Iterative learning based driver input synthesis for evaluating transient diesel soot emissions

This paper presents the synthesis of driver inputs for evaluating diesel soot emissions using iterative learning control. Transient soot emissions from diesel engine vehicles are extremely sensitive to driver aggressiveness. Using closed-loop tracking controllers to follow a vehicle over a prescribed drive cycle usually do not account for the fact that drivers potentially adapt their driving styles to a given powertrain design. This work develops an algorithm producing driver input traces that significantly reduces the soot emissions for a given drive cycle, thus providing a consistent basis for evaluating the influence of powertrain design changes on soot emissions. Possible improvements are first explored using conventional optimal techniques and results are obtained using linear programming. It is then shown that a first-order PD-type iterative learning control based algorithm can deliver good performance, substantially reducing the total soot emissions at a fraction of the computational cost.