a comparison of driving characteristics and environmental characteristics
DESCRIPTION
A Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithms. : 지능형 첨단차량을 위한 친환경 주행 모형의 개발. 2012. 10. 26. 정희진. Virginia Tech. Table of Contents. Introduction Study 1 Study 2 Study 3 Conclusion and Further Study. - PowerPoint PPT PresentationTRANSCRIPT
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
A Comparison of Driving Characteristics and Environmental Characteristicsusing Factor Analysis and K-means Clustering Algorithms: 지능형 첨단차량을 위한 친환경 주행 모형의 개발
Virginia Tech
정희진
2012. 10. 26
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm2
Table of Contents
1. Introduction2. Study 13. Study 24. Study 35. Conclusion and Further Study
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm3
Introduction: 연구 배경
첨단 기술의 발달로 과거와 달리 운전자의 행동에 직접적으로 영향을 줄 수 있는 잠재적 가능성이 높아짐 .
Driving Assistance Systems충돌 예방 경보 시스템
지능형 첨단 차량Intelligent Vehicle
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm4
첨단 교통 시스템 분야에서 운전자의 행동과 자동차 제어 전략의 변화의 환경적 영향에 관심이 높음
Eco-driving assistance Systems경제적 환경적 주행 보조 시스템
Eco-driving 의 개념은 운전자의 주행 거동의 변화를 통해 연료소모를 최소하는 것임
Introduction: 연구 배경
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm5
Eco-driving 기대 편익
소음 감소
대기 오염물질의 감축
온실가스 감축
운전기술의 강화
도로 안전의 강화
Environ-mental
Safety
운전자 및 탑승객 승차감 향상
주행 중 스트레스 감소
더욱 책임감 있는 주행
차량 유지 보수 비용 저감
사고 비용 감소
연료 소모량 감소 SocialFinan-cial
Introduction: 연구 배경
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm6
연료 소모량 배기가스배출량
엔진 온도차량 종류
구배
차량 무게
차량 연식유지 관리
노면 상태
주행 습관
Introduction: 연구 배경
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
본 연구의 목표는 eco-driving 을 모형화를 위해 개별 운전자의 주행 특성과 환경 특성을 분석하여 이상적인 주행 형태를 찾는 것임
Introduction: 연구 배경
주요 연구 내용
이상적 주행행태 도출을 통한 Eco-driving 구현
동일 교통류 상의 개별 차량 간 배기가스 배출량과 연료소모량의 차이 분석Task 1
개별 운전자의 가감속도 차이에 따른 배기가스 배출량과 연료 소모량을 비교Task 2
운전자의 유형 차이에 따른 주행 및 환경 영향 특성을 비교Task 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm8
Introduction: 전체 연구 흐름도
Trajectory Data 분석 및 처리
Microscopic emissions model 을 이용한 연료소모량 및 배기가스 배출량 산정
Study 1: 배기가스 배출량 및 연료 소모량 비교
Study 2: aggressivity 기반의 운전자 분류와 운전자 그룹의 배기가스 및 연료소모량 비교
Study 3: 주행 특성 기반 분류와 환경 영향 기반 분류의 비교 및 상관성 연구
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Used NGSIM trajectory data sponsored by FHWA (Next Generation Simulation Program.
The data was collected every deci-second for 15 minutes, and then con-verted to trajectory data including 18 items.
The site where data was collected is a segment of I-80 including 6 main stream lanes of 1650 ft and 1 on-ramp of 140 ft.
The data was collected three times on April 13, 2005:
Periods Assumed traffic condition
# of observed Cars (motorcycle/auto/ trucks & buses)
Average speed (TMS/SMS)
4:00-4:15 PM Non-congested 2052 (14/1942/96) 22.19 /17.86 MPH
5:00-5:15 PM First congested 1836 (24/1742/70) 18.72 / 14.04 MPH
5:15-5:30 PM Second congested 1790 (17/1724/49) 17.40 / 12.40 MPH
NGSIM trajectory data
Introduction: Trajectory 데이터
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
NGSIM trajectory data
Introduction: Trajectory 데이터
7 digitalcameras
segment of I-80
각 차량의 궤적 데이터 구성
속도와 가속도 선행 및 후행차량인식
Time/distance headway 와 상대속도
18 개 항목으로 구성된 trajectory data 구축
영상 처리 알고리즘을 이용한 분석
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
VT-Micro for first and second studies
Microscopic Energy and Emissions Models
Comprehensive Modal Emissions Model(CMEM) for third study
e auLauLauLauLauLauLuaLuaLuaLuLuLuLaLaLaLLMOE33
1523
143
1332
1222
112
103
92
873
62
543
32
210
e auMauMauMauMauMauMuaMuaMuaMuMuMuMaMaMaMMMOE33
1523
143
1332
1222
112
103
92
873
62
543
32
210
For accelerating (equation1):
For braking (equation 2):
Introduction: 연료소모량 및 배기가스 배출량 산정
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
연구 흐름도
Study 1
Smoothing 속도 프로파일 데이터
VT-Micro Model 을 이용한 연료소모량 및 배기가스 배출량 산정
Percentile 분석
배기 가스 배출량 및 연료 소모량 비교
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
연료 소모량 및 배기가스 배출량 예측 VT-Micro model 사용
정체 정도에 따라 3 개의 데이터 셋 사용 . 예측은 다음과 같은 전제에서 수행됨
– 모든 차량은 같은 연식 같은 타입의 승용차로 구성– 각 차량의 차이는 속도 프로파일에 의해서만 정의됨
Study 1
Percentile 분석 예측된 연료 소모량 및 배기가스 배출량을 기준으로 순서대로 재 배열
순서대로 100 개의 percentile 로 균등 분배
각 percentile 내의 차량 중 가장 많은 연료를 소모했거나 가장 많은 특정 배기가스를 배출한 차량을 대표차량으로 선정
각 percentile 의 대표차량의 연료소모량 과 배기가스 배출량을 비교
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
검증 1: 비정체 상황에서
Percent of fuel consumption and emissions for different percentile of vehicles in non-congestion condition.
High emitters
Study 1
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
검증 2: 정체 상황에서
Study 1
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
결론
비정체 , 정체 모든 상황에서 약 15% 의 차량 percentile 에 해당하는 차량들이 다른 차량에 비해 최대 300% 의 많은 배기가스를 배출하였다 .
연료소량은 배기가스 배출량에 비해 차이는 적었으나 약 5% 의 차량 percentile 이 약 두배의 연료를 소모하였다 .
그러므로 , 동일 교통류에서 배기가스 배출량과 연료소모량이 많은 주행 특성을 가지는 차량이 존재한다 .
Study 1
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
연구 흐름도
Study 2
가속도 프로파일 데이터 처리
Phase 분석을 통한 target operating acceleration modeling
Target operating acceleration 을 이용한 운전자 분류
운전자 그룹별 배기가스 및 연료 소모량 비교
운전자 주행 특성 변화에 의한 배기가스 배출량 및 연료소모량 개선 효과 분석
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Concept of the Five Pro-cesses
0 o oo
o
o o
o
o o o o
o
o
o o o
o
o
o
o
o
Constant Speed(Zero Acceleration)
Accelerating
Recovery A
Braking Recovery B
o Acceleration DataProcess
Target operating Acceleration
Target operating Acceleration
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Separating Processes Al-gorithm
Where, Acc is acceleration, and dAcc/dt is variation of acceleration
ProcessCondition 1
(value of acceleration)Condition 2
(variation of acceleration)
targetOperating
AccelerationAccelerating Positive Not Negative Highest valueBraking Negative Not positive Lowest valueRecovery A Positive Negative ZeroRecovery B Negative positive ZeroConstant Speed Zero - Zero
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm20
Phasing 된 가속도 프로파일 데이터와 target operating acceleration 의 예
Time versus acceleration diagram for an example recognized process and its target operating acceleration.
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
분류 방법
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
conditionsThe first time period The second time period The third time period
mean S.D. Mean S.D. Mean S.D.
Acceleration 6.28 0.91 5.93 0.84 6.02 0.74
Brake -6.31 0.88 -5.92 0.78 -5.93 0.74
Mean and standard deviation of operating acceleration and brakeAverage Target Operating Acceleration and Brake
Study 2
avgiiaccel AAD , avgiibrake BBD ,
Where, Daccel,i and Dbrake, i are the differences in average operating accel-eration and brake of ith vehicle from the mean of average operating accelera-tion of all vehicles respectively. Ai and Bi are the average operating accelera-tion and brake of ith vehicle respectively, and Aavg and Bavg are the mean of average operating acceleration and brake of all vehicles under consideration.
Vari-ables
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Results of Classifi-cation
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Traffic condition
The first time period
The second time period
The third time period
Total # of vehicles
Defensive 301 289 256 846Moderate 1448 1275 1291 4014
Aggressive 303 272 243 818
Total # of vehicles 2025 1836 1790 5678
Results of Classifi-cation
Study 2
Number of vehicles in each class
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Results of fuel consumption and emissions in each condition.Note: 1: number of vehicles, 2: Fuel Consumption 3: HC, 4: CO, and 5: Nox
the numbers on bar: rates in percentage
13.580
72.12
14.30
11.47
73.39
15.14
16.01
69.91
14.06
18.63
68.26
13.12
17.69
69.24
13.07
14.770
70.57
14.67
13.81
71.99
14.20
21.52
66.04
12.44
23.40
64.98
11.63
20.40
67.39
12.22
14.810
69.44
15.74
13.20
69.60
17.20
19.08
65.59
15.33
21.60
64.28
14.12
19.79
66.14
14.07
Evaluation of emissions and fuel con-sumption
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Considered two alternative scenarios.
Alternative 1 All aggressive drivers changed their driving behaviors to moderate drivers
Alternative 2 All aggressive and moderate drivers changed their driving behaviors to de-
fensive drivers
Estimation of impact of driving behavior changes
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Time period emissionsBase Alternative 1 Alternative 2
Total Total Changed % Total Changed %
The first time period
Fuel 136082ml 137797.1ml 1715.1ml 1.3% 131736.2 -4345.8ml -3.2%
HC 25866mg 23875.4mg -1990.6mg -7.7% 21953.7gm -3912.3mg -15.1%
CO 686980mg 619584mg -67396mg -9.8% 543909gm -143071mg -20.8%
NOx 56131mg 52589.3mg -3541.7mg -6.3% 46779.8mg -9351.2mg -16.7%
The secondTime period
Fuel 131991ml 134149ml 2158ml 1.6% 144298.2ml 12307.2ml 9.3%
HC 23639mg 22447.4mg -1191.6mg -5.0% 22961.5mg -677.5mg -2.9%
CO 601293mg 553803mg -47490mg -7.9% 539230.9mg -62062.1mg -10.3%
NOx 50028mg 47183.4mg -2844.6mg -5.7% 44629.1mg -5398.9mg -10.8%
The third time period
Fuel 141915ml 145243.1ml 3328.1ml 2.3% 150181.6ml 8266.6ml 5.8%
HC 23654mg 22972.4mg -681.6mg -2.9% 23267mg -387mg -1.6%
CO 587667mg 553774mg -33893mg -5.8% 538807mg -48860mg -8.3%
NOx 48731mg 46451.3mg -2279.7mg -4.7% 44563.6mg -4167.4mg -8.6%
Note: Alternative 1: Aggressive drivers change to moderate drivers, and Alternative 2: Aggres-sive and Moderate drivers change to defensive drivers.
Results of estima-tions
Study 2
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
alternative 1 에 비해 alternative 2 에서 개선 효과가 더 기대됨
비 정체상황에서 Alternative 1 의 경우 CO, HC, and NOx 각각 9.8%, 7.7%, and 6.3 % 감소하였음 .
비 정체상황에서 Alternative 2 의 경우 CO, HC, and NOx 각각 20.8%, 15.1%, and 16.7% 감소하였음 .
정체 상황에서 연료소모량과 분류된 운전자 그룹간의 상관관계는 발견되지 않았음 . 모든 교통 상황에서 배기가스 배출량과 운전자 그룹간의 상관관계가 발견되었음 .
Results of estima-tions
Study 2
차량의 속도특성이 연료 소모량의 주요인으로 추측됨 . 제안된 average target operating acceleration 은 배기가스 배출량의 차이의
주요인으로 간주될 수 있음 . 비정체 상황 보다는 정체상황에서 배기가스 배출량의 감소효과가 더 기대됨 . 운전자 교육 및 홍보를 통해 aggressive 운전자의 주행 습관을 변화시킨다면
배기가스의 유의미한 감소를 기대할 수 있음 .
Summary of Find-ings
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
연구 흐름도
Study 3
Driving data 구성
Factor Analysis
Cluster 수 결정
Driving Clustering
Mapping
Environmental data 구성
Factor Analysis
Cluster 수 결정
Environmental Clustering
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
variables min max median mean std skewness kurtosis N
avg. speed in ft/s 13.16 78.56 24.26 29.16 13.63 1.73 5.24 1940
std. speed in ft/s 2.12 23.88 7.81 8.03 2.54 0.76 5.13 1940
avg. target operating acceleration in ft/s^2 3.87 9.83 6.68 6.61 1.22 0.01 1.87 1940
std. target operating acceleration in ft/s^2 2.95 5.00 4.04 4.04 0.26 0.01 3.71 1940
avg. spacing in ft 26.51 1256.14 62.52 75.58 54.64 8.75 144.21 1940
std. spacing in ft 3.68 284.43 18.42 22.53 17.07 5.13 60.54 1940
avg. time headway in second 0.88 2077.58 3.39 105.47 262.41 3.86 19.94 1940
std. time headway in second 0.05 4055.40 1.35 507.12 828.09 1.89 6.13 1940
avg. relative speed in ft/s -17.82 24.72 0.08 0.12 2.02 1.45 34.22 1940
std. relative speed in ft/s 2.19 16.95 5.34 5.62 1.71 1.77 8.83 1940
5 Selected variables were converted to 10 variables from the trajectory data
Driving Data
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Results of Factor Analysis of Driving Data
Factor analysis of Driving Data
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
microscopic Factor
Mac
rosc
opic
fact
or
공간 기반 운전자 그룹(Spacing-based drivers group)
속도 기반 운전자 그룹(Speed-based drivers group)
Moderate group
Results of driving clus-tering
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Definition of clus-ters
Study 3
상대적으로 큰 Microscopic factor 값을 가진 운전자 그룹 .
cluster 2 의 속도 기반 운전자 그룹
상대적으로 큰 Macroscopic factor 값을 가진 운전자 그룹 .
cluster 1 의 공간 기반 운전자 그룹
두 factor 값 모두 작은 값을 가진 운전자 그룹
cluster 3 의 Moderate 운전자 그룹
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
CMEM 을 이용하여 예측한 배기가스 배출량과 연료소모량
gram/mile min max median mean std skewness kurtosis N
Hydrocarbons (HC) 0.01 1.50 0.04 0.15 0.21 2.22 8.73 1940
Carbon Monoxide (CO) 0.19 179.21 1.85 16.18 24.59 2.18 8.66 1940
Oxides of Nitrogen (NOX) 0.15 1.22 0.32 0.36 0.13 1.83 7.86 1940
Carbon Dioxide (CO2) 231.18 787.79 487.42 494.79 89.22 0.35 3.08 1940
Fuel consumption 78.60 311.50 161.20 164.14 30.47 0.58 3.81 1940
Descriptive statistics of variables as emissions and fuel consumption
Environmental Data
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Factor analysis of Environmen-tal Data
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Incomplete Combustion Factor
Fuel
Con
sum
ptio
n Fa
ctor
Moderate emitters group
High emitters group
Results of Environmental clustering
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Clustering clusters
Number of
VehiclesHC CO NOx CO2 Fuel
N % grams % grams % grams % grams % grams %
drivingclustering
Spacing-based 195 10.05 29.89 34.12 3394.24 35.75 32.17 15.39 25301.91 8.84 9689.15 10.20
Speed-based 294 15.15 8.21 9.37 850.38 8.96 32.99 15.79 52792.08 18.45 17071.44 17.97
moderate 1451 74.79 49.49 56.50 5249.33 55.29 143.81 68.82 208096.07 72.71 68250.52 71.83
Environmen-tal
clustering
high 538 27.73 56.96 65.03 6508.16 68.55 83.22 39.82 90808.00 31.73 31909.81 33.59
moderate 1402 72.27 30.62 34.97 2985.80 31.45 125.75 60.18 195382.07 68.27 63101.30 66.41
Entire vehicles 1940 100 87.58 100 9493.96 100 208.97 100 286190.07 100 95011.11 100
Comparative Analysis of Emissions and Energy
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Map-ping
Comparative analysis between (a) driving clustering and (b) environmental clustering on the incomplete combustion factor versus fuel consumption space.
Moderate driversSpeed based driversspacing based drivers
Spacing-based Drivers group
Speed-based Drivers group
Moderate drivers group
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
앞에서 제시한 전제에 근거해 약 75% 운전자들은 두 분류 방법에서 일치 되게 분류됨 .
공간 기반 운전자들은 적은 양의 배기가스를 배출하였으나 상대적으로 많은 양의 연료를 소모하는 High energy consumers’ group 으로 분류 될 수 있음 .
반면 , 속도 기반 운전자는 같은 연료 소모 성향을 가진 운전자들 중에 상대적으로 많은 배기가스를 배출하는 high emitters’ group 으로 분류할 수 있음 .
대부분의 moderate 운전자들은 그렇지 않은 운전자들에 비해 적은 배기가스 및 연료를 소모하는 성량이 있음 .
Summary of Find-ings
Study 3
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
Significant factors for driving clustering and environmental clustering was found.
While the high error rate (25%), the relationship between driving clustering and environmental clustering is significant.
A potential to estimate emissions and fuel consumption based on driving clustering is found.
The moderate drivers’ group should be similar to the eco-drivers.
Changing driving behavior to moderate drivers is recommended.
Conclusion
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
본 연구의 study 3 에서 돌출된 결론인 정체상황 하에서도 유효한지에 대한 추가 연구가 필요함 .
비선형 classification algorithm 을 사용하여 보다 정교하고 확고한 분류를 수행할 필요가 있음 .
운전자의 주행 특성을 보다 잘 반영할 수 있는 추가적인 변수의 개발이 필요함 .
Eco-driving 의 모형화 및 평가 모델의 개발이 필요함 .
운전자의 주행 환경을 반영한 eco-driving 모형의 개발이 필요함 .
Further Study
Intelligent Transportation SystemA Comparison of Driving Characteristics and Environmental Characteristics using Factor Analysis and K-means Clustering Algorithm
감사합니다 .