Fuel-efficient driving strategies

Doctoral thesis, 2020

This thesis is concerned with fuel-efficient driving strategies for vehicles driving on roads with varying topography, as well as estimation of road grade and vehicle mass for vehicles utilizing such strategies. A framework referred to as speed profile optimization (SPO), is introduced for reducing the fuel or energy consumption of single vehicles (equipped with either combustion or electric engines) and platoons of several vehicles. Using the SPO-based methods, average reductions of 11.5% in fuel consumption for single trucks, 7.5 to 12.6% energy savings in electric vehicles, and 15.8 to 17.4% average fuel consumption reductions for platoons of trucks were obtained. Moreover, SPO-based methods were shown to achieve higher savings compared to the commonly used methods for fuel-efficient driving. Furthermore, it was demonstrated that the simulations are sufficiently accurate to be transferred to real trucks. In the SPO-based methods, the optimized speed profiles were generated using a genetic algorithm for which it was demonstrated, in a discretized case, that it is able to produce speed profiles whose fuel consumption is within 2% of the theoretical optimum.

A feedforward neural network (FFNN) approach, with a simple feedback mechanism, is introduced and evaluated in simulations, for simultaneous estimation of the road grade and vehicle mass. The FFNN provided road grade estimates with root mean square (RMS) error of around 0.10 to 0.14 degrees, as well as vehicle mass estimates with an average RMS error of 1%, relative to the actual value. The estimates obtained with the FFNN outperform road grade and mass estimates obtained with other approaches.