Learning-Based Frameworks for Minimizing Pollutant Emissions in Hybrid Electric Vehicles for Dynamic Driving Conditions

Hybrid electric vehicles (HEVs) serve as a crucial intermediary step in the transition to fully electric passenger cars by enhancing the operation of internal combustion engines (ICEs) by optimizing operating strategies with the support of electric motors (EMs). Environment models, along with vehicle and driver models, further improve targeted optimization. Based on previous work, this paper employs a data-driven emission model to create an operating strategy that minimizes pollutant emissions through reinforcement learning (RL) and through a nonlinear model predictive controller (NMPC). The findings are then compared assessed, and potential enhancements are explored.