Driver Modeling and Implementation of a Fuel-Saving ADAS

Abstract

Controlling vehicle velocity, by coaching the driver to eco-drive with an advanced driver assistance system (ADAS), is a promising method to decrease fuel consumption and greenhouse gas emissions for combustion engine-driven road vehicles. By using optimal control techniques, such a system may find velocity profiles in real-time that minimize fuel consumption. This is particularly useful to recommend the optimal time to initiate coasting, which is otherwise difficult to estimate by a driver. However, this ADAS should not choose velocities and accelerations that the driver will dislike, such as those that leave too much or too little space to the preceding vehicle, or those that take corners at high speed. To remedy this, we introduce an optimal control model of acceleration that mimics drivers' behavior and combine this with a model of fuel consumption to trade-off driver preferences and fuel savings. We give examples of the velocity profiles recommended in a typical driving scenario to demonstrate the potential fuel savings. Finally, we give details of a prototype system, which has recently been implemented in the driving simulator at the University of Southampton.

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

“…However, choosing to minimise fuel consumption alone can lead to behaviour that is unnatural for a human driver, such as travelling far below the speed limit or leaving large spacings to the preceding vehicle. To address this, the authors of the present paper considered a modified optimal control problem in [28] in which the cost function has the form…”

Design and Implementation of a CANBus-Based Eco-Driving System for Public Transport Bus Services

“…Four drivers with different driving experiences were asked to drive with and without the eco-driving support system and it was seen that the eco-driving support system changes the behavior of the drivers positively. In [39], a model was created by using variables such as acceleration, speed, distance (to other vehicles) and cornering speed and hence an eco-driving support system was proposed according to this model. Zhai et al [40] investigated how the fuel efficiency of vehicle platoons can be increased and proposed a control strategy for vehicle platoon travelling on roads with varying slopes.…”

“…In fact, many of those studies (e.g. [14], [16], [17], [25]- [27], [36], [39]) do not aim at comparing eco-driving performances of different drivers. Further, neither reporting the comparison between performances nor the usage of the previous trips' data retrieved over CANBus for assisting the drivers are considered in those studies.…”

Fitting Cornering Speed Models with One-Class Support Vector Machines

“…so that a lower bound on TLC implies an upper bound to the lateral acceleration (1). This led the authors to postulate a margin of error of the driver when estimating the curvature of an upcoming corner, implying a decreasing quadratic relationship between the upper bound of lateral acceleration and vehicle velocity.…”

“…It is of considerable interest to be able to predict the speed at which a driver will negotiate a curve based on values which may be readily estimated from mapping services, such as road curvature, as this has several useful applications to Advanced Driver Assistance Systems (ADAS). The authors' main motivation in investigating this problem is to build predictive models of cornering speed for eco-driving assistance systems, which are designed to save fuel and emissions by coaching the driver to coast down and to avoid braking before corners and intersections [1]. Another potential application is in curve warning systems, which rely on estimates of the speed at which a driver will take a corner in order to identify when a warning is required [2].…”