We are pleased to present a recent publication authored by researchers from the German Research Center for Artificial Intelligence (DFKI). The work, ARISE – Adaptive Refinement and Iterative Scenario Engineering, has been accepted for publication at the IEEE Intelligent Vehicles Symposium (IV) 2026, one of the leading international conferences in intelligent mobility and autonomous driving, taking place June 22–25, 2026, in Detroit, MI, United States. The publication introduces a powerful new framework designed to enhance the creation and validation of traffic scenarios for autonomous‑driving simulations.

Understanding the ARISE Approach

The ARISE framework offers a significant step forward in converting natural‑language descriptions into executable simulation scenarios. Unlike traditional single‑pass generation tools, ARISE uses an iterative Test‑and‑Repair Loop that checks each generated Scenic script for syntactic and functional correctness within a simulator—and automatically repairs errors when they occur. This ensures higher reliability, reduced manual intervention, and greater semantic alignment between the textual description and the final scenario.

Building on the ChatScene codebase, ARISE enhances semantic extraction, expands knowledge‑base snippet retrieval, and introduces fine‑grained scenario components such as weather conditions, object behaviors, and additional traffic participants. These improvements lead to more accurate and context‑aware synthetic scenarios, especially valuable for testing rare or safety‑critical situations.

Relevance for Intelligent Mobility and BERTHA

Accurate and executable traffic scenarios are essential for evaluating autonomous‑vehicle behaviour under diverse and challenging conditions. By automating scenario generation and strengthening robustness, ARISE supports scalable simulation workflows and contributes directly to research goals shared within the BERTHA project—particularly around safety, predictable system behaviour, and human‑centred mobility technologies.