WASABI: Whole-graph Assignment-based Stabilizer for lAne topology By Inter-frame tracking

2026-07-23 · Source: cs.CV updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Software Development & Engineering · Depth: Expert, extended

Summary

WASABI is a real-time post-processing pipeline designed to stabilize lane topology outputs from perception models in autonomous driving systems. It addresses structural instabilities like missed detections, incorrect inter-lane connectivity (LCLC), over-detection, and label flicker. Operating under onboard real-time constraints (10 Hz / 20 ms / up to 200 input lanes), WASABI treats lane segments and their LCLC connectivity as joint tracking targets. The pipeline integrates segment tracking with connectivity, noise-robust topology-aware refinement, and a resource-constrained design. On internal validation data (16 sequences), WASABI improved LCLC detection F1 from 0.834 to 0.948 (+0.114, +13.6%) and reduced centerline lateral error from 2.50 m to 0.95 m. It also cut detection false-positives by 24.6%, reduced LCLC toggle rate by 63.3%, and boundary-label flicker rate by 30.2%, confirming significant across-frame stabilization.

Key takeaway

For autonomous driving engineers developing perception stacks, WASABI offers a robust post-processing solution. You should consider integrating joint lane segment and LCLC tracking to mitigate neural perception instabilities. This approach significantly reduces false positives and improves temporal consistency, leading to more reliable downstream planning. Implement resource-constrained designs and approximate Fréchet distance to meet real-time onboard requirements.

Key insights

WASABI stabilizes autonomous driving lane topology by jointly tracking segments and connectivity with noise-robust, real-time post-processing.

Principles

• Jointly track lane segments and LCLC connectivity.
• Integrate temporal history for output stability.
• Apply noise-robust topology-aware refinement.

Method

WASABI processes ego-motion compensated lane topology through frame-internal preprocessing, cost computation/association, state update via tracking buffer, output selection/correction, and conversion to NodeEdgeTopology.

In practice

• Use same-index Fréchet approximation for curve matching.
• Implement observation-adaptive miss budget for track states.
• Apply sliding-window majority vote for LCLC validity.

Topics

• Autonomous Driving
• Lane Topology
• Multi-Object Tracking
• Real-time Systems
• Fréchet Distance
• Perception Models

Best for: Research Scientist, AI Scientist, Computer Vision Engineer, Machine Learning Engineer

Related on AIssential

• WASABI: Whole-graph Assignment-based Stabilizer for lAne topology By Inter-frame tracking — WASABI stabilizes autonomous driving lane topology by jointly tracking lane segments and their inter-lane connectivity in real-time.
• WASABI: Whole-graph Assignment-based Stabilizer for lAne topology By Inter-frame tracking — WASABI stabilizes autonomous driving lane topology perception by jointly tracking lane segments and their connectivity in real-time.
• GraphBEV++: Multi-Modal Feature Alignment for Autonomous Driving — GraphBEV++ systematically mitigates multi-modal feature misalignment in BEV perception for robust autonomous driving.
• Geometry-Guided Representations for Coherent Lane and Traffic Topology Reasoning in Driving Scenes — Relational modeling, integrating geometric and contextual cues, significantly improves autonomous driving scene perception and topology reasoning.
• Time-to-Collision Based Dynamic Obstacle Avoidance Using Pretrained Vision Models for Robots in Unstructured Environments — Pretrained vision models enable data-efficient, interpretable dynamic obstacle avoidance for robots in unstructured environments.
• CLEAR: Cognition and Latent Evaluation for Adaptive Routing in End-to-End Autonomous Driving — CLEAR achieves efficient, multi-modal autonomous driving by combining single-step generative planning with LLM-driven cognitive reasoning.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.