Vous êtes sur la page 1sur 17

Fuzzy Sets and Systems 150 (2005) 457 473

www.elsevier.com/locate/fss

A car-following collision prevention control device based on the cascaded fuzzy inference system
Jeich Mar , Hung-Ta Lin
Department of Communications Engineering, Yuan-Ze University, 135 Yuan-Tung, Nei-Li, Taoyuan 320, Taiwan, ROC Received 1 October 2003; received in revised form 26 July 2004; accepted 9 September 2004 Available online 1 October 2004

Abstract A car-following collision prevention control device based on the cascaded fuzzy inference system (CFIS), consisting of a velocity fuzzy controller and an acceleration fuzzy controller, to nonlinearly control car acceleration or deceleration rate is proposed. The distance and speed relative to the car in front are measured using spread spectrum radar and applied to the collision prevention control device. The output acceleration or deceleration rate obtained from the CFIS car-following collision prevention system is based on the characteristics of the vehicle. The simulation results demonstrate that the presented CFIS control device can solve the oscillation problems for nal relative distance between the lead vehicle (LV) and following vehicle (FV) and relative speed. When the LV applies the brake suddenly or a stationary obstacle appears in front of vehicle moving at high speed on the roadway, the CFIS control device can safely avoid a collision. The CFIS car-following collision prevention control device proposed in this paper can provide a safe, reasonable and comfortable drive. 2004 Elsevier B.V. All rights reserved.
Keywords: Car-following; Cascaded fuzzy inference system; Spread spectrum radar

1. Introduction The car-following collision prevention is one of the key technologies for the intelligent vehicle-highway system (IVHS) [5,11] to control vehicle separation and trafc ow. Much work dealing with car-following collision prevention technology has been done [16,11]. In [4], the proportional, integrative plus derivative
Corresponding author. Tel.: +886 3 4638800; fax: +886 3 4639355.

E-mail address: eejmar@saturn.yzu.edu.tw (J. Mar). 0165-0114/$ - see front matter 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.fss.2004.09.004

458

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

(PID) algorithm was employed to design the combined throttle and brake controller. In [1], a Kalman lter algorithm was designed for acceleration estimation in the host car. The standard control systems require an inputoutput mathematical model for the car plant and its environment. General Motors (GM) research group [2,3] developed a car-following model that yielded various nal safe distances between the lead vehicle (LV) and following vehicle (FV) for a xed nal LV speed when the initial LV speed, FV speed or relative distance between the LV and FV was different [5]. This phenomenon was not consistent with reasonable car-following behavior in the real world. The conventional algorithms considered a deterministic one-to-one relationship existed between the LV action and FV reaction. However, the reactions of a driver to the actions of other drivers differ with drivers or conditions. Therefore, the reactions of a driver to the actions of other drivers are not based on a deterministic one-to-one relationship, but on a set of vague driving rules developed through experience from different drivers and conditions. It is assumed that a driver makes a decision as the result of a fuzzy reasoning process. The reaction possibilities for the FV are predicted based on this fuzzy reasoning process. A fuzzy system is a set of fuzzy rules that maps inputs to outputs. Fuzzy systems do not use mathematical inputoutput model or exact car parameters. An FIS car-following model with a proper membership functions and rule base is proposed in [5] which can solve both problems of the different nal relative safe distances and oscillations. Two hundred and sixteen fuzzy rules were used to update online the acceleration or deceleration rate of the FV. However, safe and comfortable drives are not considered in the car-following model. In [6], a combined brake/throttle fuzzy controller that uses a neural system to learn the fuzzy rules is designed to control the velocity and the distance between cars in single-lane platoons. An adaptive network fuzzy inference system (ANFIS) controller for a car-following collision prevention system was proposed in [7]. The consequent ANFIS parameters were updated online using a recursive least square algorithm. The frequency modulation continuous wave (FMCW) radar measured the distance and speed relative to the car in front for the ANFIS controller. Hierarchical fuzzy control was present to signicantly reduce the required rule-base size without compromising robustness and performance in [10]. In this paper, we utilize hierarchical fuzzy control principle to propose a cascaded fuzzy inference system (CFIS)-based car-following control device that receives the relative speed and distance between the LV and FV from spread spectrum radar, to nonlinearly control the acceleration or deceleration rate of the FV. The CFIS car-following control device, which consists of a 11-fuzzy-rule velocity fuzzy controller and a 3-fuzzy-rule acceleration fuzzy controller, is proposed to update online the acceleration or deceleration of the HV while following the LV. The number of fuzzy rules designed for the CFIS car-following control device is far fewer than that needed for the FIS car-following model in [5]. The proposed method saves computation time for the car-following control device. This paper is organized into six sections. In Section 2, the principle of the CFIS car-following collision prevention control device is described. Sections 3 and 4 introduce the basics of the CFIS for the car-following control device and the emergency safety device, respectively. The simulation tests are demonstrated in Section 5. Section 6 gives the concluding remarks.

2. Principle of the CFIS car-following collision prevention control device An effective car-following collision prevention system, as shown in Fig. 1, includes spread spectrum radar, a CFIS car-following control device, an emergency safety device and the vehicle interface [12]. The system measures the distance and speed relative to the car in front with spread spectrum radar mounted in

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473
Ant Antenna

459

Spread spectrum radar d,v

CFIS car-following control device

af Emergency safety device

The tune acceleration or deceleration rate ah of the host car

Fig. 1. Diagram of the car-following collision prevention system.

the front of the car that maintains a safe distance using the CFIS car-following control device. The spread spectrum radar applied to the car-following collision prevention system with different pseudo-noise (PN) codes can distinguish the echo signal of the lead vehicle from interference signals transmitted by other vehicles. The emergency safety device determines the acceleration or deceleration rate for the FV when the LV applies the brakes suddenly or an obstacle appears in front of the vehicle moving at high speed on highway. The vehicle interface connects the car-following collision prevention system with external vehicle sensors, such as the speedometer, brakes, steering sensor and warning LED display. The direct sequence spread spectrum radar employs a code division multiple access (CDMA) scheme to solve the ambiguity phenomenon of multiple targets in front in the antenna beam width area. The spread spectrum radar has a wideband digital signal processor (DSP). The DSP implements a pulse compressor that measures the relative distance between the LV and FV and uses a matched lter to calculate the relative speed between the LV and FV [8]. The direction of arrival of the reected signals from the vehicle ahead is measured using multibeam antennas installed on the front of the FV. The design objectives of the car-following collision prevention control device are to provide a safe, comfortable and reasonable ride. According to the trafc rules in Taiwan, as shown in Table 1, we set the safe distance (m) between the LV and FV as one half of the car speed (km/h). For example, if the small vehicles speed is 70 km/h, the safe distance between the FV and LV is set as 35 m. The vehicle cannot change its speed in real time maintaining a safe distance between the LV and FV when the acceleration or deceleration rate estimated by the car-following collision prevention control device is too high. The vehicle performance will be poor when rate is adjusted too low. Therefore, the acceleration and deceleration rates

460

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

Table 1 Safe distance between LV and FV The car speed (km/h) Safe distance (m) Small family cars Large family cars 60 30 40 70 35 50 80 40 60 90 45 70 100 50 80

Safe distance Velocity Fuzzy Controller Fuzzification 1 + _

X1

+
Re Relative distance d

Control rule base 1

Fuzzy inference mechanism 1

Defuzzification 1 vf + _ Relative speed v

+
X2 Acceleration Fuzzy Controller

Fuzzification 2

Control rule base 2

Fuzzy inference mechanism 2

Defuzzification 2

af

Fig. 2. Functional Diagram of the CFIS car-following control device.

must be properly chosen according to the characteristics of the vehicles. The emergency safety device to increase the safety for the CFIS control device further processes the estimated acceleration or deceleration rate. 3. CFIS car-following control device The CFIS based on the car-following control device, as shown in Fig. 2, includes a velocity fuzzy controller and an acceleration fuzzy controller, which is based on the pure fuzzy-type FIS [9]. Each of these controllers contains a single input and a single output membership function. The CFIS car-following

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

461

m1,0

m1,1 m1,2

m1,3

m1,4 m1,5 m1,6 m1,7

m1,8 m1,9

m1,10

A1,0

A1,1

A1,2 A1,3

A1,4

A1,5

A1,6 A1,7

A1,8 A1,9

A1,10

X1

Fig. 3. The Triangular membership function for velocity fuzzy controller.

control device uses the relative distance and speed, measured from the spread spectrum radars, to estimate the FVs acceleration or deceleration rate periodically. The period is 0.1 s, which corresponds to 2.78 m for the relative speed of 100 km/h between the FV and LV. The FIS velocity fuzzy controller has single input and single output. The rule base I contains eleven fuzzy if-then rules. The membership functions m1,i , i = 0, 1, . . . , 10 are triangular. The triangular-shaped membership functions m1,i , i = 0, . . . , 10, are shown in Fig. 3, where the input X1 is the difference and safe distance for the car-following de , i.e., X1 = d de . between the measured relative distance d The design strategy of fuzzy rule base I is based on the fact that the initial speed of the FV is determined by the difference between the measured relative distance and safe distance for the car-following under the comfortable driving condition. An input X1 of the membership function determines the membership grades 1,i , i = 0, . . . , 10. In Taiwans highway, the maximum vehicle speed is limited to 100 km/h for small cars. According to the Taiwan trafc rule, the car-following safe distance de should be set as 50 m. If the required time interval for the FV to recover the car-following safe distance, i.e., X1 = 0, is expected to be 5 s when X1 = 40 or 40, then the initial relative speed and the acceleration of the FV are calculated as follows: 1 a 2 V0 t+ t = X1 , 3.6 2 3.6 V0 + at = 0. (1) (2)

The initial relative speed between the FV and LV and the acceleration of the FV are solved as V0 = 57.6 or 57.6 km/h and a = 11.52 or 11.52 km/h/s, respectively, which are used to compute the values of X1 and V0 for t = 0, 1, 2, 3, 4. For example, X1 = 25.6 or 25.6 and V0 = 46.08 or 46.08 km/h are computed for t = 4. Therefore, the boundary values A1,i of the triangular membership functions can be generated by 57.6 1 11.52 2 , i = 0, 1, . . . , 5, 40 3 .6 i + 2 3.6 i A1,i = 40 57.6 (10 i) + 1 11.52 (10 i)2 , i = 6, 7, . . . , 10, 3.6 2 3.6 = {40, 25.6, 14.4, 6.4, 1.6, 0, 1.6, 6.4, 14.4, 25.6, 40}. (3)

462

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

The corresponding outputs of fuzzy inference mechanism I,C1,i , are determined as C1,i = 57.6 + 11.52 i, i = 0, 1, . . . , 5, 57.6 + (11.52) (10 i), i = 6, 7, . . . , 10, = {57.6, 46.08, 34.56, 23.04, 11.52, 0, 11.52, 23.04, 34.56, 46.08, 57.6}.

(4)

The speed of the FV increases with X1 in order to reduce the distance between the FV and LV if X1 0. The speed of the FV decreases with X1 in order to increase the relative distance between the FV and LV if X1 < 0. If the expected time interval to recover the car-following safe distance is less than 5 s for X1 = 40 or 40, a larger acceleration or deceleration rate is require. It may cause uncomfortable driving. The fuzzy rule base I uses eleven fuzzy if-then rules to produce the corresponding output Y1,i . IF X1 is m1,i THEN Y1,i is C1,i , i = 0, 1, . . . , 10. (5)

The weighted average defuzzication method is used to generate the reference relative speed vf (km/h) between the FV and LV: vf =
9 i =0 1,i Y1,i . 9 i =0 1,i

(6)

The reference relative speed vf (km/h) and the measured relative speed v between the FV and LV are used to control the reference acceleration or deceleration rate of the FV. The input X2 of the acceleration fuzzy controller is the difference between the reference relative speed and the measured relative speed between the FV and LV: X2 = vf v, (7)

, whereas X2 < 0 corresponds to vf < v . The acceleration fuzzy controller X2 0 corresponds to vf v has the same architecture as the velocity fuzzy controller. The single input is X2 and the single output is the reference acceleration or deceleration rate of the FV in km/h/s units. The rule base II contains three fuzzy if-then rules. The membership functions m2,i , i = 0, 1, 2 are triangular. When X2 > 0, the acceleration rate of the FV increases with the X2 value. When X2 < 0, the deceleration rate of the FV increases with the absolute value of X2 . The vehicle cannot change the speed in real time in order to remain the safe distance between the LV and FV when the acceleration and deceleration rate estimated by the controller are too high. The vehicle performance will be degraded when the estimated acceleration and deceleration rate are too low. Therefore, the acceleration and deceleration rate must be properly chosen according to the characteristics of the vehicles. Here we assume that the required time to accelerate the vehicles from 0 to 100 km/h is 16.67 s and the required time to decelerate the vehicles from 100 to 0 km/h is 6.67 s. Therefore, the range of output membership function is dened as 15 to 6 (km/h/s). If A2,0 > C2,0 and A2,2 < C2,2 are set, when X2 converges to the range of (15, 6), the value of the reference acceleration or deceleration rate af is too large. The oscillation of the nal safe distance between the LV and FV occurs. If the A2,0 is set too small and A2,2 is set too large, the FV needs more time to reach X2 = 0. The design strategy of the acceleration fuzzy controller is to consider the comfort of the driver in addition to guarantee the driving safety. The acceleration fuzzy controller will gradually adjust the acceleration or deceleration rate of the FV. Through the several simulation tests, the

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

463

CFIS car-following control device is determined with A2,0 = 3C2,0 and A2,2 = 3C2,2 , which can solve the oscillations. The corresponding outputs of fuzzy inference mechanism II,C2,i , are determined as C2,i = {15, 0, 6}, i = 0, 1, 2. (8)

The boundary values A2,i of the triangular membership functions are determined as A2,i = {45, 0, 18}, i = 0, 1, 2. (9)

The fuzzy rule base II uses three fuzzy if-then rules to get the corresponding output Y2,i . IF X2 is m2,i THEN Y2,i is C2,i , i = 0, 1, 2. (10)

The weighted average defuzzication method is used to generate the reference acceleration or deceleration rate af for the FV. af =
10 i =0 2,i Y2,i , 10 i =0 2,i

(11)

where 2,i is the membership grade of m2,i . In the car-following mode, the wanted tuning acceleration or deceleration rate ah (n) for the FV is ah (n) = af (n) (km/h/s). (12)

4. Emergency safety device For guaranteeing the driving safety of the CFIS system, the emergency safety device is designed in the proposed system. During the car-following modes, the driver of the LV may suddenly apply the brake or the other car in the adjacent lane may suddenly change to the same lane as the FV. The FV needs to use the larger deceleration rate, which will result in the degradation of the comfort, to avoid collision. The emergency safety control mode is on only when the distance difference X1 is negative (when the FV is very close to the LV) and the relative velocity v is positive (when the FV goes faster than the LV). The acceleration rate or deceleration rate of the LV at t=tn is determined by at (n) = t (n 1) v t (n) v (km/h/s), t (13)

where v t (n) and v t (n 1) are speeds of the LV at sample time tn and tn1 , respectively. t is the sample time interval of the relative speed and distance measured by the spread spectrum radar. v t (n) = vh (n) v(n), (14)

where vh (n) is the speed of the FV and v(n) is the measured relative speed between the LV and the FV. Since the collision probability increases during the emergency safety mode, the reference acceleration or deceleration rate must be adjusted in order to improve the driving safety of the FV. According to Fig. 4,

464

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473
relative speed df

v acceleration rate = -15 km/hr/sec

t1

t (sec)

Fig. 4. Diagram for the computation of the df reduced relative distance.

the reduced distance between the FV and LV is given as df = 1 v v 2 t1 = (m), 2 3.6 108 (15)

to zero. where t1 is the required time to reduce v For d df Lv (an average vehicle length of about 5 m) case, the wanted tuning acceleration or deceleration rate for the FV is adjusted as ah (n) = at (n) af (n) if at 15 (km/h/s) . otherwise (16)

df < Lv case, the deceleration rate of 15 km/h/s for the FV cannot maintain the car-following For d safe distance between the FV and LV. The FV needs to use the larger deceleration rate to maintain the car-following safe distance between the FV and LV. If the maximum deceleration rate of the FV is often used, it may cause uncomfortable driving. Therefore, when v = 10 km/h and at (n) 10 km/h/s, we let the FVs acceleration rate ah (n) equal at (n) 5 km/h/s, i.e., the relative acceleration rate between the FV and LV is 5 km/h/s. The relative speed v between the FV and LV will reduce to zero from 10 km/h after 2 s and the relative distance between the FV and LV decreases to 2.78 m. The acceleration or deceleration rate of the FV is determined as if v > 10 km/h, 30 km/h/s 10 km/h, at (t) 10 km/h/s, (17) ah (n) = at (n) 5 km/h/s if v af (n) otherwise, where 30 km/h/s is the maximum deceleration rate of the vehicles. 5. Simulation results To demonstrate the performance of the proposed car-following collision prevention system, several simulation tests were performed. In these tests the input relative speed and distance between the LV and

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473
25

465

Relative speed (km/hr)

20 15 10 5 0

measured relative speed real relative speed

10

12

14

16

18

20

(a)
50
measured relative distance real relative distance

Relative distance (m)

40

30 20

10 0 2 4 6 8 10 12 14 16 18 20

(b)

Time (sec)

Fig. 5. (a) Relative speed, (b) relative distance between the LV and FV in Case 1.

FV were simulated according to the output from the spread spectrum radar. The car-following scenarios are grouped into four categories: (1) distance between the LV and FV safe distance, speed of the FV speed of the LV; (2) distance between the LV and FV safe distance, speed of the FV < speed of the LV; (3) distance between the LV and FV < safe distance, speed of the FV speed of the LV; and (4) distance between the LV and FV < safe distance, speed of the FV < speed of the LV. The spread spectrum radar parameters were chosen as follows for the simulation tests. The carrier frequency fc = 60 GHz; the period of spread spectrum modulation T = 1.33 s, pulse duration Tb = 0.84667 s, and chip time Tc = 6.67 ns; the maximum detection range Rmax = 200 m; the minimum detection range R = 1 m; the maximum estimated error of the relative speed could be reduced to 0.15 km/h within 0.84667 s processing time [8]; the receiver bandwidth fb = 1/Tc = 150 MHz; and the processing gain = 21 dB. The required relative speed and distance were sent into the car collision prevention control device at a t = 0.1 second sample interval that corresponds to a relative distance of 2.78 m for 100 km/h relative speed. There are three cases (Cases 13) that correspond to the example presented in the GM and Kikuchi & Chakroborty models [5]. Cases 13 belong to the rst car-following category. Case 4 is an example related to the second, the third and fourth car-following categories. Cases 5 and 6 demonstrate that the CFIS car-following collision prevention control device can provide a safe drive when an emergency occurrs. Case 1: The initial relative distance between the LV and FV equals 40.53 m. The initial FV speed equals 48.384 km/h. The LV decelerates from 48.384 to 30.816 km/h in 2 s and remains constant. Case 2: The initial relative distance between the LV and FV equals 40.53 m. The initial speed of the FV equals 57.168 km/h. The LV decelerates from 57.168 to 30.816 km/h in 3 s and remains constant.

466
10

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

Acceleration (km/hr/sec)

0 -5 -10 0 2 4 6 8 10 12 14 16 18 20

(a)
60 55

Speed (km/hr)

50 45 40 35 30 0 2 4 6 8 10 12 14 16 18 20

(b)

Time (sec)

Fig. 6. (a) Acceleration or deceleration rate, (b) speed of the FV in Case 1.


45 case 1 case 2 40

Real relative distance (m)

35

30

25

20

15 0 2 4 6 8 10 12 14 16 18 20

Time (sec)

Fig. 7. Distance between the LV and FV for CFIS-based collision prevention control device with a different initial relative speed.

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473
45 case 1 case 3 40

467

Real relative distance (m)

35

30

25

20

15 0 2 4 6 8 10 12 14 16 18 20

Time (sec)

Fig. 8. Different initial relative distance between the LV and FV for CFIS-based collision prevention system.
15

Relative speed (km/hr)

measured relative speed real relative speed

10

0 -5 0 2 4 6 8 10 12 14 16 18 20

(a)
30

Relative distance (m)

29

28
measured relative distance real relative distance

27 26 0 2 4 6 8 10

12

14

16

18

20

(b)

Time (sec)

Fig. 9. (a) Relative speed, (b) relative distance between the LV and FV in Case 4.

468
5

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

Acceleration (km/hr/sec)

-5 -10

-15

10

12

14

16

18

20

(a)
70

Speed (km/hr)

65

60

55 0 2 4 6 8 10 12 14 16 18 20

(b)

Time (sec)

Fig. 10. (a) Acceleration or deceleration rate, (b) speed of the FV in Case 4.

Case 3: The initial relative distance between the LV and FV equals 45 m. The initial speed of the FV equals 48.384 km/h. The LV decelerates from 48.384 to 30.816 km/h in 2 s and remains constant. Case 4: The initial relative distance between the LV and FV equals 28 m. The initial speed of the FV equals 70 km/h. The initial speed of the LV equals 60 km/h and remains constant. Case 5: The initial relative distance between the LV and FV equals 50 m. The initial speed of the FV equals 100 km/h. The initial speed of the LV equals 0 km/h and the vehicle remains stationary. This case is used to test the collision prevention capability of the emergency safety device for an obstacle on the roadway in front of the vehicle when it is moving at a high speed. Case 6: The initial relative distance between the LV and FV equals 40 m. The initial speed of the FV equals 100 km/h. The LV decelerates from 90 to 40 km/h in 2 s and remains constant. The decelerate rate of the LV equals 25 km/h/s. The simulation test for Cases 1 and 2 shows the collision prevention control device performance under different initial LV speed and FV speed. Cases 1 and 3 will demonstrate the car-following performance of the collision prevention control device when initial distances between the LV and FV are different. Cases 5 and 6 will demonstrate that the emergency braking device can provide a safe drive when an emergency occurrs. Figs. 5 and 6 show the simulation results for Case 1. Figs. 5(a) and (b) illustrate that the required processing time to converge the nal distance between the LV and FV to a stable safe distance is about 10 s. This means that the distance between the LV and FV is equal to a safe distance and the relative speed is equal to zero. The Kikuchi and Chakroborty model required about 1214 s [5]. The non-smooth measured relative distance between the LV and FV was caused by the range resolution of the spread

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

469

Fig. 11. (a) Illustrating the operation of the emergency safety control device, (b) acceleration or deceleration rate, (c) speed of the FV in Case 5.

470
150

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

Relative speed (km/hr)

measured relative speed real relative speed

100

50

10

12

14

16

18

20

(a) Relative distance (m)


50 40 30 20 10 0 0 2 4 6 8 10 12 14 16 18 20
measured relative distance real relative distance

(b)

Time (sec)

Fig. 12. (a) Relative speed, (b) relative distance between the LV and FV in Case 5.

spectrum radar. No oscillation occurred in the nal relative speed and relative distance between the LV and FV. The acceleration or deceleration rate of the FV did not suddenly change, as shown in Fig. 6(a), and the curve of the FVs speed was smooth, as shown in Fig. 6(b). This means that the CFIS car-following collision prevention system provides a comfortable drive. Fig. 7 represents the simulation results for the proposed model for Cases 1 and 2. The nal distance between the LV and FV converged to a safe distance regardless of the different initial FV speeds. Oscillations in the distance between the LV and FV were not present. Fig. 8, which represents the simulation results for the proposed model for Cases 1 and 3, shows that the nal relative safe distance settled to approximately 15 m regardless of the different initial distance between the LV and FV (40.53, 45 m). Figs. 9 and 10 are the simulation results for Case 4. Figs. 9(a) and (b) show that the CFIS controller can be applied to car-following conditions that are second, third and fourth car-following categories and show that the required convergence time to a stable safe distance is about 8 s. Because the acceleration or deceleration rate of the FV does not suddenly change, as shown in Fig. 10(a), the curve of the FVs speed is smooth, as shown in Fig. 10(b), which means that the CFIS car-following collision prevention system provides a comfortable drive. Figs. 11 and 12 are df < 5 m during the simulation results for Case 5. In Fig. 11(a), because X1 < 0 m, v > 10 km/h and d time interval 03 s, the CFIS control device output is further processed by the emergency safety device to control the speed of the FV. The deceleration rate was set to 30 km/h/s. After the 3.1 second time instant, the measured relative speed between the LV and FV reduced to below 10 km/h. The CFIS control device switched to the reference acceleration or deceleration rate to smoothly control the FVs speed. Figs. 11(b) and (c) show the variations in the acceleration or deceleration rate and the FV speed during the 010 s expansion time interval, respectively. Figs. 12(a) and (b) show that the required processing time

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

471

Fig. 13. (a) Illustrating the operation of the emergency safety control device, (b) acceleration or deceleration rate, (c) speed of the FV in Case 6.

472
25

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

Relative speed (km/hr)

20 15 10 5 0 0 2 4 6 8 10 12 14

measured relative speed real relative speed

16

18

20

(a)
40

Relative distance (m)

measured relative distance realrelative distance

35

30

25

20 0 2 4 6 8 10 12 14 16 18 20

(b)

Time (sec)

Fig. 14. (a) Relative speed, (b) relative distance between the LV and FV in Case 6.

for the FV to achieve a safe distance is about 3.5 s while the relative speed is equal to zero. Figs. 13 and df 5 m 14 show the simulation results for Case 6. In Fig. 13(a), because X1 < 0 m, v > 0 km/h, d and a t 15 km/h/s during the 0 to 1.2 second time interval, the CFIS control device uses the emergency safety device to control the speed of the FV. The deceleration rate was set to 25 km/h/s. When X1 0 m, df 5 m and a v > 0 km/h, d t 15 km/h/s during 1.32 second time interval, the CFIS control device switched to the reference acceleration or deceleration rate to smoothly control the FVs speed. Figs. 13(b) and (c) show the variations in the acceleration or deceleration rate and the FV speed during the 020 s expansion time interval, respectively. Figs. 14(a) and (b) show that the CFIS control device can safely avoid a vehicle collision and make the FV smoothly achieve the nal safe car-following distance when the LV rapidly decelerates. 6. Conclusions A new car-following collision prevention control device design based on the cascade FIS has been proposed in this paper. The simulated relative distance and relative speed measured from the spread spectrum radar were applied to the car-following control devices to test the collision prevention system performance. The simulation results demonstrated that the car-following collision prevention control device proposed in this paper has the following characteristics: 1. From the simulation results of Cases 13, when the nal relative speed reaches zero, the distance between the LV and FV eventually converges to the same safe distance regardless of the initial LV

J. Mar, H.-T. Lin / Fuzzy Sets and Systems 150 (2005) 457 473

473

speed, initial FV speed and initial distance between the LV and FV. This means that the collision prevention system provides a reasonable drive. 2. No oscillation occurred in the FV speed regardless of the initial conditions. This means that the collision prevention control device provided a comfortable drive. 3. When the LV applied the brakes suddenly or a stationary obstacle appeared in front of the vehicle, the nal relative distance was always greater than zero, which means that the collision prevention control device provided a safe drive. References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] A. Bjornberg, Autonomous intelligent cruise control, 44th IEEE Vehicular Technology Conf., 1994, pp. 429433. D.C. Gazis, R. Herman, R.B. Potts, Car-following theory of steady state trafc ow, Oper. Res. 6 (1958) 165184. D.C. Gazis, R. Herman, R.B. Rothery, Non-linear follow the leader models of trafc ow, Oper. Res. 9 (1960) 545567. P. Ioannou, Z. Xu, S. Eckert, D. Clemons, T. Sieja, Intelligent cruise control: theory and experiment, Proc. of the 32nd Conf. on Decision and Control, December 1993, pp. 18851890. S. Kikuchi, P. Chrkroborty, Car-following model based on fuzzy inference system, Transport. Res. Rec. 1365 (1993) 8291. H.M. Kim, J. Dickerson, B. Kosko, Fuzzy throttle and brake control for platoons of smart cars, Fuzzy Sets and Syst. 84 (1996) 209234. J. Mar, F.J. Lin, An ANFIS controller for the car-following collision prevention system, IEEE Trans. Veh. Technol. 50 (2001) 11061113. J. Mar, H.T. Lin, The Doppler estimator operating with the matched lter for a vehicular spread spectrum radar, Electroni. Lett. 38 (16) (2002) 911912. T.J. Ross, Fuzzy Logic with Engineering Applications, University of New Mexico, McGraw-Hill, 1955. J. Stufebeam, N.R. Prasad, Hierarchical fuzzy control, Fuzzy Systems Conf. Proc., FUZZY-IEEE 99, 1999 IEEE International Published, 1 (1999) 31703175. P. Varaiya, Smart cars on smart roads: problem of control, IEEE Trans. Automat. Control 38 (2) (1993) 195207. J.D. Woll, VORAD collision warning radar, IEEE Internat. Radar Conf., 1995, pp. 367372.

Vous aimerez peut-être aussi