JPH04372409A - Vehicle motion predicting device - Google Patents

Abstract

PURPOSE:To predict a vehicle motion by detecting required parameters for a vehicle by a plural sensors, outputting actually measured data relating to past vehicle motion with use of a sampler, and outputting predicted data for the vehicle motion at a predetermined time by a neutral net. CONSTITUTION:A plural sensors 1-6 detect required parameters, that is a handle angle, a handle torque, a yaw angular speed, a lateral acceleration, a vertical acceleration and a vehicle velocity. A sampler 7 takes inputs the vehicle parameters from the sensors 1-6 and outputs past data from a current time (t) to t-nDELTAt as necessary. A neutral net 8 corrects a weight so as to reduce a difference between a self-output signal and a teacher signal as vehicle data at a time DELTAt' from the current time for achieving a learning function. As a result of this learning, data till the time nDELTAt before the current time are inputted to the neutral net 8, thereby a vehicle motion at a DELTAt'sec. from the current time can be predicted.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vehicle motion prediction device for predicting vehicle motion and controlling, for example, a damping force switching device for a damper or a cylinder pressure control device for an active suspension.

0002

[Prior Art] In order to perform control according to the movement of a vehicle,
Until now, there has been no device that can predict that movement. However, a damping force switching device for a damper has been known in the past as having a concept similar to this. This damping force switching device is
The system detects the lateral acceleration of the vehicle and sets a hard damping force when the lateral acceleration occurs to improve steering stability. On the other hand, when lateral acceleration is not occurring, that is, when the vehicle is traveling straight, the damping force is set to be soft, giving priority to ride comfort. In other words, this conventional device, depending on the lateral acceleration,
The damping force can be switched between hard and soft. However, the above-mentioned device that detects this lateral acceleration and then switches the damping force inevitably has poor responsiveness. Therefore, currently, a method is adopted in which the steering wheel angle is detected and the damping force is switched on the assumption that lateral acceleration occurs when the steering wheel is turned.

0003

[Problems to be Solved by the Invention] In the conventional device as described above, it is not always possible to accurately predict the next motion situation of the vehicle and control control equipment such as the damping force switching device of the damper. It wasn't quite satisfactory. Also, if the lateral acceleration depends only on the steering wheel angle,
There was also the problem that accurate lateral acceleration could not be detected. An object of the present invention is to provide a device that enables, for example, a damper damping force switching device, an active suspension cylinder pressure control device, etc. to be controlled in an optimal state by predicting vehicle motion. That's true.

0004

[Means for Solving the Problems] The present invention includes a plurality of sensors that detect parameters meaningful for predicting vehicle motion, such as steering wheel angle, steering torque, and yaw angular velocity, and detecting values of each of these sensors. A sampler that outputs measured data of past vehicle motion as a base, and is connected to this sampler and outputs predicted data of future vehicle motion after a time Δt′ based on the measured data from this sampler. The system is equipped with a neural network, inputs past data into this neural network, and corrects the weights so that the difference between its own output signal and the teacher signal, which is vehicle data Δt' time after the current time, becomes small. It has characteristics in points.

[0005]

[Operation] Since the present invention is constructed as described above, future predicted data after a time Δt' is obtained based on past actually measured data. Then, this predicted data is compared with the teacher signal, which is the actual vehicle data after the time Δt', and the neural network adjusts the weight by itself so that the difference becomes smaller.

[0006]

[Effects of the Invention] According to the vehicle motion prediction device of the present invention,
The future movement of the vehicle after the time Δt' can be predicted. Therefore, for example, by combining this device with a damping force switching device for a damper or a cylinder pressure control device for an active suspension, the responsiveness of these devices can be improved. Of course, the field of application of the present invention is not limited to these, but can be applied to various devices based on the prediction of vehicle conditions. However, it is natural that the vehicle motion parameters to be input/output will be different if the application target is different.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The embodiment shown in FIG. 1 provides a plurality of sensors 1-6 for detecting parameters of significance for predicting vehicle motion. Sensor 1 is for each handle θH
, sensor 2 detects the steering wheel torque Tin, sensor 3 detects the yaw angular velocity γ, sensor 4 detects the lateral acceleration αy, and sensor 5 detects the longitudinal acceleration αx.
The sensor 6 detects the vehicle speed V.

Each of the sensors 1 to 6 described above is connected to a sampler 7. This sampler 7 includes each of the above-mentioned sensors 1 to 1.
When the vehicle parameters from 6 are input, the current time t
Past data from to t-nΔt is output as necessary. For example, if n=10 and Δt=0.5 seconds,
Outputs past data in 0.5 second increments starting from 5 seconds ago.

The sampler 7 has a neural network 8
is connected to. Therefore, past data output from this sampler 7 is input to this neural network 8. Furthermore, data after a time Δt' from the current time is given to the neural network 8 as a teacher signal.

Next, the operation of this embodiment will be explained. The sampler 7, which records the vehicle parameters from the sensors 1 to 6 as described above, outputs past data from the current time t to t-nΔt as necessary. The past data outputted from the sampler 7 in this way is input to the neural network 8, which uses its own output signal and a teacher signal that is vehicle data Δt' time after the current time. The learning function is performed by correcting the weights so that the difference between them becomes smaller. As a result of this learning, by inputting data up to nΔt time before the current time into the neural network 8, it becomes possible to predict the vehicle motion Δt' seconds after the current time.

FIG. 2 shows a prediction device T consisting of the sampler 7 and a neural network 8 as a damping force switching device 1 for a damper.
This shows the state where it is connected to 0. As is clear from FIG. 2, depending on the longitudinal acceleration and lateral acceleration output from the prediction device T, the damping force of the damper is adjusted by soft S,
It can be adjusted to medium M and hard H.

FIG. 3 shows a similar prediction device T connected to a cylinder pressure control device 11 of an active suspension, which adjusts the front-to-rear distribution ratio of cylinder pressure in accordance with the rate of change in yaw angular velocity.

In any case, by using the prediction device of the above embodiment, by predicting the motion of the vehicle, it is possible to optimize, for example, the damping force switching device 10 of the damper, the cylinder pressure control device 11 of the active suspension, etc. can be controlled in a controlled manner.

[Brief explanation of drawings]

FIG. 1 is a circuit diagram.

FIG. 2 is a circuit diagram in which this device is connected to a damping force switching device of a damper.

FIG. 3 is a circuit diagram in which this device is connected to a cylinder pressure control device of an active suspension.

[Explanation of symbols]

1 Sensor 2 Sensor 3 Sensor 4 Sensor 5 Sensor 6 Sensor 7 Sampler 8 Neural network

Claims (1)

[Claims] Claim 1: A plurality of sensors that detect meaningful parameters for predicting vehicle motion, such as steering wheel angle, steering torque, and yaw angular velocity, and past vehicle motion based on the detected values of each of these sensors. A sampler that outputs measured data is connected to this sampler, and Δt is calculated based on the measured data from this sampler.
A vehicle motion prediction device equipped with a neural network that outputs predictive data of future vehicle motion in the future.