F2000I398 Seoul 2000 FISITA World Automotive Congress June 12-15, 2000, Seoul, Korea A Throttle/Brake Control Law for Vehicle Intelligent Cruise Control Kyongsu Yi1) *, Youngjoo Cho1), Sejin Lee1), Joonwoong Lee2) and Namkyoo Ryoo2) 1) School of Mechanical Engineering, Hanyang University, Seoul, 133-791, KOREA 2) Hyundai Motor Company, Kyunggi-Do, 445-850, KOREA A throttle/brake control law for the intelligent cruise control(ICC) systems has been proposed in this paper. The ICC system consists of a vehicle detection sensor, a controller and throttle/brake actuators. For the control of a throttle/brake system, we introduced a solenoid-valve-controlled electronic vacuum booster(EVB) and a step-motor controlled throttle actuator. Nonlinear computer model for the electronic vacuum booster has been developed and the simulations were performed using a complete nonlinear vehicle model. The proposed control law in this paper consists of an algorithm that generates the desired acceleration/deceleration profile in an ICC situation, a throttle/brake switching logic and a throttle and brake control algorithm based on vehicle dynamics. The control performance has been investigated through computer simulations and experiments. The results indicate the proposed ICC throttle/brake control law can provide satisfactory vehicle-to-vehicle distance and velocity control performance. Keywords: Intelligent Cruise Control, Optimal Design, Electronic Vacuum Booster, Kalman Filter, Duty Ratio 1. INTRODUCTION Sliding control has been used for vehicle longitudinal control to deal with nonlinearities in vehicle dynamics [3,8]. A neural net has been used to compute the desired vehicle acceleration for an ICC [6]. It was indicated that a linear PID-controller could not provide satisfactory performance because the controller was not able to handle the noisy sensor data, resulting in a jerky driving behavior [2]. Compared to the ordinary cruise control, the goal of an ICC is to prevent the vehicle-to-vehicle spacing from dropping an unsafe level. The throttle and brakes should be gently controlled so that the driver is aware that the ICC controller has taken over, but is not surprised by this action. In order to achieve high customer acceptance, an ICC system has to perform similarly to an experienced human driver. This paper describes a throttle/brake control law for vehicle intelligent cruise control. The ICC system considered in this study consists of a vehicle-detectionsensor, a controller and throttle/brake actuators. For the control of a throttle/brake system, a solenoid-valvecontrolled electronic vacuum booster (EVB) and a stepmotor controlled throttle actuator have been used. A nonlinear computer model for the electronic vacuum booster has been developed and the simulations were performed using a complete nonlinear vehicle model. The proposed control law in this paper consists of an algorithm that generates the desired acceleration/ deceleration profile in an ICC situation, a throttle/brake switching logic and a throttle/brake control algorithm. The control performance has been investigated through computer simulations and vehicle tests. The results indicate the proposed ICC throttle/brake control law can provide satisfactory Active safety systems and driver assistant systems have been active topics of research and development since 1990’s due to the potential for increased vehicle safety and improved driving comfort. One of the driver assistant systems currently under development by most automotive manufacturers around the world and recently commercialized by several companies is Intelligent Cruise Control (ICC) system. Ordinary cruise control systems for passenger cars are becoming less and less meaningful because the increasing traffic density rarely makes it possible to drive at a pre-selected speed. ICC systems control both speed and distance to preceding vehicles and can improve the driving comfort and reduce the danger of rear-end-collision. Although there already exist product ICC systems, the bandwidth of such systems is very low and the headway distance (the safe vehicle-to-vehicle distance to avoid collisions) is large. Therefore, these systems are almost useless on the busy urban traffic highway, and it is concluded that the bandwidth of the longitudinal vehicle control system should be increased significantly to reduce the headway distance and to be meaningful in the busy urban traffic highway. There has been a lot of research conducted on the vehicle longitudinal control including the ICC [1-7]. There have been many attempts to apply PID-type control laws [4] in the development of an ICC. Gain Scheduling and Adaptive Scheme have been used in ICC development to meet the various situations encountered by the controlled vehicle [5,6]. Linear Quadratic (LQ) and Linear Quadratic Gaussian (LQG) optimal control theories have been used to develop a car following algorithm [7]. * kyongsu@email.hanyang.ac.kr 1 performances in vehicle-to-vehicle distance control and velocity control. Intake Manifold Throttle 2. ICC VEHICLE Fig. 1 shows a vehicle longitudinal control system. The system consists of a radar sensor, a controller (ECU), a brake actuator (active booster) and a throttle actuator. An ICC vehicle considered in this study, a radar sensor, throttle and brake actuators are shown in Figure 2. The vehicle is equipped with a MMW radar distance sensor, a controller, a solenoid-valve-controlled ElectronicVacuum-Booster (EVB) and a step-motor controlled throttle actuator. Electronic Vacuum Booster air Engine solenoid Master Cylinder ICC ECU Active Booster Vacuum Booster Throttle Actuator Brake Pedal Radar Wheel Fig. 1 Vehicle longitudinal control system Fig. 3 A schematic diagram of the EVB brake actuator system Brake Actuator 3. THROTTLE/BRAKE CONTROL ALGORITHM Throttle Actuator The throttle and brakes should be gently controlled so that the driver is aware that the ICC controller has taken over and is not surprised by the control action. It has been reported that automotive decelerations up to 2.5 m / sec 2 were comfortable to human passengers [10]. The throttle/brake control algorithm has been designed such that the vehicle deceleration should not exceed this limit. The maximum of the vehicle acceleration has been limited to 1 m / sec 2 in order to prevent the kick-down of the automatic transmission. MMW Radar Fig. 2 An ICC vehicle and equipments 3.1 ICC ALGORITHM A schematic diagram of the EVB brake actuator system is shown in Figure 3. Vacuum booster differential pressure is controlled by a PWM solenoid-valve. Modeling of the EVB, simulation and experiment results on the dynamic characteristics and control performance of the EVB have been described in detail in [9]. It is indicated that steady state value of the differential pressure is proportional to the duty ratio. It should be noted that time constant varies significantly depending on the duty inputs and the steady state values of the EVB differential pressure have significant nonlinear characteristics. The nonlinear characteristics are due to friction in the valve and air flow characteristics. A stepper motor has been used as the throttle actuator. The stepping time of this motor is 4 milliseconds and the step angle is 0.75 degree. A block diagram of the vehicle and ICC algorithm is shown in Figure 4. The distance to a preceding vehicle and the relative velocity are measured using a MMW radar sensor. The distance and relative speed to the preceding vehicle and the pre-selected speed are fed to a ICC controller. Comparison of the headway distance and the distance to the preceding vehicle is used to determine control mode between the speed control and the distance control. In the case of the speed control, the controller works like a conventional cruise control. The controller controls the throttle and brakes such that the vehicle acceleration tracks the desired acceleration, which is designed so that the vehicle-to-vehicle distance converges smoothly to the headway distance. 2 In this study, ρ = 4 and r = 5 have been used. In the case that the ICC senses a cut-in vehicle, the control law, u = − k ⋅ x , demands large, uncomfortable accelerations. In order to avoid large accelerations, which deteriorate ride comfort, the desired acceleration, a des , has been obtained using a saturation function and a second order filter as follows: a des ω2 = 2 (5) u sat s + 2ς ⋅ ω ⋅ s + ω 2 Set : ( Headwaytime, Speed ) ICC controller Target detected? : Yes Target detected? : No Speed Control Optimal Design Desired Accel. / Decel. Switching Logic if u ≥ u max u max u sat = sat (u ) = u if u min < u < u max (6) u if u ≤ u min min where u = −k ⋅ x = −(k1 ⋅ x1 + k 2 ⋅ x 2 + k 3 ⋅ x3 ) . The filter damping ratio, ς , of 1 and the cutoff frequency, ω , of 5 Throttle/Brake Control Algorithm Sensors - Relative distance - Relative velocity Duty-ratio Throttle Throttle Actuator Brake Actuator rad / sec have been used. u min of –2.5 m / sec 2 and u max of 1 m / sec 2 have been used to provide comfortable ride quality and to avoid the kick-down of the automatic transmission during the speed control mode. This approach can saturate the vehicle jerk and acceleration at some maximum value. Vehicle Fig. 4 ICC algorithm 3.2 DESIRED ACCELERATION PROFILE 3.3 THROTTLE/BRAKE SWITCHING LOGIC Linear optimal control theory has been used to design a desired acceleration. Using integrators to model the vehicles, a state space model for the ICC and preceding vehicles can be written as follows: 0 1 0 0 0 x = Ax + Bu + Γw = 0 0 0 x + − 1u + 1 w (1) 0 0 0 1 0 The states are x T = [ x1 x2 x3 ] = [ x p − xcc v p − vcc Depending on the desired acceleration that the ICC vehicle must follow, the ICC controller applies throttle or brake control. Fig. 5 shows a switching line with hysteresis. The switching line indicates the vehicle acceleration (the minimum acceleration, a min ) when the throttle is closed ( α = 0 ) for a given vehicle velocity. The minimum acceleration line has been used as a switching line in the throttle/brake controls. The ICC controller applies throttle control when a des ≥ a min + h or brake control when vcc ] , the input, u , is the ICC vehicle acceleration and the disturbance, w , is the preceding vehicle acceleration. x and v indicate position and velocity, respectively. Subscripts, p and cc , indicate the preceding and ICC vehicles, respectively. The gains for the state feedback law, u = − k ⋅ x , are chosen to minimize the cost function: ∞ ( ) ∞ ( a des ≤ a min − h for a given vehicle speed. Switching logic with hysteresis is necessary to avoid frequent switching between throttle and brake controls. 0.0 ) 2 2 2 Acceleration [m/sec ] J = E[∫ ed + ρ ⋅ ev + r ⋅ u2 dt] =E[∫ xT Q x + uT R u dt] (2) 0 0 ed and ev are the distance and velocity errors, respectively, and defined as follows: ed = d h − ( x p − xcc ) = x3 ⋅ t h − x1 , ev =v p −vcc = x2 (3) where d h and t h (=1.2 seconds) are the headway time distance and headway time, respectively. The weighting factors, ρ and r , are chosen to give a tradeoff between performance and ride comfort. The matrices, Q and R , are defined as follows: 0 − th 1 ρ 0 , R = [r ]. Q= 0 (4) − t h 0 t h 2 -0.2 Throttle Control -0.4 h h α=0 -0.6 Brake Control -0.8 0 20 40 60 80 100 120 140 Velocity [km/hr] Fig. 5 Throttle/brake switching 3.4 THROTTLE CONTROL 3 160 180 At the low level of acceleration, wheel slip is quite small. The no-slop assumption has been incorporated in previous throttle/brake control designs for vehicle longitudinal control in Intelligent Cruise Control or in Automated Highways [3,8]. A throttle control law has been derived under a no-slip condition of the driving wheels. A block diagram of the throttle control algorithm is shown in Fig.6. The dynamic equation for a vehicle model of a typical passenger car encountering a grade with angle θ is[1] dv Mv = Ft − Fa − M v g sin θ = Ft − FL (7) dt where M v is the vehicle mass, v the vehicle speed, Ft the throttle opening angles as a function of the engine speed and torque. The throttle angle command, α des , has been computed from the computed throttle angle and the vehicle acceleration feedback using the Proportional-plusIntegration control as follows: ∫ α des = α f + K p × ( ades − a ) + K i × (ades − a )dt (14) where a is the vehicle acceleration and K p , K I gains. It has been recognized that the control law with large gains results in a very jerky driving behavior due to the torque production delay of the engine. Small gains have been used in this study. The actual throttle angle is controlled by the stepper motor to minimize the error between the actual and the desired throttle angles. tire longitudinal force at the wheel, Fa the aerodynamic drag force, g the gravitational constant, θ the road grade and FL the driving resistance load. Since the inertia of the Desired Acceleration wheel and axle, J w , is relatively small compared to the vehicle mass, i.e.: J (8) M v >> w2 r where r is the effective tire radius, the tire longitudinal force (the tire tractive force), Ft , can be written as follows: 1 1 Ft ≈ Ts − Tb (9) r r where Ts is the driving axle shaft torque and Tb the total brake torque. For a given desired acceleration, the required tire tractive force is computed as follows: Ft = M v a des + FL (10) When the desired acceleration for a given vehicle velocity is greater than the switching line, i.e., the throttle control region, the desired shaft torque, Ts , is computed from the equations (9) and (10) as follows: Ts = r ( M v a des + FL ) (11) Desired Tractive Force ades Ftr r Desired Shaft Torque Rd Rgi Desired Turbine Torque Torque-Converter Model Tt Tp Ts Engine Map αf PI control K p × (ades − a) + Ki × ∫ (ades − a)dt + + αdes Actual Acceleration a Fig. 6 Throttle control algorithm 3.5 BRAKE CONTROL The brake torque is applied only when the engine braking is not sufficient to follow the desired acceleration profile. A block diagram of the brake control algorithm is shown in Fig.7. When the desired acceleration for a given vehicle velocity is smaller than the switching line, i.e., the brake control region, the desired brake torque, Tb ,des , is computed from the equations (9) and (10) as follows: Tb ,des = −r ( M v a des + FL ) + Ts (15) The shaft torque, Ts , is computed using the engine map as follows: 1 Ts = Tnet (ω e ,0) (16) Rg The driving resistance load, FL , changes relatively slowly and the estimated value of the load can be used for the computation of the desired shaft torque, Ts . It has been presented in recent researches that the estimated value of the driving resistance load can be obtained from the vehicle longitudinal acceleration, engine speed and wheel speed measurements. The desired engine net torque, Tnet , can be computed from the desired shaft torque as follows: Tnet = R g Ts (12) Since the total brake torque is proportional to the brake pressure, the desired brake pressure, p w,des , can be obtained by the equation: 1 p w,des = Tb,des Kb (17) where K b is the lumped gain for the entire brake system. where R g is the gear ratio from the engine to the wheels. K b lumps all the uncertainties in the brake model from the The engine net torque is represented as a function of engine speed, ω e , and throttle angle, α , as follows: brake pressure to the brake torque. The parameter, K b , has been obtained from experimental data. A value of K b = 850 Nm / Pa was used and it provides a good fit to one set of the experimental results. Since the brake line pressure is equal to the brake master cylinder pressure at low frequency actuation, the desired master cylinder pressure, p mc,des , is set to be the brake pressure, i.e.: Tnet = Tnet (ω e ,α ) (13) Typically the engine map is provided by the engine manufacturer as a look up table. The throttle angle, α f , for the desired net engine torque for a given engine speed can be computed from the engine map that shows the 4 u = g −1 ( pd ,des ) + P ⋅ ( pd ,des − pd ) + I ⋅ ∫ ( pd ,des − pd )dt 1 Tb,des (18) Kb Since the master cylinder pressure is directly connected to the vacuum booster, the desired vacuum booster pressure, is given by: A p d ,des = mc p mc,des (19) Ad pmc,des = p w,des = + D ⋅ ( p d ,des − p d ) where u means the applied duty input to the EVB solenoid valve, p d the measured EVB differential pressure, P, I , D the gains and g (⋅) the function representing the relationship between the duty input to the EVB solenoid valve and the steady state values of the EVB differential pressure, i.e.: p d = g (u ) (21) where Amc is the area of the master cylinder and Ad the area of the diaphragm of the vacuum booster. Since the vehicle acceleration is limited in a range [-2, 1] m / s 2 in an ICC, the proportioning valve is not activated so both the rear and the front brakes experience the same brake pressure. In addition, since the brake pressure is controlled at a low frequency range in an ICC, the inertial effect of the connecting rod between the master cylinder and the vacuum booster is not significant. Therefore the simplified equations (16), (17) and (18) represent accurately the brake system in case of an ICC. 4. VEHICLE TESTS Numerical simulations and vehicle tests have been done to evaluate the performance of the proposed control algorithms. Vehicle tests have been conducted using a test vehicle, a 2000cc passenger car equipped with a MMW radar distance sensor, a controller, an EVB brake actuator and a step motor controlled throttle actuator. The differential pressure, p d , of the vacuum booster was controlled by a PWM solenoid valve. The pressure was proportional to the duty ratio input to the solenoid valve. A pressure sensor was installed on the EVB to measure the differential pressure and the measured pressure was used as the feedback in the brake control. The already existing wheel speed sensors, engine RPM sensor, and a Throttle Position Sensor (TPS) have used to estimate the vehicle accelerations and to implement the control laws. Optimal Design : Desired Deceleration udes = −k1 ⋅ ( x p − xcc ) − k2 ⋅ (v p − vcc ) − k3 ⋅ vcc Saturation udes Filter ωn 2 2 s + 2ζωn s + ωn 2 ades 4.1 VEHICLE SPEED CONTROL In the case of the speed control, the controller controls the throttle angle such that the vehicle acceleration tracks the desired acceleration profile, which is designed so that the vehicle speed converges to the set-speed. The desired acceleration, a des , has been obtained using the equations (5) and (6) with u = k 2 ⋅ (v set − vcc ) Desired Wheel Pressure Pw,des = 1 [−rM v ades + Ts − rFL ] Kb Desired Differential Pressure Pd ,des (20) where v set is the set-speed determined by the driver. Fig. 8 shows a comparison of vehicle test and simulation results. Vehicle speeds and throttle angles measured by the TPS are compared. The ICC vehicle’s initial speed has been set to be 80 km / h and the set-speed has been increased to 100 km / h at 2 seconds. As soon as the setspeed is increased, the stepper motor is controlled based on the throttle control algorithm. The maximum vehicle accelerations have been limited to 1 m / s 2 in the simulations and vehicle tests. As indicated in the Fig.8, the simulation results predict the vehicle test results very closely. A = mc Pw,des Ad Duty-ratio : Feedforward + PID control u = g −1 ( Pd , des ) + P × ( Pd , des − Pd ) + I × ∫ ( Pd ,des − Pd )dt + D × ( Pd ,des − Pd ) Fig. 7 Brake control algorithm As indicated in section 2, the EVB brake actuator system dynamics is not negligible and the EVB shows nonlinear characteristics. Therefore, a feed forward plus proportional-integral-derivative (PID) control law is used to control the EVB pressure: 5 A throttle/brake control law for vehicle ICC has been presented. The control law was developed considering the throttle and brake actuators characteristics. The performance of the proposed control laws was investigated via simulations and vehicle tests. The control laws were implemented on a test vehicle. An ICC system used in the tests consists of a MMW radar sensor, a stepper-motorthrottle actuator, an EVB brake actuator, and a controller. The ICC system was implemented on the test vehicle. The vehicle test results show that vehicle speed and distance control performances are satisfactory. The throttle and brakes were gently controlled and the driver is not surprised by the control action. The vehicle accelerations were limited to [1~ -2.5] m / sec 2 in order not to deteriorate ride comfort. The simulation and vehicle test results have shown that the proposed throttle/brake control laws can provide the ICC vehicle with an optimized compromise between safety and comfort. Experiment Simulation Experiment Simulation Fig. 8 Vehicle speed control: Simulations and vehicle tests 4.2 VEHICLE DISTANCE CONTROL Vehicle distance control tests were done using two vehicles: the ICC vehicle and a cut-in-vehicle. Fig. 9 shows the test results. The ICC vehicle’s set-speed was 100 km / h and a vehicle of 80 km / h had appeared in the front of the ICC vehicle at 5.5 seconds. The initial relative distance was approximately 133 m and the headway time of 1.5 seconds has been used in this test. Since the speed of the preceding vehicle is smaller than that of the ICC vehicle, the ICC controller activates the brake control such that the relative distance converges to the headway time distance and the ICC vehicle speed converges to the preceding vehicle’s speed. As illustrated in the Fig. 9, the ICC vehicle speed converges smoothly to the preceding vehicle speed and the relative distance converges to the headway time distance. In this test, the headway time distance was computed using the preceding vehicle speed estimated using the ICC vehicle speed and the measured relative speed. REFERENCES [1] Winner, H., Witte, S., Uhler, W., and Litchtenberg, B., “Adaptive Cruise Control System Aspects and Development Trends,” SAE paper No. 961010, 1996. [2] Muller, R. and Nocker, G., “Intelligent Cruise Control with Fuzzy Logic,” In Intelligent Vehicles ’92 Symposium, pp. 173-178, Detroit, 1992, IEEE Industrial Electronics Society. [3] Choi, S. and Devlin, P., “Throttle and Brake Combined control for Intelligent Vehicle Highway Systems,” SAE paper No. 951897, 1995. [4] Chien, C.C., Ioannou, P., and Lai, M.C., “Entrainment and Vehicle Following Controllers Design for Autonomous Intelligent Vehicles,” Proceedings of the 1994 American Control Conference, pp.6-10, Baltimore, Maryland, June 1994. [5] Ioannou, P., Xu, Z., Eckert, S., Clemons, D., and Sieja, T., “Intelligent Cruiss Control: Theory and Experiment,” Proceedings of the 32nd Conference on Decision and Control, volume 2, pp.1885-1890, San Antonio, Texas, December 1993. [6] Germann, St. and Isermann, R., “Nonlinear Distance and Cruise Control for Passenger Cars,” Proceedings of the 1995 American Control Conference, pp. 3081-3085, Seattle, Washington, June 1995. [7] Elliasson, A., “A Controller for Autonomous Intelligent Cruise Control-a Preliminary Design,” In 3rd International Conference on Vehicle Navigation and Information Systems, pp. 170-175, IEEE, 1992. [8] Hedrick, J.K., “Longitudinal Control and Platooning,” In TOPTEC:Collision Avoidance Systems for Intelligent Vehicles, Washington, D.C., April 1993, SAE. [9] Yi, K., Lee, S.J., Lee, C., and Kong, Y., “Modeling and Control of an Electronic-Vacuum Booster for Vehicle Longitudinal Control,” paper in progress. [10] Goldman, D.E., and von Gierke, H.E. Chapter 44 in Shock and Vibration Handbook, Cyril M. Harris(ed), McGraw-Hill Book Company, 3rd edition, 1988. ICC vehicle speed Preceding vehicle speed Relative distance Headwaytime distance Fig. 9 Vehicle tests results 5. CONCLUSIONS 6