignores safety, traffic-law, and comfort constraints. We present \textsc{RECTOR} (Rule-Enforced Constrained Trajectory Orchestrator), a post-generation reranking layer that scores candidates against a tiered rulebook (Safety~$\succ$~Legal~$\succ$~Road~$\succ$~Comfort) via differentiable proxies and a scene-conditioned applicability mechanism, then selects with a deterministic $\varepsilon$-lexicographic rule that preserves cross-tie
The increasing maturity of autonomous driving systems is pushing the need for robust, safety-critical decision-making layers beyond basic model confidence, especially as these systems near wider deployment.
This development addresses a critical gap in autonomous driving: how to reliably integrate complex, hierarchical rules (safety, legal, comfort) into real-time trajectory selection, enhancing trustworthiness and regulatory acceptance.
Autonomous driving systems can now incorporate a more sophisticated, rule-based reranking layer, moving beyond simple model outputs to prioritize critical constraints deterministically.
- · Autonomous Driving Developers
- · Automotive OEMs
- · AI Safety Researchers
- · Regulatory Bodies
- · Companies relying solely on end-to-end black-box models for AD decision-making
- · Traditional AD testing methodologies
Improved safety and reliability metrics for autonomous vehicles, leading to faster development cycles and reduced accident rates.
Increased public and regulatory trust in autonomous driving technology, potentially accelerating widespread adoption and reducing insurance premiums.
This rule-based, hierarchical decision-making framework could inspire similar architectures in other safety-critical AI applications, such as medical robotics or industrial automation, where compliance and explicit constraints are paramount.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG