Accuracy of a Driver Model with Nonlinear AutoregRessive with eXogeous Inputs (NARX)

Akihiro Miyata, Masato Gokan, Toshiya Hirose

Research output: Contribution to journalConference article

Abstract

Most driving assist systems are uniformly controlled without considering differences in characteristics of individual drivers. Drivers may feel discomfort, nuisance, and stress if the system functions differently from their characteristics. The present study reduced these side effects for systems with a highly accurate driver model. The model was constructed using Nonlinear AutoregRessive with eXogeous inputs (NARX), which has a learning function and estimates the driving action of a driver. The model was constructed for one driving condition yet can be applied to other driving conditions. If one model can be applied to many driving conditions, a system can construct as minimum requirements. The driver decelerated while approaching the target at the tail of a traffic jam on a highway. A driver model was constructed for the driver's braking action. The experimental condition was 11 data measurements from 50 to 130 km/h made at intervals of 10 km/h. A model was constructed with 1-10 data. Analysis clarifies the number of data points needed to construct the model. The accuracy of the model was confirmed from 50 to 130 km/h at intervals of 10 km/h. Analysis clarifies on model accuracy when there is velocity difference. The accuracy of the model improved as the volume of learning data increased. The relationship between the volume of data and the model accuracy was clarified. The accuracy decreased as the difference in velocity increased, and this tendency was more obvious when the subject vehicle travelled at low velocity.

Original languageEnglish
JournalSAE Technical Papers
Volume2018-April
DOIs
Publication statusPublished - 2018 Jan 1
Event2018 SAE World Congress Experience, WCX 2018 - Detroit, United States
Duration: 2018 Apr 102018 Apr 12

Fingerprint

Braking

ASJC Scopus subject areas

  • Automotive Engineering
  • Safety, Risk, Reliability and Quality
  • Pollution
  • Industrial and Manufacturing Engineering

Cite this

Accuracy of a Driver Model with Nonlinear AutoregRessive with eXogeous Inputs (NARX). / Miyata, Akihiro; Gokan, Masato; Hirose, Toshiya.

In: SAE Technical Papers, Vol. 2018-April, 01.01.2018.

Research output: Contribution to journalConference article

@article{4d60060867024daeacfd48a3ae8edb7e,
title = "Accuracy of a Driver Model with Nonlinear AutoregRessive with eXogeous Inputs (NARX)",
abstract = "Most driving assist systems are uniformly controlled without considering differences in characteristics of individual drivers. Drivers may feel discomfort, nuisance, and stress if the system functions differently from their characteristics. The present study reduced these side effects for systems with a highly accurate driver model. The model was constructed using Nonlinear AutoregRessive with eXogeous inputs (NARX), which has a learning function and estimates the driving action of a driver. The model was constructed for one driving condition yet can be applied to other driving conditions. If one model can be applied to many driving conditions, a system can construct as minimum requirements. The driver decelerated while approaching the target at the tail of a traffic jam on a highway. A driver model was constructed for the driver's braking action. The experimental condition was 11 data measurements from 50 to 130 km/h made at intervals of 10 km/h. A model was constructed with 1-10 data. Analysis clarifies the number of data points needed to construct the model. The accuracy of the model was confirmed from 50 to 130 km/h at intervals of 10 km/h. Analysis clarifies on model accuracy when there is velocity difference. The accuracy of the model improved as the volume of learning data increased. The relationship between the volume of data and the model accuracy was clarified. The accuracy decreased as the difference in velocity increased, and this tendency was more obvious when the subject vehicle travelled at low velocity.",
author = "Akihiro Miyata and Masato Gokan and Toshiya Hirose",
year = "2018",
month = "1",
day = "1",
doi = "10.4271/2018-01-0504",
language = "English",
volume = "2018-April",
journal = "SAE Technical Papers",
issn = "0148-7191",
publisher = "SAE International",

}

TY - JOUR

T1 - Accuracy of a Driver Model with Nonlinear AutoregRessive with eXogeous Inputs (NARX)

AU - Miyata, Akihiro

AU - Gokan, Masato

AU - Hirose, Toshiya

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Most driving assist systems are uniformly controlled without considering differences in characteristics of individual drivers. Drivers may feel discomfort, nuisance, and stress if the system functions differently from their characteristics. The present study reduced these side effects for systems with a highly accurate driver model. The model was constructed using Nonlinear AutoregRessive with eXogeous inputs (NARX), which has a learning function and estimates the driving action of a driver. The model was constructed for one driving condition yet can be applied to other driving conditions. If one model can be applied to many driving conditions, a system can construct as minimum requirements. The driver decelerated while approaching the target at the tail of a traffic jam on a highway. A driver model was constructed for the driver's braking action. The experimental condition was 11 data measurements from 50 to 130 km/h made at intervals of 10 km/h. A model was constructed with 1-10 data. Analysis clarifies the number of data points needed to construct the model. The accuracy of the model was confirmed from 50 to 130 km/h at intervals of 10 km/h. Analysis clarifies on model accuracy when there is velocity difference. The accuracy of the model improved as the volume of learning data increased. The relationship between the volume of data and the model accuracy was clarified. The accuracy decreased as the difference in velocity increased, and this tendency was more obvious when the subject vehicle travelled at low velocity.

AB - Most driving assist systems are uniformly controlled without considering differences in characteristics of individual drivers. Drivers may feel discomfort, nuisance, and stress if the system functions differently from their characteristics. The present study reduced these side effects for systems with a highly accurate driver model. The model was constructed using Nonlinear AutoregRessive with eXogeous inputs (NARX), which has a learning function and estimates the driving action of a driver. The model was constructed for one driving condition yet can be applied to other driving conditions. If one model can be applied to many driving conditions, a system can construct as minimum requirements. The driver decelerated while approaching the target at the tail of a traffic jam on a highway. A driver model was constructed for the driver's braking action. The experimental condition was 11 data measurements from 50 to 130 km/h made at intervals of 10 km/h. A model was constructed with 1-10 data. Analysis clarifies the number of data points needed to construct the model. The accuracy of the model was confirmed from 50 to 130 km/h at intervals of 10 km/h. Analysis clarifies on model accuracy when there is velocity difference. The accuracy of the model improved as the volume of learning data increased. The relationship between the volume of data and the model accuracy was clarified. The accuracy decreased as the difference in velocity increased, and this tendency was more obvious when the subject vehicle travelled at low velocity.

UR - http://www.scopus.com/inward/record.url?scp=85045433718&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85045433718&partnerID=8YFLogxK

U2 - 10.4271/2018-01-0504

DO - 10.4271/2018-01-0504

M3 - Conference article

AN - SCOPUS:85045433718

VL - 2018-April

JO - SAE Technical Papers

JF - SAE Technical Papers

SN - 0148-7191

ER -