Energy-efficient autopilot (EcoPilot)

Research Project, 2022 – 2026

The goal of this project is developing an intelligent, energy efficient, autonomous driving system, hereafter termed as EcoPilot. Novel solutions will be sought to address multiple challenges that originate from the long list of system requirements. The EcoPilot shall incorporate an AI policy that is safe, energy-efficient, learning-based and real-time implementable. The policy shall autonomously make a tactical decision for choosing whether to follow the leading vehicle, whether to engage eco-driving by, e.g., slowing down when climbing uphill and speeding up when rolling downhill, whether to change lane, or take an emergency action. The policy shall simultaneously plan longitudinal and lateral control and energy use in a complex fuel-cell hybrid electric vehicle (FCHEV) powertrain. The policy shall account for uncertainty in environment, it shall learn from observations and shall make an interaction-aware prediction of surrounding road users to be able to adapt to developing traffic situations. The policy shall be real-time implementable, which is one of the major challenges throughout this project.
The idea is to investigate a combination of model-based and machine learning techniques to make the problem tractable, while maintaining interpretability.