Adaptive Driving Agent

ABSTRACT Adaptive Cruise Control (ACC) is a technology that allows a vehicle to automatically adjust its speed to maintain a preset distance from the vehicle in front of it based on the driver's preferences. Individual drivers have different driving styles and preferences. Current systems do not distinguish among the users. We introduce a method to combine machine learning algorithms with demographic information and expert advice into existing automated assistive systems.

ArXiv, 2020

This paper issues an integrated control system of self-driving autonomous vehicles based on the personal driving preference to provide personalized comfortable driving experience to autonomous vehicle users. We propose an Occupant's Preference Metric (OPM) which is defining a preferred lateral and longitudinal acceleration region with maximum allowable jerk for users. Moreover, we propose a vehicle controller based on control parameters enabling integrated lateral and longitudinal control via preference-aware maneuvering of autonomous vehicles. The proposed system not only provides the criteria for the occupant's driving preference, but also provides a personalized autonomous self-driving style like a human driver instead of a Robocar. The simulation and experimental results demonstrated that the proposed system can maneuver the self-driving vehicle like a human driver by tracking the specified criterion of admissible acceleration and jerk.

2012

Abstract Traditionally, vehicles have been considered as machines that are controlled by humans for the purpose of transportation. A more modern view is to envision drivers and passengers as actively interacting with a complex automated system. Such interactive activity leads us to consider intelligent and advanced ways of interaction leading to cars that can adapt to their drivers.

Advances in Human Factors of Transportation

Automation in the road transport system is coming faster than expected being influencing and shaping the future of mobility. However, very few is known about the impact of automatic driving on traffic and how drivers will accept, use, trust and interact in traffic when driving a vehicle with a certain level of automation. Additionally, most of the potential users have unrealistic representations of autonomous vehicles, the driver's role in automation or the impacts of full automation on the road transport system. Aiming at better understanding the drivers' behavior when dealing with automated driving, this paper addresses the following issues based on a state of the art on automated driving: drivers' preferences for the automation levels across different categories of drivers; limits of the technology; needs for changes in traffic laws, as well as licensing and training; driver's promptness to resume the vehicle control following a long period of autonomous driving.

… Journal of Cognitive …, 2001

The research reported in this article considers the effects of automating driver's control tasks. Driving can be broken down into 3 general subtasks: navigation, control, and hazard avoidance. Control can be further subdivided into lateral control (position in lane) and longitudinal control (speed and leading headway). Lateral control can be automated by an active steering (AS) system, and longitudinal control can be automated by an adaptive cruise control (ACC) system. Previous research has used driving simulators to consider the effects of driver workload and the ability to reclaim control with these systems. There are, however, some questions about the validity of driving simulators, and this research sought to validate a driving simulator. This was achieved by comparing responses on a secondary task and driving style questionnaire in both a road car and a driving simulator. When validity was established, a comparison of 4 levels of automation was undertaken: manual, ACC, AS, and ACC plus AS. The results showed no reduction in workload associated with ACC over manual driving, but reduction in workload associated with AS and further reduction in workload associated with ACC plus AS. Despite these reductions in workload, there were no adverse affects on normal driving performance. However, the presence of vehicle automation seemed to make drivers less likely to reclaim control in an emergency-braking scenario.

2017

Fully automated driving of road vehicles on highways consists of different vehicle control tasks that are currently evolving from research to series production. These tasks are mainly covered by the three following driving systems: Adaptive Cruise Control -ACC, Lane Keeping Assist -LKA and Lane Change Assist -LCA. Among these systems, different levels of automation are available, e.g. warning/informing, intervene on demand and also partially automated systems. This research presents the first part of a feasibility study about an intuitive switching mechanism using a Human-Machine Interface (HMI) that combines aforementioned systems in order to adjust the level of automation of SAE levels 0 to 3, (Smith 2013), according to the preferences of a human driver. Driving systems of the levels 0, 2 and 3 together with a novel HMI concept are implemented in a driving simulator and tested with respect to acceptance and comfort as well as to reliability by a sample of 20 volunteer drivers. The proof of concept of each system as well as the testing form the basis for a later implementation of the presented intuitive switching mechanism.

Driver's training is essential in order to assess and provide the driver with sufficient skills to handle vehicles in complex and dynamic environment. These skills are related to the cognitive factors that will influence the automaticity of the driver to make effective decision. To illustrate these skills, simulation scenarios based on driver's training has been conducted. It has pointed out that the simulation results are related to the existing concepts that can be found in literatures. 1 Introduction Driving is an activity that requires human cognition where the driver continuously encounters tasks in need of solutions. The ability to master these tasks will depend on the knowledge, the skills the person has, and the person's intellectual abilities. Furthermore, it aims to remove the barrier between knowledge and the skills required to drive safely and efficiently when commencing training where these skills must be known, for the training to be appropriate [1]. Moreover, the objective of training for critical decision making is to provide the learner with experiences, and instruction on cues, patterns, mental models, and actions that could efficiently establish a collection of well-learned concepts which enable the operator to perform mainly at the skill-based level of processing, while providing adequate knowledge-based foundation to perform well in new situations [2]. Thus, training is needed in recognizing situations, in communicating situation assessment, and in acquiring the experience to conduct mental simulation of options through the act of human cognitive unconscious decision making, or automaticity [3], [4]. Having analyzed this ability, it will provide a good perspective towards driving assisting systems. To address this issue, this paper proposes a computational agent model of automaticity, which integrates related dynamic factors based on cognitive and psychology theories to describe basic training process that enhances the automaticity required by a driver to make fast decision in a driving domain. The organization of the remaining part of this paper is as follows. The underlying concepts and a computational agent model of automaticity are discussed in Section 2. The study scenario and

Transportation Research Part F: Traffic Psychology and Behaviour, 2019

To evaluate human-machine interfaces for automated driving systems, a robust methodology is indispensable. The present driving simulator study investigated the effect of practice on behavioral measures (i.e., experimenter rating, reaction times, error rate) and the development of the preference-performance relationship for automated driving humanmachine interfaces. In a within-subject design, N = 55 participants completed several transitions between manual, Level 2 and Level 3 automated driving. Behavioral measures followed the power law of practice with exception of transitions to manual and error rates for Level 3 automation. After the first block of interactions, preference no longer predicted performance. The preference-performance relationship remained stable after the second block of interactions, which is mainly due to a stabilization in behavioral parameters. To get a deeper insight into the evaluation of human-machine interfaces for automated driving, the results suggest the application of multi-method approaches. Furthermore, we found evidence for the influence of initial interactions for self-reported usability.

2021

This paper presents a research process aiming at studying the activity of drivers in automated vehicles (AV). To do this, we carried out in situ observations and conducted interviews with two different populations: professional and novice drivers in automated driving (AD). The results obtained by triangulation highlight a series of “sensitive” situations specific to automated driving. The clinical analysis of these situations shows changes in the relations and mediations involved. Some of them have common characteristics, making it possible to classify the sensitive situations identified. These changes require a potential adaptation of “traditional” driving schemes, necessary for the appropriation of the AV by the drivers. These results allow us to provide recommendations for improving AV prototypes, and to consider the design of a learning tool to support the appropriation of these systems. This device should, at a minimum, make it possible to familiarize vehicle drivers with sensi...

Abstract. Adaptive Cruise Control (ACC) is a technology that allows a vehicle to automatically adjust its speed to maintain a preset distance from the vehicle in front of it based on the driver's preferences. Individual drivers have different driving styles and preferences. Current systems do not distinguish among the users. We introduce a method to combine machine learning algorithms with demographic information and expert advice into existing automated assistive systems.