Integrated DYC and ASR Mixed Control of Distributed Drive Electric Vehicle Using Co-Simulation Methodology

The improvement of handling and stability performance of distributed drive electric vehicle (DDEV) is analyzed, visualized and designed by proposing and deploying the mixed control strategies in this paper including Direct Yaw Control (DYC), Anti-slip Regulation (ASR) and dual-mode switching control. The practicability and real time visualization of driving efficiency and timeliness of DDEV is achieved to reduce the margin of error for the desired torque value by employing the DYC strategy which uses fuzzy PID algorithm. Furthermore the ASR strategy which adopts the optimal slip rate algorithm to determine the requirement of desired torque value based on the different road conditions is used to reduce slip phenomenon effectively and to maintain handling control of DDEV. In response to different scenes especially conflict and coexistence between DYC and ASR, the dual-mode switching control strategy is applied to find more suitable slip rate range by using the root mean square error method (REME). Finally, co-simulation platform of ADAMS/Car and MATLAB/Simulink is built to simulate the mixed control strategies by integrating dual-mode switching control, DYC and ASR. The simulation results show that this strategy has a more significant control effect needed to meet the requirements of normal vehicle handling and stability. The mixed control strategy is adopted and downloaded into the electronic control unit of our student type formula vehicle called Flash V6 which was designed and developed by a team of students, the ZUST ATTACKER Team.


Introduction
Distributed drive electric vehicle (DDEV) is an important development and design direction of future electric vehicles. Their main features are that, the drive motor is embedded in the wheel, the transmission chain is shorter, speed response is faster and the torque control accuracy is high 1 . The wheels equipped with the drive motors can not only be independently controlled but also the wheel torque can be arbitrarily distributed to achieve optimum and better control effect. Based on these characteristics, a torque optimization control algorithm was designed 2 and combined with the quadratic programming method to design torque average distribution algorithm 3 . Zhu et al. 4 proposed a particle swarm optimization and nonlinear predictive control algorithm through torque optimization and distribution to improve the vehicle stability while Liu et al. 5 used the expected additional torque obtained by a linear quadratic regulator (LQR).
The more common schemes in torque control include uses the yaw rate as the feedback object which is obtained by the optimal allocation strategy 6 and combined with the slip rate normally used to construct ABS which prevent vehicle instability 7 . The yaw rate and the side slip angle are not only used as parameters to observe the vehicle response 8−9 but also combined with sliding mode controller. These strategies based on multiple sliding mode controls 10−12 were used to design a torque distribution system needed for torque redistribution while ensuring the stability of vehicle braking.
Meanwhile, the slip rate can also be used as a reference value to study the ASR strategy. Based on the independent driving electric vehicles models, the conventional method of slip rate calculations were used to improve the vehicle longitudinal driving stability 13 . Through fuzzy PID controller, the motor torque is adjusted to obtain the best slip rate control 14−16 which shortens the driving acceleration distance and improves the steering stability 17 . Based on fuzzy PID control theory 18 , the integrated control strategy of ASR and DYC with a feed forward-feedback method were designed as the active front wheel steering controller (AFS) and DYC which are used as new controllers to obtain a target drive torque that could effectively improve vehicle stability 19 . A dynamic DYC combined with feed forward and feedback control system was proposed to improve the stability of the vehicle 20 . The path tracking and MPC controller were introduced in DYC to achieve the purpose of correcting the driving state of the vehicle 21 . Multi-mode control strategies were designed to improve the longitudinal stability of the vehicle 22 . A multi-mode switching control strategy was designed based on the longitudinal acceleration characteristics of the vehicle. The effectiveness and practicability of the control strategies were finally verified using a real vehicle.
Based on the above excellent research works, a mixed control strategy can be integrated into the dual-mode switching control of DYC algorithm with fuzzy PID controller and ASR algorithm with optimal slip rate PID controller. The strategy can be established based on the DDEV under normal motion conditions. A co-simulation platform can be set up to simulate the vehicle model with single lane change and step steer tests using ADAMS®/Car and MATLAB®/Simulink.
The results from the mixed control strategy integration consisting of DYC and ASR can effectively prevent DDEV from entering the dangerous limit driving state when turning and driving. The outcome can greatly improves the driving safety of the vehicle.

Vehicle modeling
The electric vehicle model with distributed driven wheels was constructed using ADAMS/Car software while the motor model and control strategy with the MATLAB/Simulink software. Setting the relevant parameters at the communication interface between ADAMS/Car and MATLAB/Simulink, the real-time data interaction channel of ADAMS/Car and MATLAB/Simulink was established, and the two parts of the model were jointly simulated. The overall architecture is illustrated as shown in in Fig.1.   In order to design the control model, the 2-DOF linear vehicle model 9 is used as the reference model:

Establishment of ADAMS/Car model
According to the transformation formula: Where v is the longitudinal speed; y v is the velocity of the center of mass; f  is the front wheel angle.
According to formula (2), f  with v are used as input variables; the desired yaw rate d  and side slip angle d  are obtained by establishing 2-DOF linear vehicle model in Simulink.

Modeling of simulink motor
Emrax 188 motor is used in the DDEV, because of the reliability of permanent magnet synchronous motors. Their mechanical characteristics include simple structure, small volume, less cumbersome, rapid response, high power density and efficiency. According to the relevant parameters in Table 2, a simplified motor model is established as the torque demand command actuator in the vehicle control strategy. By adjusting the current in the motor model, the motor can output smooth torque to the driving wheel.  In order to make the motor torque command response as close to the real motor characteristics as possible, the motor driver is regarded as the first-order inertial link when building the simplified model of the motor as shown in the formula: Where w T is the actual torque of the motor, τ is the time constant obtained from the motor characteristic parameters, m T is the external characteristic torque of the motor.
In order to describe the characteristics of the motor, the transfer function equation of the motor is established and combined with the motor parameters. The model diagram is shown in Fig. 3. Where dl T is target torque of left wheel, dr T represents right wheel target torque, T e is the torque difference, x U is the motor terminal voltage value, U e is the voltage difference, I is the armature current, e K is the electromagnetic torque constant, w K is the induced electromotive force constant, w U is the induced electromotive force value, ml T is the left wheel motor output torque, mr T is the right wheel motor output torque, rr w actual angular velocity of right wheel, rl w actual angular velocity of left wheel, rl T is the actual torque of left wheel, rr T is the actual torque of right wheel.

Vehicle model
As shown in Fig  Where K is the pedal opening valve, c  is the steering wheel angle, and  is the adhesion coefficient.

Control strategy design
In this section, the DYC and ASR are explained, and the effectiveness of the control strategies is verified by simulation.

DYC module
The DYC module is mainly used to calculate the additional yaw torque produced by the vehicle around the center of mass when turning, adjust the yaw motion state of the vehicle, and then participate in the driving torque distribution module of the vehicle.
Based  Defined variable: Where 1  and 2  are the weight coefficients, which are obtained by the established fuzzy rule controller in Table 3.  Table 3. Fuzzy controller rules According to formula (4) and Table 3, the 1 M  is obtained: Where 1 k is the fuzzy function known as Mamdani fuzzy rule controller. Neglecting the road interference in the process of vehicle driving, the 1 M  obtained needs to be converted to the torque value acting on the driving wheel and the following formula is used : Where F  is the corrected driving force. It is noted that, the relationship between driving torque and driving force of wheel is given as follows: Where r is the wheel rolling radius, and T  is the torque the motor need to adjust when maintaining the body posture under the current driving conditions.
The redistributed torque values are obtained by combining equations (6) and (7): It can be seen , in the first second of simulation, the actual value of yaw rate of vehicles without DYC is significantly higher than the expected value and seriously deviates in the later stage which may lead to over steering and other dangerous situations while the vehicle with DYC has better yaw rate tracking performance. The difference is within 0.1rad and the control response is fast making the control effect obvious. And the wheels of the vehicle can be well attached to the road surface after the slip rate of the vehicle is controlled.

ASR module
The purpose of ASR module is to obtain the additional yaw torque value that needs to be applied to prevent the wheel from slipping during the driving process and use the road surface information to obtain the force required to control the normal steering of the vehicle to improve the vehicle handling stability and driving safety. The direct driving force of the vehicle comes from the friction reaction between the driving wheel and the ground and its limit value is also called adhesion. The mathematical formula of adhesion is described as follows: F is the ultimate adhesion between the wheel and the ground; z F is the vertical load of the wheel. In order to prevent the relative slip movement between the wheel and the ground, the vector sum of the driving torque provided by the motor for the wheel and the lateral force on the tire should be less than the limit adhesion force between the wheel and the ground.
Where μ F is the wheel ground utilization force; y F is the wheel lateral force. When the driving force on the wheel exceeds the adhesion limit, the wheel will be in slip state. When the vehicle is running normally, it is necessary to consider the ability of the wheel of the DDEV to resist external interference. The tire is required to have sufficient reserve adhesion while meeting the vehicle longitudinal driving and yaw torque requirements. The optimal slip rate value opt  is obtained based on the adhesion coefficient curve of slip rate 11 as shown in Fig.8 which is located and selected at point A because the longitudinal adhesion coefficient of the tire at this point is the maximum.  as shown in Fig.9.
Where zrl F is the vertical load of the left rear wheel; zrr F is the vertical load of the right rear wheel; g is the acceleration of gravity; x a is the longitudinal acceleration; y a is the lateral acceleration. When the vehicle is steering, the load transfer will occur and the limit adhesion of the inner wheel will decrease while that of the outer wheel will increase. Therefore, in order to prevent the inner wheel from slipping, the driving torque of the inner wheel should be limited.
If the above formula is satisfied, then:  T T  T T  M  T ,  + T  T  T  T  T  M  T ,  -T  T  T  T  T If not, then: The co-simulation model of ADAMS/Car and MATLAB/Simulink is built based on the ASR as shown in Fig.10. It can be seen, the vehicle enters the steering state in the first second and its yaw rate changed greatly. After the ASR control, the vehicle can follow the expected yaw rate very well, and the difference of d  and  is controlled at 0.5rad/s, making the control effect good; the yaw rate of the uncontrolled vehicle changes relatively more, which may appear in the later stage. And the slip rate of the driving wheel of the vehicle under the control of ASR changes very little and tends to be stable quickly, which prevents excessive wheel slip and ensures the dynamic performance and stability of the vehicle.

Design of dual-mode switching control
In order to better realize the demand of handling and stability control in the driving process of DDEV, a dual-mode switching control strategy based on DYC and ASR is proposed in this section.

Mode switching design
Single control cannot adapt to all working conditions of vehicle. According to the evaluation index of vehicle handling and stability established by academician Guo Konghui 9 and the characteristics of the two control modes, the slip rate is taken as the reference characteristic quantity for switching between the two control modes.
We can set the test and simulation conditions as shown in Fig.12 to confirm the characteristic slip rate threshold when the two control modes are switched.
The simulation parameters in Table 4 are set for DYC.

9/14
Test schemes Parameters Value and units Step steer   REME  as ordinate, r w as the sum of curves in abscissa and  as ordinate while r w is the abscissa curve. In Fig.13 the curve under the step steer test condition is analyzed when  is above 0.15,   REME  is small and the DYC effect at this stage is better. The curve analysis under single lane changes the test condition when  is above 0.15 and the corresponding   REME  curve is smaller which shows that the DYC effect is significant in this stage. Under the constant radius cornering test condition, the   REME  curve is obviously small with wr and when  is between 0.15 and 0.4, the   REME  curve suddenly changes from a larger value to a smaller value. The analysis shows that the DYC has more advantages when  is between 0.15 and 0.4.  Table 5    In Fig.14, it can be seen from the curve under the step steer test condition that  increases with r w when  is higher than 0.4. The control effect is better when  is above 0.4. Analyzing the curve of single lane change test, it can be concluded that the corresponding   REME  curve is smaller than other stages when λ is above 0.4 which indicates that the ASR effect is more significant in this stage. Under the constant radius cornering test condition, the corresponding   REME  curve is obviously smaller than other stages with wr When  is above 0.4 the   REME  curve changes greatly at this stage. It can be seen that the effect of ASR has been shown. In conclusion, when λ is between 0.4 and 1 the   REME  curve tends to decline or is in a low range which indicates that the ASR effect is superior.
Combining with the simulation curve analysis in Fig.13 and Fig.14 .When  is between [0.15 0.4], the effect of DYC strategy is better. When  is between [0. 4 1], the effect of ASR strategy is better. Therefore, the threshold values of mode switching are 0, 0.15, 0.4 and 1.

Mode switching judgement
In order to prevent the frequent change of  which is caused by complex road surface that eventually lead to misjudgment of vehicle ASR and DYC system.  is used as the evaluation reference when 0 15 0 4 . .

 
, DYC is switched to apply torque to reduce  . When 0 4 1 .  , ASR is switched to control opt  . Once the corresponding control mode is switched in the whole process, it will keep running until it is judged that  is less than 0.15.
Mode switching means that after switching the control system, the torque command of the drive motor is output by the corresponding control module. When the control mode is exited, the torque command of the drive motor is not directly output through the control module. The switching process is designed as 01 y , y  and 2 y  , which is shown in Fig.15.

Mode switching simulation test
The purpose of simulation test is to verify the effectiveness of the two control modes. The single lane change test simulation means that the driving vehicle completes the lane changing action during the driving process which is of great significance to the vehicle stability. According to the relevant standard 10 , v is designed to be 50km/h,  is 0.4, and c  within 1s was set to 2  rad to the left and returns to normal at 7s for 8s run. The curve is shown in Fig.16.

12/14
According to the yaw rate response curve in Fig.17, it can be seen that when the vehicle is in 1s due to c  input, the vehicle may be turning or overtaking and its  changes greatly. After about 1s adjustment, the controlled  quickly reaches stability and the steady state value is 0.35rad/s making the control effect very good. After the steering action of the uncontrolled vehicle is finished in a second, the  curve is still in the growth trend and the whole vehicle has the tendency to side slip which is not conducive for stable driving of the vehicle. It can be seen from the curve in Fig.17 that the uncontrolled and controlled changes have taken place near 1.5s and the controlled  curve change is controlled within 0.005rad making the stability of the vehicle better. The change of  curve without control to -0.025rad is 5s of the maximum value of  with control. The x a curve also shows that the controlled vehicle value is 1/2 of the uncontrolled value, indicating that the vehicle has better handling stability after control.

Driver in the loop simulation test
In order to further verify the real effect of the DYC strategy, it is necessary to observe the control effect via real vehicle test. However, due to the difficulty of real vehicle verification, the existing driver simulator Logitech G27 equipment is used to verify the control strategy to continuously optimize the strategy.
In Fig.18, the serial communication port between G27 and MATLAB/Simulink is established first and the joystick output module in MATLAB/Simulink is used to adjust the c  of Logitech G27 Pedal (switch) opening K and other signals are collected. For the vehicle model in this paper, we mainly collected the c  of G27 and K signals. In particular, due to the c  of Logitech G27, the total travel range is 5π rad in the vehicle's mathematical model in MATLAB/Simulink. The travel range of c  is 6πrad making it is necessary to collect the c  of joystick output signal when processed by gain to satisfy the c  of vehicle mathematical model. Based on G27, the c  is designed to test conditions of turning  rad to the left and then turning  rad to the right as shown in the Fig.19.

13/14
It can be seen from the curves of torque variation of the rear wheels in Fig.19 that the torque values on the uncontrolled wheels rapidly decrease and then increase at about 4s, 8s and 11s. The change in trend of θc curve shows that when the vehicle is steering, the driving torque of the wheel fluctuates violently. After controlling the torque, fluctuations on the rear wheels decrease and are made relatively gentle which indicates that the DYC strategy and the design redistributes the torque to the wheels to improves the handling stability of the vehicle under the premise of meeting the normal driving demand of the vehicle.

Conclusion
The control effects of DYC and ASR on handling and stability performance are presented based on the DDEV model in this study . Firstly, the subsystem model of each part of the vehicle is established using the ADAMS/Car software while the motor model is established based on MATLAB/Simulink software. The DYC model and the ASR model are designed, built and used in the established co-simulation platform.
The analysis from the simulation is carried out by referencing the two models which led to the design of dual-mode switching control strategy. The results of simulations from the curves guaranteed a satisfactory control effect. Finally the results from the driver in the loop simulation of the DYC strategy, real vehicle test simulation and the curve analysis show that the DYC strategy has superior control effect on vehicle handling stability which provides a robust methodology for the development of vehicle control strategy in future. Figure 1 Schematic diagram of vehicle and control models architecture.