DXC-Luxoft

Solution

DXC-Luxoft built scenARI.Lux as a generative AI chatbot application for autonomous-driving validation. The system ingests human-readable traffic situation descriptions and routes them through multiple bots coordinated by a dedicated agent application so the workflow can support different LLM foundation models. To improve answer quality and latency, the pipeline uses Retrieval-Augmented Generation, feeding OpenX standards definitions into the process as an authoritative knowledge base. DXC-Luxoft also fine-tunes the underlying model with manually created scenarios and, optionally, client scenarios. The application generates ASAM OpenDRIVE and ASAM OpenSCENARIO DSL/XML outputs that simulators can execute, turning free-text inputs into standardized scenario artifacts for virtual testing. DXC-Luxoft positions those generated scenarios as inputs to its broader autonomous-driving validation toolchain, including collaboration with MicroVision for real-world scenario abstraction.

Data flow

Human-readable traffic descriptions are submitted to scenARI.Lux, where multiple bots coordinated by a dedicated agent application process the input with LLMs. RAG injects OpenX standards definitions, and fine-tuning on manually created and optional client scenarios improves generation. The system then emits ASAM OpenDRIVE and ASAM OpenSCENARIO DSL/XML files for simulators and downstream virtual-validation workflows.