Professional Documents
Culture Documents
Chenchao 2016
Chenchao 2016
Abstract˖Generally, vehicle automatic emergency braking system (AEBS) evaluation is based on traffic accident data
statistics, to calculate the weight of each test condition or directly evaluate it’s qualified or not. Due to each test condition of
choice is typical speed for the accident conditions, thus has certain one-sidedness in the evaluation. Aiming at this problem,
this paper puts forward the theory of analytic hierarchy process (APH) for vehicle AEBS evaluation. Get the following results:
Hierarchical structure model of AEBS evaluation was set up, based which, get the judgment matrix of weight calculation;
Weights of working condition of each test was calculated; According to the theory of hierarchical analysis judgment, calculate
the consistency of judgment matrix is good; Finally, consistency evaluation of same speed and different speed was added in the
AEBS evaluation.
Key words˖
˖AEBS evaluation˗AHP˗hierarchical structure model˗judgment matrix˗consistency judgement˗plus evaluation
7HVWVSHHGNPK
2.2 The construction of judgment matrix and the Target vehicle
static 7HVWVSHHGNPK
7HVWVSHHGNPK
Analytic hierarchy structure model is completed, we
need to build more out of the judgment matrix of each
evaluation index. In the hierarchical structure model for a Fig 2.1: Hierarchic APH model
layer element, the relative importance between the
hierarchical element numerical comparison, and in the form In analytic hierarchy process (AHP), the importance of
of matrix, said the judgment matrix. Judgement matrix the matrix elements of the element and element j I more
structure need to be on the two comparison between the value, and both are reciprocal relationship, namely
various indicators, and then press 9 points a ratio scheduled 1
order the relative weight of each evaluation index, construct bij ˄4.3˅
the evaluation index in turn. Assume that figure 2.1 establish
b ji
the hierarchical structure of the model from top to bottom 1 1 1 1
three layers are expressed with A, B, C, namely target layer, Under which take 9, 7, 5, 3, 1, , , , a total of 9
criterion layer B, plan C.The objective judgment matrix for 3 5 7 9
A relative criterion layer B: scale, in turn, said the weight from heavy to light. Table 2.1
for the meaning of the value scale [20] [21]:
§1 b12 b13 b14 b15 ·
¨ ¸
¨ b21 1 b23 b24 b25 ¸ Calibration, bij The meaning of calibration
˄4.1˅
A= ¨ b31 b32 1 b34 b35 ¸ 1 bi and bj have the same importance
¨ ¸ 3 bi is a bit important than bj
¨ b41 b42 b43 1 b45 ¸ 5 bi is Quite important than bj
¨b b52 b53 b54 1 ¸¹ 7 bi is very important than bj
© 51
Similarly, change lanes conditions, for example, 9 bi is extremely important than bj
standard layer relative to the plan B to C of judgment
matrix is: Tab 2.1: scale meaning of APH
§ 1 c12 c13 ·
¨ ¸ ˄4.2˅ According to the national traffic accidents between
B4 = ¨ c21 1 c23 ¸ vehicle traffic accident statistic data to construct the target
¨c ¸ layer and rule layer of comparative judgment matrix. Table
© 31 c32 1 ¹
2.2 is all kinds of working condition of the traffic accident statistic data on average for 2011-2014.
Tab 2.2: All kinds of working condition of the traffic accident statistical data on average of 2011-2014
Target layer for criterion of the judgment matrix: road speed static conditions (test speed 0 to 50 km/h) and the
comprehensive accident casualty ratio statistics calculation
§ 1 1 1· under the working condition of each accounted for,
¨1 7 2
3
2¸ according to the different degrees of injury serious situation
¨ ¸ different set minor injuries, serious injuries, death in the
¨7 1 5 9 5¸
¨ ¸ ˄4.4˅ proportion of casualties to 1, 1.2, 1.5, after calculating the
1 proportion; Due to traffic accident statistics are not specific
¨2 1 3 1¸
A ¨ 5 ¸ to each working condition of the speed of the car accident
¨1 1 1 1¸ statistics of economic property losses, so can't by accident
¨ 1 ¸
¨3 9 3 3¸ economic loss to calculate weight of the working condition
¨ 1 ¸ of the situation. But in the process of accident statistics
¨2 1 3 1¸ analysis can be found that with the improvement of the
© 5 ¹
speed of the car accident caused economic losses will
increase gradually, so when calculating the economic loss of
Through computing the largest eigenvalue of matrix
weight is introduced into a linear with the increase of the
MATLAB lambda m = 5.0999, the normalized eigenvector k
speed conditions, meter rule layer in the hierarchy analysis
= (0.0878, 0.5888, 0.0878, 0.5888, 0.1389).The car static,
diagram is 1, each working condition of minimum speed
goals, low speed of target, the target car to slow down,
economic loss every time, increase the speed of the car
change lanes and target bias the weight of five conditions
economy loss weight increased by 0.1.Finally using the
were 0.0878, 0.5888, 0.1387, 0.0457, 0.1389.
analytic hierarchy process (AHP) to calculate weight of the
The weight of the total target of relative criterion layer
rule of each layer corresponding solutions.
after calculation, weights of criterion layer relative to their
respective solutions are also calculated. Due to the
particularity of AEBS evaluation, when calculating the rule
7KHDFFLGHQWUDWH
Speed km/h 10 20 30 40 50
Ratio of accidents 1.1% 9.5% 23.4% 23.7% 31.6%
Ratio of casualties 1.1% 9.5% 23.4% 29.15% 45%
Ratio of economic
16.7% 18.3% 20% 21.7% 23.3%
losses
Tab 2.3: Target car static condition scheme layer element weights
AEBS born to solve the problem of driving safety, so 50 km/h the weight of five conditions were 0.0381, 0.0804,
that when calculating the each speed integrated accident of 0.1996, 0.2375, 0.4444.
safety issues should be considered, so this article set weights ˄2˅Target vehicle at low speed condition
number is 0.2, the economic loss of ratio 0.2, casualties of According to (1) the target vehicle scheme layer
more than 0.6.Proportion of weighted scheme layer is shown theory and calculation of the static working condition,
in table 2.4. working condition of low speed target car scheme layer
weights using the same theory and method to calculate.
Speed Figure 2.3 for vehicle ahead slow accident rate and the
km/h number of casualties accident statistic histogram, the figure
Ratio shows as long as the distribution in vehicle ahead slow
traffic accident within 60-70 km/h.
Tab. 2.4 Working condition of a target car static solution
7KHDFFLGHQWUDWH
§ 1 1 1 1·
¨1 3 5 6 9¸
¨ ¸
¨3 1 1 1¸
1 ˄4.5˅
¨ 3 4 5¸
¨ ¸ ަԆ
B1 ¨5 1¸ 6SHHGNPK 0LQRULQMXU(G 6HULRXVO\LQMXUHG 'HDWK
3 1 1
¨ 3¸
¨ 1¸ Fig 2.3: A target car speed and working condition of
¨6 4 1 1 ¸
¨ 2¸ slow traffic accident casualties statistical histogram
¨9 5 3 2 1 ¸¹
©
Through computing the largest eigenvalue of matrix According to figure 2.3 of section on the statistical
MATLAB lambda m = 5.0918, the normalized eigenvector k data of traffic accidents and economic loss of weighted
= (0.0693, 0.1464, 0.0693, 0.1464, 0.8091).Which is 10 to proportion of calculation scheme layer number, the number
of casualties and economic losses are shown in table 4.5.
§ 1 1·
Speed km/h 40 50 60 70 ¨1 1 3 5 ¸
¨ ¸
Ratio of accidents 6.8% 9% 23.1% 34.5% ¨1 1 1 1 ¸ ˄4.6˅
Ratio of casualties 7.6% 9.9% 24.4% 41.6% B2 ¨ 3 5¸
¨ ¸
Ratio of economic ¨3 3 1 1 ¸
21.7% 23.9% 26.1% 28.3% ¨ 2¸
losses ¨5 5 2 1 ¸
© ¹
Tab 2.5: Target car lower speed condition scheme layer
element weights Through computing the largest eigenvalue of matrix
MATLAB lambda m = 4.0042, the normalized eigenvector k
Proportion of weighted scheme layer is shown in table = (0.1629, 0.1629, 0.4674, 0.1629).The 40-70 km/h the
2.6. weight of four conditions were 0.09893, 0.09893, 0.2838,
0.5183.
Spee ˄3˅The working condition of a target vehicle deceleration
d For the working condition of rule layer slows down
40 50 60 70 layer weighting from the front two kind of working
km/
h condition. Because everything slows down operating mode
Rati 10.27 12.51 24.48 37.73 choice is the speed of 50 km/h, so for this rule layer
o % % % % calculated when accounted for only refer to the actual urban
road traffic accident reduction speed of vehicles is
Tab 2.6: Working condition of a lower target car calculated.
solution layer weighted proportion
shown in table 2.8.
Speed
10/0 20/0 30/0 40/0 50/0 50/20
km/h
Ratio
7DUJHWYHKLFOHGHFHOHUDWLRQm/s2
Matrix derived 1 2 3 4 5 6 7 8 9
RI number 0 0 0.58 0.9 1.12 1.24 1.32 1.41 1.45
Tab 4.9: Hierarchy total sorts RI values mean random consistency index
CI value is level, total sorts consistency index is defined as: strategy with the standard of the same score, or reduce the
CI=˄¬P-n˅/˄n-1˅ ˄4.10˅ score accordingly.
The lambda m is the biggest characteristic value of In addition to alarm signal should be red color, voice
judgment matrix, n to judge the order of the matrix. signal should be enough to alert the driver.
Finally it is concluded that consistency indicators:
CI Om n ˄4.11˅
CR / RI
RI n 1
If CR<0.1ˈThe judgment matrix consistency is good.
In calculating target layer relative criterion layer,
criterion layer relative solution consistency of judgment
matrix is as follows:
Target layer relative criterion of consistency indicators:
CI Om n 5.0999 5
CR / RI /1.12 0.0222 0.1 Fig 4.5: AEBS best alarm strategy ADAC suggested
RI n 1 5 1
Rule layer target vehicle static condition judgment
matrix consistency indicators: 2.4 Bonus evaluation method
CI Om n 5.0918 5
CR / RI /1.12 0.0205 0.1 When AEBS evaluation, due to the working condition
RI n 1 5 1 of each test is a test, so there is a chance, this test method is
Rule layer target vehicle working condition of low proposed to increase AEBS points test evaluation method.
speed consistency of judgment matrix indicators: Considering the AEBS algorithm and work stability,
CI Om n 4.0042 4 increase the working condition of a target vehicle static
CR / RI / 0.9 0.0016 0.1
RI n 1 4 1 repeat test evaluation, test car speed of 30 km/h repeated
Rule layer target vehicle deceleration condition five times test, record test vehicles AEBS alarm time TTC,
judgment matrix consistency indicators: AEB -f TTC, test the consistency of the parameters of the
CI Om n 44 working condition of each and the consistency of different
CR / RI / 0.9 0 0.1 working conditions, if the consistency is good AEBS
RI n 1 4 1
Rule layer target vehicle bias condition judgment evaluation plus 1.
matrix consistency indicators:
CI Om n 6.2416 6 3 Results
CR / RI /1.24 0.039 0.1
RI n 1 6 1
In conclusion, the judgment matrix is conforming. (1) The introduction of analytic hierarchy process
(AHP), analytic hierarchy structure model is established, a
2.3 False alarm and HMI evaluation method target vehicle static condition, target vehicle at low speed
conditions, working condition of a target vehicle
AEBS in vehicles in danger, the driver to take the
emergency brake, when braking or insufficient braking deceleration condition and target vehicle bias conditions,
deceleration of the general will reach more than 8 m/s2, so if change lanes rule of five layers.
incurred during the process of actual driving false alarm and
emergency brake will not only seriously affect the driver's (2) According to the hierarchical structure model,
driving experience and comfort, and more likely to cause respectively, set up the total target layer relative to the
traffic accidents such as rail after. In AEBS evaluation, criterion of judgement matrix and guidelines for each
straight false alarm condition and corners false alarm
conditions was not happen false alarm points, in case of scheme layer of judgment matrix.
false alarm will deduct points. (3) Calculates the maximum characteristic value of each
For evaluation of HMI, European ADAC suggested that judgment matrix and weight. Calculated according to the
AEBS alarm strategy adopted as shown in figure 4.5, when
vehicles in danger ahead, AEBS first signal light, sound or theory of analytic hierarchy process (AHP) consistency
vibration alarm to remind the driver in front of the danger, if judging the consistency of judgment matrix is good.
the driver didn't immediately take braking or other risk of
(4) The last increase AEBS HMI, false alarm and bonus
driving behavior, AEBS will adopt a snub braking action
again to remind the driver, if the driver hasn't been any evaluation method.
response, AEBS will be a part of the brake, brake
completely until the vehicle stop collision avoidance or
reduce collisions [65].For this alarm strategy, this article
choose to use this strategy as evaluation AEBS alarm
strategy evaluation method, if the test AEBS use the alarm
References [12] Assessment methodologies for forward looking
integrated pedestrian and extension to cyclists
[1] Gao Yun, Xiao Fen.Factors that affect China's private safety.D1.3[R].2011.
car ownership and its change path analysis [J]. [13] Assessment methodologies for forward looking
Industrial technology economy, 2013, 9:41-47. integrated pedestrian and extension to cyclists
[2] AEB vulnerable road user assessment protocol for safety.D1.4[R].2011.
inclusion in assessment protocol – pedestrian [14] DOT-HS-810-905-Integrated Vehicle-Based Safety
protection[S].2014.9. Systems - Human Factors and Driver-Vehicle Interface
[3] EUROPean new car assessment programme-test (DVI) Summary Report[R].2008.2.
protocol – AEB VRU systems[S].2014.9 [15] 100875-IVBSS Verification Test Plans for Heavy
[4] IIHS Autonomous Emergency Braking Test Protocol Trucks[R].2008.5
(Version I)[S].2013.11 [16] 100876-VBSS Verification Test Plans for Light
[5] National Highway Traffic Safety Administration Vehicles[R].2008.5
forward collision warning system confirmation [17] GB Intelligent Transportation Systems — Forward
test[S].2013.2
vehicle collision warning systems — Performance
[6] European Parliament and of the Council with respect to
type-approval requirements for certain categories of requirements and test procedures [S]. ICS 03.220.20˗
motor vehicles with regard to advanced emergency 43.040.20.
braking systems[S].2012.4.16 [18] Made in China in 2025 - technology innovation green
[7] Intelligent transport systems — Forward vehicle book of key areas [Z]2015.10:114-116
collision mitigation systems — Operation, performance,
and verification requirements[S].2013.6.1 [19]Satty,T.L.The Analytie Hierarehy
[8] ADAC. Comparative test of advanced emergency Proeess[M].NewYOrk:MeGraw111,198.
braking systems[R].2012.7 [20] Huang Bin. Evaluation methods of passenger car
[9] ASSESS
D1.1-FINAL-Preliminary_Test_Scenarios-PUBLIC[R]. driving safety [D].2014.6:27-27.
2009.11.19. [21] RI values mean random consistency index [Z].baidu
[10] ASSESS Project Overview Active Test AHP: http://baike.baidu.com/link?url.
workshop[R].2012
[11] ASSESS D4.1b - FINAL - Action plan pre-crash
evalution PUBLIC[R].2010.03.01.
Acknowledgement
Thanks to all the colleagues for contributions to this thesis, evaluation method especially.
thanks to Xia Qin with the help of the calculation of AEB
Chen Chao
Intelligent Vehicle Testing & Evaluation Center Engineer
China Automotive Engineering Research Institute Co, Ltd.
TEL:15730255450
FAX:86-23-63423954
E-mail:chenchao@caeri.com.cn
Xia Qin
Intelligent Vehicle Testing & Evaluation Center Engineer
China Automotive Engineering Research Institute Co, Ltd.
TEL:18623198434
FAX: 86-23-63423954
E-mail:xiaqin@caeri.com.cn