DeFT: Maintaining Determinism and Extracting Unit Tests for Autonomous Driving Planning

This repository corresponds to the ICSE 2026 Research Track paper and its accompanying artifact, DeFT, a tool and methodology designed to improve testing reliability in autonomous driving systems by addressing non-determinism in planning tests. Traditional system-level scenario tests often produce varying outcomes, making failure reproduction and debugging challenging. DeFT is a methodology that converts non-deterministic system-level scenario tests into deterministic module-level tests by extracting and reconstructing module inputs.