JP2000207002A - Optimum control method for shock absorber - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve comfortableness of riding on a vehicle by optimizing control parameters by genetic algorithm by using as an evaluation function the difference in time differentiation between the entropy in a shock absorber and entropy given to the shock absorber from a controller. SOLUTION: An optimization part provided to a learning part of a control system optimize the control parameters by making the tutor signal (input/output value of fuzzy neural network) of a controller for learning genetically evolve by using as the evaluation function the difference between the time differential dSc/dt of the entropy from the controller for learning and the time differential dSs/dt of the internal entropy of a vehicle and a suspension to be controlled which is obtained from a motion model and by using a genetical algorism so as to reduce the time differential difference. Consequently, the turning performance of the vehicle and comfortableness of riding the vehicle can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for optimally controlling a shock absorber having a non-linear motion characteristic.

[0002]

2. Description of the Related Art Although optimization control for optimizing the operation characteristics of a controlled object has been used for a long time, optimization control of a controlled object having non-linear operation characteristics has not been fully developed. Conventionally, there is no general analysis method for non-linear operation characteristics. Therefore, when controlling such a control target, a controller is often designed by linear approximation.
That is, since it was difficult to evaluate the non-linear operating characteristics, when controlling a controlled object having non-linear operating characteristics, an appropriate equilibrium point of the operating characteristics of the controlled object was found, and an appropriate equilibrium point near the equilibrium point was found. A method of performing control while linearizing operation characteristics of a control target and evaluating pseudo operation characteristics has been used.

[0003]

However, in the conventional method of performing control by linearly approximating nonlinear motion characteristics,
Accurate control can be performed in a region near the equilibrium point to be linearized, but accuracy degrades as the distance from this region increases, and there are problems such as inability to cope with large changes in the environment surrounding the control target. . Shock absorbers used in automobiles and motorcycles are also one of the controlled objects that have non-linear operating characteristics, and the characteristics of the damping force, which is the output of the shock absorbers, are used to determine the turning performance and riding Since it affects the comfort, it is desired to optimize the characteristics.However, as described above, it is difficult to evaluate the non-linear operation characteristics, and there is a problem in the method of controlling by linear approximation. Optimum control could not be performed. In order to solve this problem, instead of directly evaluating the motion characteristics of the damping force of the shock absorber, the acceleration was evaluated, and a genetic algorithm was evolved to minimize this acceleration to achieve optimal damping force control. It is proposed that this method, for example,
If the control is performed in response to a large acceleration, an overshoot occurs, resulting in a poor ride quality and not practical. An object of the present invention is to solve the above-mentioned conventional problems and to provide a control method capable of performing optimal control of a shock absorber having a non-linear motion characteristic.

[0004]

In order to achieve the above-mentioned object, an optimal control method for a shock absorber according to the present invention is provided by a method for controlling a time derivative of entropy inside a shock absorber and a control device for controlling the shock absorber. The method is characterized in that a difference between a given entropy and a time derivative is obtained, and the control parameter of the control device is optimized by a genetic algorithm using the difference as an evaluation function.

[0005]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an embodiment of a method for optimally controlling a shock absorber according to the present invention will be described with reference to an embodiment shown in the accompanying drawings. FIG.
1 is a schematic block diagram of a control system that executes a shock absorber optimal control method according to the present invention. This control system is divided into an actual control part which is actually mounted on the vehicle and executes the control of the vehicle and the suspension, and a learning part for optimizing the control device of the actual control part based on the motion model of the vehicle and the suspension. After optimizing the learning control device in the learning part by simulation, the parameters of the control device in the actual control part are determined using the optimized parameters of the learning control device. In the present embodiment, a shock absorber of a damping coefficient control method is employed as a shock absorber of a suspension, and a control device of an actual control portion outputs a signal for controlling a throttle amount of an oil passage in each shock absorber. Specifically, the control device of the actual control unit uses a plurality of sensors provided on the vehicle and the suspension to determine the vertical position z0 of the vehicle, the pitch angle β, the roll angle α, the suspension angle η of each wheel, and the arm angle of each wheel. Information about the angle θ, the suspension length z6 of each wheel, and the deflection z12 of each wheel is detected, and based on the detection result, an optimum throttle amount of each shock absorber is estimated and output according to a predetermined fuzzy rule. A fuzzy controller is used (see FIG. 2A). The learning part includes a vehicle and suspension motion model of the real control part, a learning control device including a fuzzy neural network corresponding to the control device of the real control part (see FIG. 2B), and the learning control device. And an optimizing part for optimizing. The optimization part is the time derivative of the entropy from the learning controller (d
Sc / dt) and the time derivative (dSs / dt) of the internal entropy of the control object (ie, vehicle and suspension) obtained from the motion model as an evaluation function, so that the difference in the time derivative of the entropy becomes small. Using a genetic algorithm, a teacher signal (input / output values of a fuzzy neural network) of a learning control device is genetically evolved to optimize the learning control device. As described above, the fuzzy rule of the fuzzy inference device as the control device of the actual control portion is determined based on the fuzzy neural network as the learning control device optimized using the vehicle and suspension motion models in the learning portion.

Here, in the above-mentioned optimization part, the principle of using the difference between the time derivative of the entropy from the control device (dSc / dt) and the time derivative of the internal entropy of the controlled object (dSs / dt) as an evaluation function will be briefly described. Will be described.

[0007]

In the present invention, the waste such as disturbance of the entire control system of the plant is determined by the difference between the time derivative dSu / dt of the entropy of the plant to be controlled and the time derivative dSc / dt of the entropy of the control system of the plant. Then, evaluation is performed in relation to the stability of the controlled object represented by the Lyapunov function. That is, the operation of the plant becomes more stable as the time derivative of the Lyapunov function, in other words, the difference in entropy becomes smaller.

Next, a method of calculating a motion model of a vehicle and a suspension used in the learning control device will be described. FIG. 3 is a diagram showing parameters of a motion model of a vehicle and a suspension in an automobile, and FIGS. 4 to 7 are enlarged views of each wheel in FIG.
1. First, regarding the transformation matrix, the coordinates of each unit represented by the generalized coordinates Xn are represented by coordinates in an inertial system. 1.1
One point in the inertial system, here, the geometric center Pc of the vehicle body
Or represented by coordinates xr, yr, zr {r} just below
Is used as a base point to obtain a conversion matrix from Or to each part. Note that {2} is the local coordinates with the center of gravity of the vehicle as the origin,
{7} is the local coordinate with the center of gravity of the suspension as the origin, {10} is the local coordinate with the center of gravity of the arm as the origin,
{12} is the local coordinate with the center of gravity of the wheel as the origin,
{13} is local coordinates having the origin at the point of contact of the wheel with the road surface, and {14} is local coordinates having the origin at the stabilizer connection point.

The learning section acquires a pseudo sensor signal based on the vehicle and suspension motion model obtained as described above, activates the learning control device based on the pseudo sensor signal, and in the optimization section. The time derivative of entropy from the learning control device and the time derivative of entropy inside the controlled object are calculated. In this embodiment, the entropy inside the controlled object is obtained from the motion model as described above, and the time derivative dScs /
A value dSs / dt obtained by adding dt and the time derivative dSss / dt of the entropy related to the suspension is used. In this embodiment, a shock absorber of a damping coefficient control method is employed as a shock absorber, and a learning control device (a control device of an actual control portion) controls a throttle amount of an oil passage in the shock absorber. The output of the device does not include a velocity component, and thus the entropy of the learning control device is always zero. Therefore, the optimizing part uses the difference between the time derivative of the entropy from the learning control device and the time derivative of the entropy inside the controlled object as an evaluation function, and calculates the difference (that is, the time of the entropy inside the controlled object in this embodiment). A teacher signal (input / output values of a fuzzy neural network) in the learning control device is genetically evolved by a genetic algorithm so as to reduce the differentiation, and the learning control device learns based on the teacher signal and optimizes. Then, based on the optimized learning control device, the parameters of the control device of the actual control portion (in this embodiment, the fuzzy rule in the fuzzy inference device) are determined, whereby the non-linearity is determined. It is possible to optimally control a suspension having the characteristics described above.

In the embodiment described above, a fuzzy neural network is used as a learning control device in a learning portion, and a fuzzy controller is used as a control device in an actual control portion. However, this configuration is limited to this embodiment. Instead, for example, the control device in the learning portion and the control device in the actual control portion may have the same configuration.
Further, in this embodiment, the control system is divided into a learning part and an actual control part, and only the actual control part is mounted on an actual vehicle, and the learning part is configured to perform optimization using a motion model. However, this is not limited to the present embodiment, and the optimization of the learning portion may be performed using an actual vehicle, and accordingly, both the learning portion and the actual control portion may be performed. You may comprise so that it may be mounted in a vehicle.

[0012]

As described above, the optimal control method of the shock absorber according to the present invention is based on the time derivative of the entropy inside the shock absorber and the time derivative of the entropy given to the shock absorber from the control device for controlling the shock absorber. Since the difference is obtained and the control parameter of the control device is optimized by a genetic algorithm using the difference as an evaluation function, it is possible to optimally control the shock absorber that has been difficult to perform the optimization control because of having a non-linear characteristic. Will be possible. According to the optimum control method of the second aspect of the present invention, the output of the control device is a control signal containing no speed component, and the time derivative of entropy given to the shock absorber from the control device is always zero. This has the effect that optimization can be performed using only the time derivative of the entropy inside the shock absorber. Further, claim 3 of the present invention
According to the optimization method according to the control device is configured by a fuzzy neural network or a fuzzy controller,
The genetic algorithm optimizes the value of the coupling coefficient or the fuzzy rule, so that the genetic algorithm finds a global optimal solution, and a local optimal solution can be obtained with a fuzzy neuron based on the optimal solution. Thus, it is possible to efficiently find the optimal solution. Further, according to the optimization method of claim 4 of the present invention, the control parameters of the learning control device are controlled by a genetic algorithm using the evaluation function using the learning control device and the real control control device. Since the control parameters of the actual control device are determined based on the optimized control parameters of the learning control device, and the shock absorber is controlled by the actual control device, the actual control device is It will always be possible to operate with the optimized control parameters. Further, according to the optimizing method according to the fifth aspect of the present invention, the optimization of the learning control device is performed by the simulation using the motion model related to the shock absorber, so that the learning control device can be easily optimized. This has the effect. According to the optimization method of claim 6 of the present invention, the shock absorber has a structure in which the damping force changes by changing the cross-sectional area of the oil passage, and the control device controls the cross-sectional area of the oil passage. Is controlled, so that the output of the control device becomes a control signal containing no speed component, and the time derivative of entropy given to the shock absorber from the control device can always be made zero. Therefore, there is an effect that the optimization can be performed using only the time derivative of the entropy inside the shock absorber.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of a control system that executes a shock absorber optimal control method according to the present invention.

FIG. 2A is a schematic diagram of a control device in an actual control portion, and FIG. 2B is a schematic diagram of a control device in a learning portion.

FIG. 3 is a diagram showing parameters of a motion model of a vehicle and a suspension in an automobile.

FIG. 4 is an enlarged view of a right front wheel in FIG. 3;

FIG. 5 shows an enlarged view of the left front wheel in FIG. 3;

FIG. 6 is an enlarged view of the right rear wheel in FIG. 3;

FIG. 7 shows an enlarged view of the left rear wheel in FIG. 3;

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G05B 13/04 G05B 13/04 F-term (Reference) 3D001 AA02 CA01 DA03 DA17 EB32 ED13 ED17 3J069 AA34 EE63 5H004 GB12 HA07 HA10 HB07 JB18 JB30 KC02 KC08 KC12 KC33 KD03 KD05 KD14 KD18 KD23 KD43 KD67 LA05 LA12

Claims (6)

[Claims] The difference between the time derivative of the entropy inside the shock absorber and the time derivative of the entropy given to the shock absorber from the control device for controlling the shock absorber is determined, and the difference is used as an evaluation function by a genetic algorithm to perform the control of the control device. An optimal control method for a shock absorber, characterized by optimizing control parameters. 2. An optimal control method of a shock absorber, wherein an output of the control device is a control signal containing no speed component, and a time derivative of entropy applied from the control device to the shock absorber is always zero. 3. An optimal control method for a shock absorber, wherein the control device is constituted by a fuzzy neural network or a fuzzy controller, and a value of a coupling coefficient or a fuzzy rule is optimized by a genetic algorithm. 4. The learning device according to claim 1, wherein the control device comprises a learning control device and a real control device, and the control parameters of the learning control device are optimized by a genetic algorithm using the evaluation function. The control parameter of the real control device is determined based on the control parameter of the real control device, and the control of the shock absorber is performed by the real control device. Optimal control method of shock absorber. 5. The optimal control method for a shock absorber according to claim 4, wherein optimization of the learning control device is performed by simulation using a motion model for a vehicle equipped with the shock absorber. 6. The shock absorber has a structure in which a damping force is changed by changing a cross-sectional area of an oil passage, and the control device controls a throttle amount of a valve device capable of changing a cross-sectional area of the oil passage. The optimal control method for a shock absorber according to any one of claims 2 to 5, wherein: