This project integrates concepts from VALAWAI and SymAware projects to develop a unified framework for value- and risk-aware decision-making in autonomous vehicles. By combining machine learning and Large Language Models (LLMs), the project aims to create personalised driver models that balance safety requirements with individual preferences like speed patterns, route choices, and risk tolerance. For example, one driver may prioritise user comfort over speed in one context, or speed over eco-driving in another. The novelty of this booster project will be in integrating drivers' preferred values, while prioritising functional priorities, like safety.

Key activities include:

- Driving style learning and modelling: We will define personalised values, specified through driving styles, along with risks in a measurable way by training machine-learning models based on historical data, and using real-time LLM-mediated interactions to extract context-related data from driver.

- Hybrid optimisation: We will develop decision-making algorithms that reconcile functional priorities (safety) with personalised values, addressing scenarios where preferences may enhance or compromise safety.

- Validation: We will implement prototype simulations to assess improvements in user satisfaction and safety compliance.