APP下载

A novel CMAC-PID combined control strategy for the Lane Keeping Assist System in automobiles

2020-07-24YanningWANG

机床与液压 2020年12期

Yan-ning WANG

(Guangdong University of Science and Technology, Dongguan 523083, China)

Abstract: In order to effectively improve the vehicle driving safety, Lane Keeping Assist System (LKAS) arises as one of the most important components of the vehicle safety assist driving system. Considering the human-machine conflict in the vehicle control weight allocation of the auxiliary control system, in this paper, the decision-making of vehicle status judgment and the auxiliary control strategy are studied. Firstly, the vehicle-road reference model is established. Then, the monitoring strategy of vehicle status is proposed based on the preview theory. Furthermore, the decision-making rule of vehicle deviation is designed according to the torque applied on the steering wheel. Furthermore, the self-learning function of the cerebellar model articulation controller (CMAC) neural network is applied to adjust PID parameters in real time, which aims to realize the active control of vehicle steering system. Finally, the simulation is carried out by CarSim and Matlab. Simulation results illustrate that, compared with the traditional PID control, the CMAC-PID-combined control algorithm can correctly recognize the state of driver operation and vehicle status. Once the vehicle deviates, the proposed algorithm can effectively generate early warning and output active deviation correction control. Simulation results demonstrate that the proposed control algorithm achieves advantages of high accuracy and fast response speed, thus ensure the driving safety and reduce the human-computer conflict.

Key words: Lane Keeping Assist System (LKAS), CarSim, CMAC neural networks, Vehicle status

Lane Keeping Assist System (LKAS) is one of the most important components of the vehicle assist driving system. It is of great significance to improve vehicle lateral stability, reduce driver’s operation burden, and avoid traffic accidents. Due to the driver’s negligence or change of driving conditions, there is a risk of lane deviation during the driving process of the vehicle [1-3]. LKAS can be adaptively modified to cope with the driver’s operation, in which way to improve the cooperation between the auxiliary control system and the driver. And an alarm arises once the vehicle unconsciously deviated from the original lane while no correction activities were taken. In such a way, LKAS is designed to remind the driver of the current risk of vehicle deviation. Meanwhile, the vehicle steering is under active control of the LKAS controller according to the pre-established control strategy, which intends to provide more reaction time for the driver. In this manner, the vehicle can be driven back to the centerline of the lane. Therefore, the collision accidents caused by the vehicle deviation are greatly reduced, and also the comfort and driving safety of the vehicle are further improved [4].

At present, the bottleneck of LKAS remains on the control weight allocation problem between the driver and the steering system, which is sensitive to cause man-machine conflict and affects the driver’s driving experience [5-6]. On the other hand, as the vehicle steering system is a linear, multivariable and complex one, the control parameters of the system need to be adjusted online in real time. To this end, we propose a monitoring method of vehicle running status based on Preview theory, design a lane departure decision-making model according to the torque applied on the steering wheel, and apply the CMAC self-learning function to adjust PID parameters in real-time. Thus, the proposed CMAC-PID combined control strategy can realize precise control of vehicle lateral position, reduce the possibility of conflict between the auxiliary control system and driver, and also improve the effectiveness of LKAS in restraining the unconscious vehicle deviation.

1 Vehicle-road reference model

1.1 Vehicle dynamics model

In order to evaluate the performance of LKAS, the lateral motion of the vehicle is taken into consideration. In this paper, a two degree of freedom vehicle dynamics model is introduced. Practically, only the lateral motion along theY-axis of the vehicle and the yaw motion around theZ-axis of the vehicle are both concerned in the LKAS applications. The state equation of the vehicle is shown as follows [7].

(1)

whereα1andα2are the side deflection angles of the front wheels and rear wheels, whileaandbindicate the wheelbase of the corresponding wheels, respectively.δdis the front wheel angle,vxandvyrepresent the lateral speed and longitudinal speed of the vehicle, respectively.θis defined as the vehicle heading angle.

(2)

whereβis the eccentric angle of the center of mass,mis the mass of the vehicle,Izis the moment of inertia of the vehicle around theZ-axis,C1andC2are the side deflection stiffnesses of front tires and rear tires, respectively.

1.2 Path preview model

“Preview” is the main idea of vehicle path tracking, which can monitor the driving status and path of the vehicle on the road in real time. In this paper, the “Preview” method is adopted to analyze the driving status of the vehicle, and the principle of the method is shown in Fig.1.

Fig.1 Principle of preview method

According to Fig.1, the geometric position of the vehicle with respect to the lane is obtained as follows [8].

(3)

whereθ′ is the yaw angle of the vehicle;Sis the “Preview” distance;ysis the lateral position deviation, andρis the road curvature.

2 Analysis of LKAS framework

The lane departure warning and lane keeping active control functions are immediately required in LKAS. And the driving state of the vehicle is determined by the status monitoring module. The LKAS framework is shown in Fig.2.

Fig.2 Functional framework of LKAS

The functional framework of LKAS in Fig.2 is further detailed as follows.

The environment information is obtained by the environmental information awareness module, which mainly includes the position, direction and other state information of the vehicle with respect to the lane centerline. The vehicle speed, yaw rate, steering wheel torque, turn on signal status, and other information are obtained through the vehicle status monitoring module.

Among the lane departure warning module, driver status recognition module, and vehicle control weight allocation module, the highest priority of vehicle control should be allocated to the driver during this process. Once the vehicle departure condition is met, a warning signal will be sent to remind the driver in time. Furthermore, if the active control is activated, then the LKAS system can drive the vehicle in time back to the proper lane.

According to the proposed lane keeping control strategy, the motors output lane keeping control and correct the driving attitude of the vehicle, so as to ensure that the vehicle always is forced to the lane centerline. LKAS can not only improve driving safety but also gain the benefit of human-computer sharing.

3 Driver awareness recognition

In LKAS, the driver’s active operation should be allocated with the highest priority of vehicle control. At this point, it is of great importance to correctly recognize the driver’s operation status, which leads to a direct influence on human-computer coordination and LKAS control performance. Thus, the driver’s status recognition module should be loaded before the lane departure auxiliary control. The principles of driver status recognition and vehicle control weight allocation are shown in Fig.3.

Fig.3 Driver status recognition and vehicle control weight allocation principle

The principles of driver awareness recognition are as follows.

In the driver status recognition system, the driver’s control status is recognized according to whether the turn signal is on or not. Once the turn signal is turned on, it is confirmed that the driver is actively manipulating the vehicle. At this time, in order to avoid the interference between LKAS and the driver, the vehicle control weight is allocated to the driver. The system changes the driving status of the vehicle according to the intention of the driver, without any warning and control.

There are two cases to be discussed when the turn signal is turned off. On the one hand, the driver takes the initiative to operate the vehicle while forgetting to turn on the turn signal or failing to turn on the turn signal in an emergency. On the other hand, when the driver does not actively control the vehicle, the vehicle deviates from the lane. In the first case, a certain amount of steering torque is usually applied to the steering wheel. While there is usually no torque applied to the steering wheel in the second case. At this point, the driver’s operation status can be determined by measuring the torque signalTse. Given a torque thresholdT, it is means that the driver is operating the vehicle ifTse>T, and thus LKAS does not perform the assist control. ForTse

4 Lane keeping assistant control strategy based on CMAC-PID

4.1 PID control algorithm

PID control is a kind of control method which combines proportion, integral and differential modules. It is widely used in various industrial fields. Recently, PID is also used in the lateral deviation control of the vehicle, and the control law is formulated as follows.

(4)

whereKp,KiandKdare the proportional coefficient, integral time constant and differential time constant, respectively.up(t) is the output variable, representing the target steering torque. Ande(t) is the tracking error between the expected output and the actual output.

Despite the good control performance of PID, it does not gain the self-adaptive ability. In another words, PID control cannot adjust the parameters in real time when the external environment varies, thus the control accuracy cannot be kept sustainably. Therefore, PID is only applicable for the system with fixed model and parameters, as well as the occasions with rarely low control requirements.

4.2 CMAC neural networks

Cerebellar model articulation controller (CMAC) is a neural network model describing the function and principle of the cerebellar in limb motion control. To the best knowledge of authors’, it can be used to learn any nonlinear mapping, and also gain the advantages of fast convergence and robust adaptive ability. Therefore, CMAC is widely applied in real-time control systems [9-10]. The structure of CMAC neural networks is shown in Fig.4.

Fig.4 The structure of CMAC neural networks

4.3 CMAC-PID combined control strategy

Relying on the self-learning function of CMAC, we take the advantages of both CMAC and PID control methods to design the steering system of the vehicle. The proposed CMAC-PID combined control strategy is shown in Fig.5.

Fig.5 CMAC-PID combined control diagram

In Fig.5, the CMAC-PID combined control diagram is a concurrent structure with both feedforward and feedback modules. The output of PID is used to train the CMAC neural networks, and the feedback control module of PID is applied when there is any tracking error between the output of the steering system and the expected control performance. Meanwhile, the adjust control signal is transferred to the CMAC controller for neural networks training. The corresponding parameters in the steering system are jointly determined by CMAC and PID. In this way, the tracking error between the expected signal and the actual signal is designed as the input of PID, then the output signal and the reference command of PID are set as the input of CMAC. Furthermore, the output of PID and the output of CAMC are combined to be the control signal of the steering system. Consequently, the output of the system can be expressed as follows.

(5)

u(t)=uc(t)+up(t)

(6)

whereαiis the selected vector,uc(t) is the output of CMAC control,cis the generalization coefficient of CMAC.

In the control process, with the help of CMAC self-learning function, the network weights are continuously updated to realize the real-time online adjustment of PID controller parameters, so as to meet the requirements under different working conditions. The weights update law is given as follows.

(7)

(8)

ω(k)=ω(k-1)+Δω(k)+α[ω(k)-ω(k-1)]

(9)

whereηis the online learning rate,ω(k) is the weight, Δω(k) is the weight change,αis the coefficient of inertia.

5 Simulation analysis

In order to verify the effectiveness of the proposed model and control strategy, the joint simulation is carried out by CarSim/MATLAB platforms under multi-conditions. We build the whole vehicle model in CarSim, while EPS model, driver model and control algorithm module in MATLAB. The simulation parameters for the vehicle are shown in Table 1.

Table 1 Simulation parameters for the vehicle

Step signal is chosen as the input for the steering control system, and the results are shown in Fig.6.

Fig.6 Simulation results of step signals

The simulation result demonstrates that, compared with the conventional PID control, the proposed CMAC-PID combined control strategy has smaller overshoot and faster response speed, which can fully satisfy the control requirements of LKAS.

The object of the following simulation is to track and monitor the front tire deflection angle that is transformed via the lateral position and the offset of the vehicle from the centerline of the lane. In such a way, we can adjust the steering moment of the steering wheel via the proposed controller such that the LKAS performance is fulfilled.

The simulation experiments are carried out at the speed of 60km/h and 80km/h, respectively. Assuming that the vehicle is given a right sudden turn (i.e. step signal) to the steering wheel without turning on the steering signal, which simulates the vehicle’s right departure from the lane. Meanwhile, the torque acting on the steering wheel is less than the set threshold. The simulation results under the proposed control strategy adopted in LKAS are shown in Fig.7~10.

Fig.7 Decision state of LKAS

Fig.8 Vehicle lateral deviation

Fig.9 Auxiliary torque of the motor

Fig.10 Steering wheel angle

As demonstrated in the simulation results, the proposed control strategy in LKAS monitors the driving status of the vehicle in real-time. Once the decision signal indicates that the vehicle has deviated without the driver’s consciousness, an active control decision is made and then a control command is generated from the control system. It is shown that, the greater the vehicle speed, the earlier the decision is made. Subsequently, by giving the control priority to LKAS, the ESP system begins to work such that the vehicle can be driven back to the correct route in time. With the increase of vehicle speed, the lateral deviation of the vehicle, the auxiliary torque produced by the motor and the steering wheel angle of the vehicle also increases. In the case of 60 km/h, the maximum lateral deviation is 580 mm. Also, for 80 km/h, the maximum lateral deviation is 720 mm.

As discussed above, the proposed lane departure decision strategy and CMAC-PID combined control strategy are feasible and effective, with high control accuracy and fast response speed, which can avoid human-computer conflict and ensure driving safety.

6 Conclusions

In order to further improve the safety of vehicle driving, this paper studies the decision strategy and control strategy for the lane keeping assistant system of vehicles. The following conclusions can be drawn.

Firstly, the vehicle-road reference model is established. Meanwhile, with the help of the ‘Preview’ theory, the monitoring method of vehicle status and the decision-making strategy of vehicle deviation are proposed.

Secondly, by combining the self-learning function of CMAC neural networks and PID control, real-time online update law for PID parameters is carried out. And also combined simulation experiments are carried out by CarSim and MATLAB platforms.

Furthermore, the research results illustrate that, compared with the conventional PID control, the CMAC-PID combined control strategy achieves the advantages of small overshoot and fast response speed. Moreover, once the vehicle deviates, the proposed control strategy can effectively generate an early warning and active deviation correction control, which can reduce the human-machine conflict and ensure driving safety.


登录APP查看全文