JP7312707B2 - suspension controller - Google Patents

Description

The present disclosure relates to a suspension control device mounted on a vehicle such as an automobile.

Conventional suspension control detects or estimates a vehicle state and performs feedback control according to the detected state (see Patent Document 1). The feedback control uses, for example, the Skyhook control law or BLQ (Bi-linear Optimal Control). Also, in recent years, model predictive control and direct optimal control have been proposed (see Non-Patent Document 1).

JP 2014-69759 A M.Witters , J.Swevers, “A case study towards the application of NMPC to a continuously variable semi-active suspension,” : Proceedings of ISMA2010 including USD2010, 2010, pp. 479-494.

By the way, the suspension control device disclosed in Patent Document 1 uses controls such as the skyhook control law and BLQ. However, since these controls are based on linear/bilinear systems, they are not necessarily optimal controls.

On the other hand, model predictive control is said to have high control performance. However, model predictive control requires complicated models and controllers for control design. Further, direct optimization control requires optimization for each step, and cannot be realized by an ECU (Electronic Control Unit).

SUMMARY OF THE INVENTION An object of an embodiment of the present invention is to provide a suspension control system capable of optimal control according to vehicle conditions.

A suspension control device according to one embodiment of the present invention is provided between a vehicle body and wheels of a vehicle and includes a force generation mechanism capable of adjusting a force between the vehicle body and the wheels; a vehicle state quantity acquisition unit for detecting or estimating a vehicle state quantity;The AI learning unit of the controller learns the command value obtained by the optimization method so as to minimize the evaluation function in advance and the acquisition result of the vehicle state quantity acquisition unit.It is characterized by

A suspension control device according to one embodiment of the present invention is provided between a vehicle body and wheels of a vehicle and includes a force generation mechanism capable of adjusting a force between the vehicle body and the wheels; a vehicle state quantity acquisition unit for detecting or estimating a vehicle state quantity; a road surface profile acquisition unit for detecting or estimating a road surface profile; An AI learning unit that learns a command value for the force generation mechanism based on the acquisition result of the acquisition unit.The AI learning unit of the controller learns command values obtained by an optimization method that minimizes a predetermined evaluation function, and acquisition results of the vehicle state quantity acquiring unit and the road surface profile acquiring unit.It is characterized by

According to one embodiment of the present invention, optimum control according to the vehicle state becomes possible.

1 is a diagram schematically showing a suspension control device according to a first embodiment; FIG. FIG. 4 is an explanatory diagram showing a procedure for learning a DNN of a controller; FIG. 4 is an explanatory diagram showing ride comfort performance of the first embodiment and a comparative example; FIG. 4 is a characteristic diagram showing temporal changes in road surface displacement, sprung acceleration, sprung jerk, piston speed, current value, and damping force in the first embodiment and the comparative example. FIG. 5 is a diagram schematically showing a suspension control device according to a second embodiment; FIG. FIG. 7 is a diagram schematically showing a suspension control device according to a third embodiment; FIG. FIG. 12 is an explanatory diagram showing an analysis model used for DNN learning according to the fourth embodiment ;

A suspension control system according to an embodiment of the present invention will now be described in detail with reference to the accompanying drawings, taking as an example a case where it is applied to a four-wheeled vehicle.

In FIG. 1, for example, left and right front wheels and left and right rear wheels (hereinafter collectively referred to as wheels 2) are provided on the underside of a vehicle body 1 that constitutes the body of the vehicle. These wheels 2 are configured including tires 3 . The tire 3 acts as a spring that absorbs fine irregularities on the road surface.

The suspension device 4 is interposed between the vehicle body 1 and the wheels 2 . The suspension device 4 includes a suspension spring 5 (hereinafter referred to as a spring 5) and a damping force adjustable shock absorber (hereinafter referred to as a variable damper 6) provided between the vehicle body 1 and the wheel 2 in parallel with the spring 5. Note that FIG. 1 schematically illustrates a case where one set of suspension devices 4 is provided between the vehicle body 1 and the wheels 2 . In the case of a four-wheeled vehicle, a total of four sets of suspension devices 4 are provided individually and independently between the four wheels 2 and the vehicle body 1 .

Here, the variable damper 6 of the suspension device 4 is a force generation mechanism that generates an adjustable force between the vehicle body 1 side and the wheel 2 side. The variable damper 6 is configured using a damping force adjustable hydraulic shock absorber. The variable damper 6 is provided with a variable damping force actuator 7 comprising a damping force adjusting valve or the like in order to continuously adjust the characteristics of the generated damping force (that is, the damping force characteristics) from hard characteristics (hard characteristics) to soft characteristics (soft characteristics). Note that the variable damping force actuator 7 does not necessarily have to be configured to continuously adjust the damping force characteristics, and may be capable of adjusting the damping force in a plurality of steps, for example, two or more steps. Also, the variable damper 6 may be of the pressure control type or the flow rate control type.

The sprung acceleration sensor 8 detects the vertical acceleration of the vehicle body 1 (sprung). The sprung acceleration sensor 8 is provided at an arbitrary position on the vehicle body 1 . The sprung acceleration sensor 8 is attached to the vehicle body 1 at a position near the variable damper 6, for example. The sprung acceleration sensor 8 detects vertical vibration acceleration on the vehicle body 1 side, which is the so-called sprung side, and outputs the detection signal to an electronic control unit 11 (hereinafter referred to as an ECU 11).

A vehicle height sensor 9 detects the height of the vehicle body 1 . A plurality of (for example, four) vehicle height sensors 9 are provided corresponding to the respective wheels 2 on the vehicle body 1 side, which is the upper side of the spring, for example. That is, each vehicle height sensor 9 detects the relative position (height position) of the vehicle body 1 with respect to each wheel 2 and outputs the detection signal to the ECU 11 . The vehicle height sensor 9 and the sprung acceleration sensor 8 constitute a vehicle state quantity acquisition unit that detects the state quantity of the vehicle. The state quantity of the vehicle is not limited to the vertical acceleration of the vehicle body 1 and the height of the vehicle body 1 . The state quantity of the vehicle may include, for example, the relative velocity obtained by differentiating the height (vehicle height) of the vehicle body 1, the vertical velocity obtained by integrating the vertical acceleration of the vehicle body 1, and the like. In this case, in addition to the vehicle height sensor 9 and the sprung acceleration sensor 8, the vehicle state quantity acquisition unit also has a differentiator that differentiates the vehicle height and an integrator that integrates the vertical acceleration.

The road surface measurement sensor 10 constitutes a road surface profile acquisition unit that detects the profile of the road surface. The road surface measurement sensor 10 is composed of, for example, a plurality of millimeter wave radars. The road surface measurement sensor 10 measures and detects the road surface condition in front of the vehicle (specifically, including the distance and angle to the road surface to be detected, and the screen position and distance). The road surface measurement sensor 10 outputs a road profile based on the detected values of the road surface.

The road surface measurement sensor 10 may be, for example, a combination of a millimeter wave radar and a monaural camera, or may be configured by a stereo camera including a pair of left and right imaging elements (digital camera, etc.) as described in Japanese Patent Application Laid-Open No. 2011-138244. The road surface measurement sensor 10 may be configured by an ultrasonic distance sensor or the like.

The ECU 11 is mounted on the vehicle body 1 side of the vehicle as a control device that performs behavior control including attitude control of the vehicle. The ECU 11 is configured using, for example, a microcomputer. The ECU 11 has a memory 11A capable of storing data. The ECU 11 has a controller 12 .

The input side of the ECU 11 is connected to the sprung acceleration sensor 8 , the vehicle height sensor 9 and the road surface measurement sensor 10 , and the output side is connected to the damping force variable actuator 7 of the variable damper 6 . The ECU 11 outputs the road surface profile and the vehicle state quantity to the controller 12 based on the vertical vibration acceleration detected by the sprung acceleration sensor 8, the vehicle height detected by the vehicle height sensor 9, and the road surface detected by the road surface measurement sensor 10. The controller 12 obtains the force to be generated by the variable damper 6 (force generating mechanism) of the suspension device 4 based on the road surface profile and the vehicle state quantity, and outputs the command signal to the damping force variable actuator 7 of the suspension device 4 .

The ECU 11 stores the data of the vehicle state quantity and the road surface input in the memory 11A for several seconds while the vehicle travels about 10 to 20 meters. As a result, the ECU 11 generates time-series data (road surface profile) of the road surface input and time-series data of the vehicle state quantity when the vehicle travels a predetermined distance. The controller 12 controls to adjust the damping force to be generated by the variable damper 6 based on the road surface profile and the time-series data of the vehicle state quantity.

The controller 12 has a trained DNN 13 (deep neural network). The DNN 13 is an AI learning unit, and is composed of, for example, a multi-layered neural network of four or more layers. Each layer has a plurality of neurons, and neurons of two adjacent layers are connected by a weighting factor. Weighting factors are set by prior learning. The controller 12 acquires time-series data of the road surface input (road surface profile) and time-series data of the vehicle state quantity based on the vertical vibration acceleration detected by the sprung acceleration sensor 8, the vehicle height detected by the vehicle height sensor 9, and the road surface detected by the road surface measurement sensor 10. The controller 12 outputs time-series data of the optimum command value based on the time-series data of the road surface input and the time-series data of the vehicle state quantity. At this time, the latest optimum command value corresponds to the current optimum damping force command value. As a result, the controller 12 outputs the most appropriate damping force command value for the current vehicle and road surface. The damping force command value corresponds to the current value for driving the damping force variable actuator 7 .

Next, the learning method of the DNN 13 of the controller 12 will be described with reference to the explanatory diagram shown in FIG. The DNN 13 is constructed by executing the processes of (1) direct optimal control command value search, (2) command value learning, and (3) weighting factor download.

First, an analysis model 20 including a vehicle model 21 is constructed in order to directly search for the optimum control command value. FIG. 2 illustrates a case where the vehicle model 21 is a one-wheel model. The vehicle model 21 may be, for example, a pair of left and right two-wheel models. The vehicle model 21 receives the road surface input and the optimum command value from the direct optimum control unit 22 . The direct optimum control unit 22 obtains the optimum command value according to the procedure of direct optimum control command value search described below.

(1) Direct optimal control command value search The direct optimal control unit 22 searches for the optimal command value by iterative calculation using the analysis model 20 including the vehicle model 21 in advance. The search for the optimum command value is formulated as the optimum control problem shown below, and obtained numerically using an optimization method.

It is assumed that the motion of the target vehicle is represented by Equation 1 using the equation of state. Note that the dot in the formula means the first order differentiation (d/dt) with respect to time t.

Here, x is a state quantity and u is a control input. The initial conditions of the state equation are given as in Equation 2.

The equality constraint conditions and inequality constraint conditions imposed between the initial time t ₀ and the end time t _f are expressed by Equations 3 and 4.

The optimum control problem is a problem of finding a control input u(t) that minimizes the evaluation function J shown in Equation 5 while satisfying the state equation shown in Equation 1, the initial conditions shown in Equation 2, and the constraint conditions shown in Equations 3 and 4.

It is very difficult to solve such constrained optimal control problems. For this reason, a direct method is used as an optimization method because it can easily handle constraint conditions. This method converts an optimal control problem into a parameter optimization problem and obtains a solution using an optimization method.

To transform the optimal control problem into a parameter optimization problem, we divide the initial time _t0 to the final time _tf into N intervals. If the end times of each section are represented _by t ₁ , t ₂ , .

A continuous input u(t) is replaced by a discrete value u _i at the end time of each interval, as shown in Equation (7).

_Numerical integration is performed on _the state equations for the inputs _{u 0 , u 1} , _. At this time, the input in each interval is obtained by linearly interpolating the input given at the end time of each interval. As a result of the above, the state quantity is determined for the input, and the evaluation function and the constraint conditions are expressed by this. Thus, the transformed parameter optimization problem can be expressed as follows.

Assuming that the parameters to be optimized are collectively X, the equation (8) is obtained.

Therefore, the evaluation function shown in Equation 5 is expressed as in Equation 9.

Also, the constraint conditions shown in equations 3 and 4 are expressed as equations 10 and 11.

In this way, the optimum control problem as described above can be converted into a parameter optimization problem represented by the equations (8) to (11).

The evaluation function J for formulating the problem of obtaining the optimum control command according to the road surface as an optimum control problem is defined as shown in Equation 12 so that the vertical acceleration Az is minimized to provide a comfortable ride and the control command value u is reduced. Here, q ₁ and q ₂ are weighting factors. q ₁ and q ₂ are set in advance based on, for example, experimental results.

The direct optimum control unit 22 obtains the parameter optimum problem formulated in this manner through numerical analysis using an optimization technique, and derives optimum command values for various road surfaces.

(2) Command value learning The optimal command value derived by direct optimal control command value search is used as the output, and the road surface profile and vehicle state quantity at that time are used as inputs, and the input and output of various road surfaces are learned by the DNN 23, which is artificial intelligence. The DNN 23 is a deep neural network for learning and has the same configuration as the in-vehicle DNN 13 . The time-series data of the road surface input and the time-series data of the vehicle state quantity are input to the DNN 23 as the road surface profile. At this time, weighting coefficients between neurons in the DNN 23 are obtained using the time-series data of the optimum command values corresponding to the road surface input and the vehicle state quantity as teacher data.

(3) Weighting Coefficient Download The weighting coefficient of the DNN 23 learned by the command value learning is set in the DNN 13 which is the command value determination unit of the actual ECU 11 . Thereby, the DNN 13 of the controller 12 is configured.

(4) Optimal Command Value Calculation The controller 12 including the DNN 13 is mounted on the vehicle. The input side of the controller 12 is connected with the sprung acceleration sensor 8 , the vehicle height sensor 9 and the road surface measurement sensor 10 . The output side of the controller 12 is connected to the damping force variable actuator 7 of the variable damper 6 . The controller 12 acquires road surface inputs and vehicle state quantities based on detection signals from the sprung acceleration sensor 8 , vehicle height sensor 9 and road surface measurement sensor 10 . The controller 12 inputs to the DNN 13 the road surface input time-series data and the vehicle state quantity time-series data as the road surface profile. When the DNN 13 receives the time-series data of the road surface input and the vehicle state quantity, the DNN 13 outputs a command value for the variable damper 6 as an optimum command according to the learning result.

In this way, the direct optimum control unit 22 derives the direct optimum control command by off-line numerical optimization under various conditions. The artificial intelligence (DNN 23) is made to learn the road surface profile, vehicle state quantity and optimum command at that time. As a result, the controller 12 (ECU 11) equipped with the DNN 13 can directly implement optimum control without performing optimization for each step.

Next, in order to confirm the effect of the DNN 13 on the ride comfort performance, the feedback control based on the conventional skyhook control rule as the control of the comparative example and the control by the DNN 13 of the first embodiment were compared in terms of ride comfort performance and the like by simulating the vehicle. The simulation conditions were vehicle specifications assuming, for example, an E-segment sedan. As a simulation model, a 1/4 vehicle model considering sprung mass and unsprung mass was used. For the road surface, we set up an undulating road to check the basic damping performance of the sprung mass.

Simulation results are shown in FIGS. 3 and 4. FIG. FIG. 3 shows the sprung vertical acceleration RMS value (effective value). The vertical axis in FIG. 3 indicates the effective value of the first frequency region (0.5-2 Hz), which is the fluffy feeling region. The horizontal axis in FIG. 3 indicates the effective value of the second frequency range (2-50 Hz), which is a vibration range other than the fluffiness. As shown in FIG. 3, the control according to the first embodiment reduces vibration in both the first frequency region and the second frequency region compared to the control (skyhook control rule) according to the comparative example. As a result, it can be seen that the control according to the first embodiment achieves both low-frequency sprung damping and high-frequency vibration isolation, and has improved performance compared to conventional control (Skyhook control rule).

FIG. 4 shows the results of temporal changes in sprung acceleration and the like. As a result, the control according to the first embodiment sets the damping force higher before the road surface input changes, compared to the conventional control (Skyhook control rule). Therefore, in the control according to the first embodiment, although the acceleration is large until around 0.6 seconds, the peak value of the acceleration is small after that, the waveform is smooth, and the convergence after passing through the undulating road is also improved in the control according to the first embodiment, compared to the control according to the comparative example (Skyhook control rule). From these, it can be seen that the control according to the first embodiment achieves high damping performance and smoothness compared to the control (skyhook control rule) according to the comparative example.

Thus, according to the present embodiment, the suspension control device is interposed between the vehicle body 1 and the wheels 2 of the vehicle, and controls the variable damper 6 capable of adjusting the force between the vehicle body 1 and the wheels 2, the vehicle state quantity acquisition unit (sprung acceleration sensor 8 and vehicle height sensor 9) for detecting the vehicle state quantity, the road surface profile acquisition unit (road surface measurement sensor 10) for detecting the profile of the road surface, and the force generated by the variable damper 6 based on the results acquired by the vehicle state quantity acquisition unit and the road surface profile acquisition unit. a controller 12; The controller 12 has a DNN 13 serving as an AI learning section that learns a command value for the variable damper 6 based on the acquisition results of the vehicle state quantity acquisition section and the road surface profile acquisition section. At this time, the DNN 13 learns the command value obtained by the optimization method in advance so as to minimize the evaluation function J, and the obtained results of the vehicle state quantity obtaining section and the road surface profile obtaining section.

As a result, it becomes possible to perform control with a true optimum command according to each road surface, so that it is possible to improve ride comfort and steering stability performance. In addition, even for a non-linear controlled object such as a semi-active suspension (variable damper 6), control can be constructed without complicated modeling. It is impossible to implement direct optimum control in the ECU 11 considering the characteristics of each road surface, vehicle, and variable damper 6 with the current technology due to the calculation time. However, an artificial intelligence (DNN 13) that has learned the direct optimum command in advance can be incorporated in the ECU 11 (controller 12). Therefore, direct optimum control can be realized by the ECU 11 .

Next, FIG. 5 shows a second embodiment. A feature of the second embodiment resides in the addition of a feedback control section that performs feedback control of the variable damper based on the vertical behavior of the vehicle. In addition, in 2nd Embodiment, the same code|symbol shall be attached|subjected to the component same as 1st Embodiment mentioned above, and the description shall be abbreviate|omitted.

The ECU 30 according to the second embodiment is configured similarly to the ECU 11 according to the first embodiment. Therefore, the ECU 30 has a memory 30A. In addition to this, the ECU 30 has a controller 31 .

The controller 31 has a learned DNN 13 . In addition to this, the controller 31 has a feedback control section 32 . The feedback control unit 32 is configured in the same manner as the controller disclosed in Patent Document 1, for example. Therefore, the feedback control unit 32 calculates the target damping force from the sprung acceleration output from the sprung acceleration sensor 8 as the vehicle state quantity acquiring unit, based on, for example, the skyhook control law. The feedback control section 32 controls the damping force of the variable damper 6 based on the calculated target damping force. Note that the feedback control unit 32 may control the variable damper 6 based on other control laws such as bilinear optimum control, H∞ control, etc., in addition to the skyhook control law.

At this time, the ECU 30 obtains the relative speed between the sprung and unsprung states based on the vehicle height output from the vehicle height sensor 9 as a vehicle state quantity obtaining unit. When the relative speed is low, such as when the stroke is reversed between the extension stroke and the contraction stroke of the shock absorber, the feedback control unit 32 calculates a corrected damping force by reducing the target damping force, and outputs a control signal corresponding to this corrected damping force to the shock absorber. As a result, it is possible to reduce the occurrence of noise and jerk caused by a sudden change in damping force.

On the other hand, when the relative speed is high, the feedback control unit 32 calculates a corrected damping force with a relaxed restriction on the target damping force compared to when the relative speed is low, and outputs a control signal corresponding to the large value of the corrected damping force. As a result, when the relative speed is high, a large damping force can be generated by the shock absorber, damping performance can be secured, and ride comfort can be improved.

The controller 31 selects one of the DNN 13 and the feedback control section 32 according to the learning state of the DNN 13 and the like. For example, when the DNN 13 has been learned, the controller 31 selects the DNN 13 . In this case, the controller 31 controls the damping force of the variable damper 6 using the DNN 13 .

On the other hand, for example, when the DNN 13 has not learned, the controller 31 selects the feedback control section 32 . Further, even if the DNN 13 has already learned, the controller 31 selects the feedback control section 32 even when the vehicle travels on a road surface that is significantly different from the learned road surface profile. In these cases, the controller 31 controls the damping force of the variable damper 6 using the feedback control section 32 .

Thus, substantially the same effects as those of the first embodiment can be obtained in the second embodiment as well. Further, in the second embodiment, the DNN 13 serving as an AI learning unit is added with a feedback control unit 32 that performs feedback control of the variable damper 6 based on the vertical behavior of the vehicle. Thereby, the controller 31 controls the variable damper 6 by the feedback control section 32 when the DNN 13 has not yet learned. As a result, in the second embodiment, it is possible to ensure ride comfort control performance comparable to that of the conventional technology. In addition, by combining the AI control by the DNN 13 and the feedback control by the feedback control unit 32, it is possible to cope with changes in vehicle specifications even on a learned road surface.

Note that when the vehicle travels on an unlearned road surface, the road surface profile may be stored or transmitted to an external server. In this case, the direct optimum control unit obtains the optimum command value based on the newly acquired unlearned road surface profile. After that, the road surface profile and the optimum command value are added, and the weighting coefficients between neurons of the DNN are learned again. After the learning is completed, the weighting coefficients of the vehicle-mounted DNN are updated. As a result, the damping force of the variable damper 6 can be optimally controlled using the DNN the next time the vehicle travels on the same road surface.

Next, FIG. 6 shows a third embodiment. The feature of the third embodiment resides in setting different DNN weights for the front wheels and the rear wheels and independently controlling the variable dampers for the four wheels. In addition, in 3rd Embodiment, the same code|symbol shall be attached|subjected to the component same as 1st Embodiment mentioned above, and the description shall be abbreviate|omitted.

The suspension control system according to the third embodiment includes sprung acceleration sensors 8 _FL , 8 _FR , and 8 _RR that detect the vertical acceleration of the vehicle body 1 of the three wheels (front left wheel FL, front right wheel FR, and rear right wheel RR), and vehicle height sensors 9 _FL , 9 _FR , 9 _RL , and 9 _RR that detect the vehicle height of the four wheels (front left wheel FL, front right wheel FR, rear left wheel RL, and rear right wheel RR). , a road surface measurement sensor _10L for measuring the road profile measured by the left wheel (LH) tire, and a road surface measurement sensor _10R for measuring the road surface profile measured by the right wheel (RH) tire. The road surface measurement sensor _10L outputs a road surface profile on which the left wheel tire travels. The road surface measurement sensor _10R outputs a road surface profile on which the right wheel tire runs. A single road surface measurement sensor may be provided instead of the road surface measurement sensor 10- _L and the road surface measurement sensor 10- _R . In this case, a single road surface measurement sensor outputs a road surface profile on which the left wheel tire runs and a road surface profile on which the right wheel tire runs.

The ECU 40 according to the third embodiment includes a four-wheel acceleration calculator 41 that calculates the vertical acceleration of the four wheels from the vertical acceleration of the three wheels. In this case, the sprung acceleration sensors 8 _FL , 8 _FR , 8 _RR , the vehicle height sensors 9 _FL , 9 _FR , 9 _RL , 9 _RR , and the four-wheel acceleration calculator 41 constitute a vehicle state quantity acquisition unit.

The ECU 40 includes delay units _42L and _42R . The input side of the delay unit _42L is connected to the road surface measurement sensor _10L . The output side of the delay unit _42L is connected to the rear wheel _DNN45L . The input side of the delay section _42R is connected to the road surface measurement sensor _10R . The output side of the delay section _42R is connected to the rear wheel _DNN45R . The delay units _42L , _42R delay the road surface profile by a value (time) obtained by dividing the wheelbase by the vehicle speed, and output the delayed road surface profile to the rear wheel DNNs _45L , _45R .

The input side of the ECU 40 is connected to the sprung acceleration sensors _8FL , _8FR , _{8RR , the} vehicle height sensors _{9FL , 9FR , 9RL} , _9RR and _the road surface measurement sensors _10L , _10R . , 7 _{RL and} 7 _RR . The ECU 40 outputs the road surface profile and vehicle state quantity to the controller 43 based on the values detected by the sprung acceleration sensors _8FL , _8FR , _8RR , the vehicle height sensors _9FL , _9FR , _9RL , _9RR and the road surface measurement sensors _10L , _10R . The controller 43 obtains forces to be generated by the variable dampers 6 _FL , 6 _FR , 6 _RL and 6 _RR (force generating mechanisms) based on the road surface profile and the vehicle state quantity, and outputs command signals thereof to the damping force variable actuators 7 _FL , 7 _FR , 7 _RL and 7 _RR .

The controller 43 includes learned front wheel DNNs _44L , _44R and rear wheel DNNs _45L , _45R . The front wheel DNN 44 _L , 44 _R and the rear wheel DNN 45 _L , 45 _R are constructed in the same manner as the DNN according to the first embodiment. For this reason, the front wheel DNNs 44 _L and 44 _R and the rear wheel DNNs 45 _L and 45 _R are AI learning units, which are composed of, for example, four or more multilayer neural networks. Each layer has a plurality of neurons, and neurons of two adjacent layers are connected by a weighting factor. Weighting factors are set by prior learning. For the front wheel DNNs 44 _L and 44 _R and the rear wheel DNNs 45 _L and 45 _R , for example, similar to the first embodiment, weighting factors obtained by learning using a one-wheel vehicle model are set. Preferably, the front wheel DNNs 44 _L and 44 _R and the rear wheel DNNs 45 _L and 45 _R are set with weighting factors obtained by learning using a four-wheel vehicle model.

The controller 43 acquires time-series data (road surface profile) of the road surface input and time-series data of the vehicle state quantity for each of the four wheels based on the values of vertical vibration acceleration detected by the sprung acceleration sensors 8 _FL , 8 _FR , and 8 _RR , the vehicle height detection values of the vehicle height sensors 9 _FL , 9 _FR , 9 _RL , and 9 _RR , and the road surface detection values of the road surface measurement sensors 10 _L and 10 _R.

At this time, the front wheel DNN 44 _L outputs time-series data of optimum command values for the left front wheel based on the time-series data of the road surface input and the time-series data of the vehicle state quantity. The front wheel DNN 44 _R outputs time-series data of optimum command values for the right front wheel based on the time-series data of the road surface input and the time-series data of the vehicle state quantity. The rear wheel DNN 45 _L outputs time-series data of optimum command values for the left rear wheel based on the time-series data of the road surface input and the time-series data of the vehicle state quantity. The rear wheel DNN 45 _R outputs time-series data of optimum command values for the right rear wheel based on the time-series data of the road surface input and the time-series data of the vehicle state quantity.

Thus, even in the third embodiment, substantially the same effects as in the first embodiment can be obtained. Further, in the third embodiment, front wheel DNN _44L , _44R and rear wheel DNN _45L , _45R are provided. Therefore, the variable dampers 6 _FL , 6 _FR , 6 _RL and 6 _RR can be separately controlled for the front wheels and the rear wheels. In addition to this, in the third embodiment, different weights are set for the front wheel DNNs _44L , _44R and the rear wheel DNNs _45L , _45R , and the variable dampers _6FL , _6FR , _6RL , _6RR are independently controlled for the four wheels. The road surface measurement sensors 10 _L and 10 _R output a road surface profile on which the left (LH) and right (RH) tires run. Therefore, the suspension control system of the third embodiment can cope with different road surfaces for the left and right wheels. If the road surface profiles of the left and right wheels are different, the behavior of each wheel will be different. Therefore, in order to cope with this, in the third embodiment, the vehicle height information and acceleration information of each wheel are input to the DNN of each wheel (front wheel DNN _44L , _44R and rear wheel DNN _45L , _45R ), and the command value of each wheel is calculated. The calculated command values are output to the variable dampers 6 _FL , 6 _FR , 6 _RL and 6 _RR of each wheel for control.

Next, FIGS. 1 and 7 show a fourth embodiment. A feature of the fourth embodiment is that the optimum command value is derived using a full vehicle model. In addition, in 4th Embodiment, the same code|symbol shall be attached|subjected to the component same as 1st Embodiment mentioned above, and the description shall be abbreviate|omitted.

When learning the weighting coefficients of the DNN 54 of the controller 12, a direct optimum control command value search is performed. At this time, in the first embodiment, the vehicle model 21 included in the analysis model 20 is a one-wheel model. In contrast, in the fourth embodiment, a four-wheel model is used as the vehicle model 51 included in the analysis model 50 . Therefore, the vehicle model 51 includes a left front wheel 2 _FL , a right front wheel 2 _FR , a left rear wheel 2 _RL and a right rear wheel 2 _RR , which are wheels. The vehicle model 51 also includes a suspension system 4 _FL for the front left wheel 2 _FL , a suspension system 4 _FR for the front right wheel 2 _FR , a suspension system 4 _RL for the rear left wheel 2 _{RL ,} and a suspension system 4 _RR for the rear right wheel 2 RR. The suspension devices 4 _FL , 4 _FR , 4 _RL and 4 _RR include springs 5 and variable dampers 6 _FL , 6 _FR , 6 _RL and 6 _RR . In addition to this, the vehicle model 51 is a model that takes into account, for example, the pitch and roll of the vehicle. At this time, a front wheel stabilizer 52 _F is provided between the left front wheel 2 _FL and the right front wheel 2 _FR . A front wheel stabilizer 52 _R is provided between the left rear wheel 2 _RL and the right rear wheel 2 _RR . Therefore, the vehicle model 51 can obtain the vehicle state quantity in consideration of the stabilizer reaction force. The vehicle model 51 is not limited to the four-wheel model shown in FIG. 7, and various four-wheel models can be used.

A vehicle model 51 receives a road surface input and an optimum command value from a direct optimum control unit 53 . As the direct optimum control command value search, the method for obtaining the optimum command value by the direct optimum control unit 53 is the same as the method for obtaining the optimum command value by the direct optimum control unit 22 according to the first embodiment.

As in the first embodiment, the optimal command value derived by the direct optimal control command value search is used as the output, and the road surface profile and vehicle state quantity at that time are used as the input, and various inputs and outputs of the road surface are learned by the DNN, which is artificial intelligence. The DNN is a deep neural network for learning, and has the same configuration as the vehicle-mounted DNN 54 . Time-series data of the road surface input and time-series data of the vehicle state quantity are input to the DNN as the road surface profile. At this time, weighting coefficients between neurons in the DNN are obtained using the time-series data of the optimum command values corresponding to the road surface input and the vehicle state quantity as teacher data.

The DNN weighting coefficients learned by the command value learning are set in the DNN 54 which is the actual command value determination unit of the ECU 11 . Thereby, the DNN 54 of the controller 12 is configured.

Thus, even in the fourth embodiment, substantially the same effects as in the first embodiment can be obtained. In the fourth embodiment, a vehicle model 51 consisting of a full vehicle model is used to find the optimum command value. Therefore, the DNN 54 can output an optimum command value that takes into consideration the behavior of the actual vehicle, and can enhance the effect of suppressing pitch and roll, for example.

In each of the above-described embodiments, the vehicle state quantity acquisition unit is provided with the sprung acceleration sensor 8 and the vehicle height sensor 9 . The present invention is not limited to this, and the vehicle state quantity acquisition unit may include, for example, in addition to the sprung acceleration sensor 8 and the vehicle height sensor 9, a portion that calculates the vehicle state quantity based on the vertical vibration acceleration detected by the sprung acceleration sensor 8 and the vehicle height detected by the vehicle height sensor 9. In addition to detection signals from the sprung acceleration sensor and the vehicle height sensor, the vehicle state quantity acquisition unit acquires information related to the vehicle state, such as vehicle speed, based on signals from the CAN (Controller Area Network), and may calculate or estimate the state quantity of the vehicle in consideration of these information. In this case, the vehicle state quantity acquisition unit is configured by the calculation units in the ECUs 11, 30, and 40 in addition to various sensors.

In each of the embodiments described above, the road surface profile acquisition unit detects the road surface profile by the road surface measurement sensor 10 . The present invention is not limited to this, and the road surface profile acquisition unit may acquire information from a server based on GPS data, for example, or may acquire information from another vehicle through inter-vehicle communication. The road surface profile acquisition unit may estimate the road surface profile based on the vertical vibration acceleration detected by the sprung acceleration sensor 8 and the vehicle height detected by the vehicle height sensor 9 . In this case, the road surface profile acquisition unit is configured by the arithmetic units in the ECUs 11, 30, and 40 in addition to various sensors.

In each of the above-described embodiments, the suspension control device has the vehicle state quantity acquisition section and the road surface profile acquisition section. The present invention is not limited to this, and the suspension control device may not have the road surface profile acquisition section. In this case, the controller of the suspension control device adjusts the force generated by the force generating mechanism based on the result obtained only by the vehicle state quantity obtaining section. The controller has an AI learning section that learns a command value for the force generating mechanism based on the acquisition result of the vehicle state quantity acquisition section. The AI learning unit of the controller learns the command value obtained by the optimization method so as to minimize a predetermined evaluation function and the acquisition result of the vehicle state quantity acquisition unit.

In each of the above-described embodiments, the variable damper 6, which is a semi-active damper, is used as the force generating mechanism. The present invention is not limited to this, and an active damper (either an electric actuator or a hydraulic actuator) may be used as the force generating mechanism. In each of the above-described embodiments, the case where the force generating mechanism that generates an adjustable force between the vehicle body 1 side and the wheel 2 side is configured by the variable damper 6 composed of a damping force adjustable hydraulic shock absorber has been described as an example. The present invention is not limited to this, and for example, the force generating mechanism may be configured by an air suspension, a stabilizer (Kinesus), an electromagnetic suspension, etc., in addition to the hydraulic shock absorber.

In each of the above-described embodiments, the vehicle behavior device used for a four-wheeled vehicle has been described as an example. However, the present invention is not limited to this, and can be applied to, for example, a two-wheeled or three-wheeled vehicle, a work vehicle, a truck, a bus, or the like, which is a transport vehicle.

It goes without saying that each of the above-described embodiments is an example, and partial replacement or combination of configurations shown in different embodiments is possible.

Next, as a suspension control device included in the above-described embodiment, for example, the aspects described below are conceivable.

A suspension control device according to a first aspect is provided between a vehicle body and wheels of a vehicle and includes a force generation mechanism capable of adjusting a force between the vehicle body and the wheels; a vehicle state quantity acquisition unit for detecting or estimating a vehicle state quantity; and

A suspension control device according to a second aspect includes a force generation mechanism interposed between a vehicle body and wheels and capable of adjusting a force between the vehicle body and the wheels, a vehicle state quantity acquisition unit for detecting or estimating a vehicle state quantity, a road surface profile acquisition unit for detecting or estimating a road surface profile, and a controller for controlling to adjust the force generated by the force generation mechanism based on results obtained by the vehicle state quantity acquisition unit and the road surface profile acquisition unit, wherein the controller comprises the vehicle state quantity acquisition unit and the road surface profile acquisition unit. It is characterized by having an AI learning unit that learns a command value for the force generation mechanism based on the acquisition result of the unit.

As a third aspect , in the first aspect , the AI learning unit of the controller learns a command value obtained by an optimization method so as to minimize a predetermined evaluation function and the acquisition result of the vehicle state quantity acquisition unit.

As a fourth aspect , in the second aspect , the AI learning unit of the controller learns command values obtained by an optimization method that minimizes a predetermined evaluation function, and acquisition results of the vehicle state quantity acquiring unit and the road surface profile acquiring unit.

As a fifth aspect , in any one of the first to fourth aspects , the controller further includes a feedback control section that calculates a target damping force obtained from the vehicle state quantity obtained by the vehicle state quantity acquisition section.

1 vehicle body 2 wheel 2 _FL left front wheel (wheel)
2 _FR right front wheel (wheel)
2 _RL left rear wheel (wheel)
2 _RR right rear wheel (wheel)
3 tires 4, 4 _FL , 4 _FR , 4 _RL , 4 _RR suspension devices 5 suspension springs (springs)
6, 6 _FL , 6 _FR , 6 _RL , 6 _RR variable damper (force generation mechanism)
7 damping force variable actuator 8, 8 _FL , 8 _FR , 8 _RR sprung acceleration sensor (vehicle state quantity acquisition unit)
9, 9 _FL , 9 _FR , 9 _RL , 9 _RR vehicle height sensor (vehicle state quantity acquisition unit)
10, 10 _L , 10 _R road surface measurement sensor (road surface profile acquisition unit)
11, 30, 40 ECU
12, 31, 43 controller 13, 23, 54 DNN (AI learning part)
20, 50 analysis model 21, 51 vehicle model 22, 53 direct optimum control unit 32 feedback control unit 44 _L , 44 _R front wheel DNN (AI learning unit)
45 _L , 45 _R rear wheel DNN (AI learning part)

Claims (3)

a force generating mechanism interposed between a vehicle body and wheels of a vehicle and capable of adjusting a force between the vehicle body and the wheels;
a vehicle state quantity acquisition unit that detects or estimates the state quantity of the vehicle;
a controller that controls to adjust the force generated by the force generation mechanism based on the result obtained by the vehicle state quantity obtaining unit;
The controller is
an AI learning unit that learns a command value for the force generation mechanism based on the acquisition result of the vehicle state quantity acquisition unit ;
The suspension control device, wherein the AI learning unit of the controller learns a command value obtained by an optimization method so as to minimize a predetermined evaluation function and the obtained result of the vehicle state quantity obtaining unit. a force generating mechanism interposed between a vehicle body and wheels of a vehicle and capable of adjusting a force between the vehicle body and the wheels;
a vehicle state quantity acquisition unit that detects or estimates the state quantity of the vehicle;
a road surface profile acquisition unit that detects or estimates a road surface profile;
a controller that controls to adjust the generated force of the force generation mechanism based on the acquisition results of the vehicle state quantity acquisition unit and the road surface profile acquisition unit;
The controller is
an AI learning unit that learns a command value for the force generation mechanism based on the acquisition results of the vehicle state quantity acquisition unit and the road surface profile acquisition unit ;
The AI learning unit of the controller learns a command value obtained by an optimization method that minimizes a predetermined evaluation function, and the obtained results of the vehicle state quantity obtaining unit and the road surface profile obtaining unit. Suspension control device. 3. The suspension control device according to claim 1 , wherein the controller further comprises a feedback control section that calculates a target damping force obtained from the vehicle state quantity acquired by the vehicle state quantity acquisition section.

Images