Incorporating Driver Preferences Into Eco-Driving Assistance Systems Using Optimal Control

Abstract

Recently there have been several proposals for 'ecodriving assistance systems', designed to save fuel or electrical power by encouraging behaviours such as gentle acceleration and coasting to a stop. These systems use optimal control to find driving behaviour that minimises vehicle energy losses. In this paper, we introduce a methodology to account for driver preferences on acceleration, braking, following distances and cornering speed in such eco-driving optimal control problems. This consists of an optimal control model of acceleration and braking behaviour containing several physically-meaningful parameters to describe driver preferences. If used in combination with a model of fuel or energy consumption, this can provide an adjustable trade-off between satisfying those preferences and minimising energy losses. We demonstrate that the model gives comparable performance to existing car-following and cornering models when predicting drivers' speed in these situations by comparison with real-world driving data. Finally, we present an example highway braking scenario for an electric vehicle, illustrating a trade-off between satisfying driver preferences on vehicle speed and acceleration and reducing electrical energy usage by up to 43%.

An Overview of Eco-Driving Theory, Capability Evaluation, and Training Applications

“…The difficulty for drivers to drive at a given speed trajectory due to the large external interference seems to be the main reason for the small improvement of fuel consumption in urban driving. Meanwhile, when making eco-driving recommendations, the driver’s driving preferences (such as longitudinal and lateral acceleration, vehicle spacing) should be considered to improve the driver’s acceptance of eco-driving recommendations [ 59 , 60 ]. Eco-driving operation recommendations should be the result of trading off driver preference and fuel economy.…”

“…Drivers’ driving behavior has a certain tendency, and it is very difficult to change driving habits. Even if it changes in a short amount of time due to external factors, when those external factors are eliminated, the effect will decline over time, and they will even return to their original driving habits [ 59 ]. Eco-driving training did indeed achieved good energy-saving effects in the short term, but the long-term effect is not good.…”

Attention Horizon as a Predictor for the Fuel Consumption Rate of Drivers

“…For this reason, in the past decade there has been a surge of studies that try to determine the driving behavior that maximizes fuel efficiency [ 8 , 9 , 10 , 11 , 12 , 13 ]. The term “eco-driving” [ 8 , 13 , 14 , 15 , 16 , 17 , 18 , 19 ] is often used to describe this type of efficient driving, although it also includes strategic and tactical decisions that lead to less fuel consumption, besides driving behavior itself. In the past, relatively simple data from controlled vehicle testing was often used to determine the optimal behavior [ 1 , 20 ].…”

“…In particular, combining high-resolution sensor data with additional heterogeneous sources, such as map information, provides unique novel opportunities. The obtained driver behavior model can either be used as an algorithm for a cruise control system or for suggestions in a driver assistant system [ 17 , 18 , 22 , 23 ], indicating what actions to take to improve fuel efficiency. For example, there is a study that gives gear shift guidance to the driver according to the timing of the traffic lights [ 9 ].…”

Adjusting the need for speed: assessment of a visual interface to reduce fuel use

“…This paper seeks to extend previous research McIlroy, Stanton & Godwin, 2017) by presenting an evaluation of a novel assistance system that provides feedforward information based on a model of vehicle fuel consumption, prediction of the position of the preceding vehicle given its present velocity, and upcoming road geometry (Fleming, Yan, Allison, Stanton & Lot, 2018). Uniquely, this system incorporates a model of typical driver preferences on aspects such as following distances and cornering speeds to 'coach' the driver into more fuelefficient behaviour as a modification of their natural driving style (Fleming, Yan, & Lot, 2020). Using a repeated-measures study in a fixed-base driving simulator, its efficacy is compared to traditional eco-driving advice in terms of fuel economy, effects on driver speed, acceleration and braking, and measures of cognitive workload.…”