Predictive Longitudinal Control of Heavy-Duty Vehicles Using a Novel Genetic Algorithm and Road Topography Data

Abstract

Fuel costs account for approximately one third of the total cost of haulage contractors. This makes it very lucrative from both the contractors’ and hence Scanias’ perspective to reduce the vehicles’ fuel consumption. With limited power-to-mass ratio of heavy-duty vehicles, anticipatory control is crucial for fuel- and time-efficient manoeuvring. Solutions addressing this problem are already in production, but with ever-increasing system complexity the usefulness of conventional mathematical methods is suffering. As an alternative approach, this thesis is aimed at investigating the applicability of a real-time genetic algorithm (GA) to the domain of longitudinal control of heavy-duty vehicles for fuel-saving adaption to road topography data. Known to be computationally heavy, an as lightweight as possible algorithm is developed and aimed at optimising the engine torque by model predictive control. The final algorithm uses a vehicle prediction model of fuel-consumption data including a gear prediction model. Validated through simulation this novel approach displays a clear improvement over a similar MPC-controller utilising a QP-solver and a cost function similar to that of the GA.