You are on page 1of 5

ControlEng.Practice,Vol. 3, No. 5, pp.

723-727, 1995

Pergamon
0967-0661 (95)00053-4

Copyright 1995 Elsevier Science Ltd Printed in Great Britain. All fights re~'ved 0967-0661/95 $9.50 + 0.00

CONTINUING EDUCATION FOR APPLYING CONTROL THEORY TO AUTOMOBILES AT TOYOTA


H. K o u s a k a * , M . Y a m a s h i t a * , Y. H i r a n o * * , A. O h a t a * * a n d S. Ijichi*

*Toyota Motor Corporation, Aichi 471, Japan **Higashi-Fuji Technical Center, Toyota Motor Corporation, Susono, Shizuoka 410-11, Japan

(Received September 1994; in final form March 1995)

Abstract: Since 1990, the Toyota Motor Corporation has been conducting internal education programs on control technology for its engineers, using the MATLAB and inverted pendulum experimental devices. With three years of experience, some major complaints were voiced concerning these programs, such as "not practical", or "not helpful to understand the basic theory of control". As countermeasures for these problems, SIMULINK has been introduced as a new tool for constructing models easily. "One-year Project" has also been started which enables the participants to solve the problems they face in their daily work with the help of university researchers and Toyota's experts. Key Words: Automobiles; Education; Optimal control; Robust control; Adaptive control

1. INTRODUCTION Vehicles, which are Toyota's products, consist of countless technologies. With the rapid progress and expansion of automobile technologies, there are some areas where the individual engineer or group has diffculty in obtaining new skills as a result of their own efforts. In order to encourage engineers to obtain the new skills which were required in various departments, but which had not yet been acquired in spite of their importance, the company started its own "Technical Education Programs". At present, there are three courses; "Electronics", "Materials" and "Control". Each course consists of lecture sessions, given by University researchers, and training sessions, led by internal experts. The duration varies from three to twenty-five days. The curriculums are studied and decided through meetings between representatives of the departments concerned and the University researchers. The opinions of participants are collected after the programs, and are reflected in the next year's programs. The Control Course started in 1990 to fulfil the increasing need for people skilled in control engineering. For instance, as the legal restrictions on emissions and fuel consumption are getting tougher, more-

advanced and complicated control technologies are required. Also, as shown in Table 1, various vehicle control systems such as Four Wheel Steering, Traction Control Systems, Anti-lock Braking Systems and Active Suspensions have been introduced in a very short period, to improve safety and comfort. Today, control technology has a big influence on vehicle performance and production costs. On the other hand, some surveys have Woved that in the company, there are not enough engineers who are able to keep up with the very rapid progress in the area of control technology. With this background, the Control Course started in 1990 to develop engineers who are able to utilize control theories in the actual development of vehicle components. 2. CURRICULUM OF CONTROL COURSES IN TECHNICAL EDUCATION PROGRAMS A lot of vehicle control objects are stable and SISO systems; heuristic methods have therefore often been used in the development of control systems. This means that mathematical models and control theory are not utilized. This type of development procedure is no longer effective in the current situation, where vehicle control must be far more advanced and complicated to meet various requirements, such as clean emissions, safety, customer satisfaction, and cost effi-

723

724

H. Kousaka et al.

Table 1 Development of electronic control in the automobile.


1970
I

19~
I

1~0
l

Engine Control

*Electric Fuel Injection ( l ~ l ) ', i p ! *Spark AdvanCe COnu'ol t : *Knock Control i * Vmiable Cylinder Control [ Vitiable Induetion,Control *Auto Drive

Drive Train Control

Suspension Control

i *VariabldValve "timing ,' *Lean Bm'nEngiue ," ', *Electronically ConlrolledT,~msmission(ECT) i *Full-thne4WD *Activ~ TorqueSplit 4WD i *Bk~monicHeightControl(EHc) ElectronicallyM o d ~ SuspensionflEMS) Air su~emdbn TEMS
tActibe Suspend'on ',

:
*Speed Seining PS (On-off solenoid)

: *Pk~o TEMS

Power Steering (PS)

4 Wheel Steering (4WS)

*Progressive PS (Line~ solenoid) *HydraulicP~aefionPS *ElectrictllyactuatedPumpPS *EleoCr6mnguetk:MO~ PS *Nisum H1CA~ *Mazda4W~ *To~ta 4WS *,'NiuanSuper-HICAS ' *Toyota Active4WS i (yaw rate feedback)
*4WD',ABS
l i

Anti-lock Brake

(ABS)

* R ~ WheelABS

*4 W l ~ l ABS

*TRC (Engine) :
Traction Control (TRC) *TRC (Engine & Throttle) *TRC (Engine & Brake)

ciency. However, since vehicle control systems can be designed by heuristic methods to some extent, the engineers generally do without control theory. To improve this situation, two courses were planned as follows. The Basic Course This was planned to teach the basic procedures for designing control systems, such as "Modeling =~ Identification =~ Controller Design ~ Simulation", and how to use the latest design tools. The Advanced Course This was planned to promote understanding of the latest control theories. Both courses include not only the lecture sessions, but also training sessions using MATLAB (Math Works Inc., 1992a) and inverted pendulum experimental devices, referred to here as IPED. The IPED was selected because it is easy to handle, and its feedback effects are visible. It is generally recognized as the most popular experimental device for control theory education. It provides a common object which can be used to teach control theory to engineers from various areas: power train, chassis, factory engineering, and so
Off.

many cases, the number of state variables is implicitly reduced by sliding surface, and case classification is done by if-then logic. Therefore, regulators such as Linear Quadratic methods cannot handle all the complicated vehicle control systems. To design excellent control systems, the following contents should be clearly understood. Analvsis of the control objects This includes deciding the control goals, studying the disturbance characteristics, modeling the objects, identification of parameters and uncertainties, and so
on.

D e s i ~ of the controller This covers decisions concerning estimators and regulators, and the design of pre/post compensators, including the feedforward and linearizations. Analysis of the control systems This means the evaluation of performance and robustness in closed-loop systems, and so on. Table 2 Contents of the curriculum.
B u i c Course f~3nte~lt Pertod Introduction. O.b day How to use MATLAB. 0.5 Programming a real-time controller. 2 Design of an LQ controller, and implementation. 2 Advanced Course t~3fltent Linem" control theory. Robmt control theory. Feedback linearization. Sliding control. Adaptive control. Fuzzy control. l.~tio~i 2 clay 2 1 1 1 1

In most vehicle control systems, two-degree-of-freedom controllers with feedforward and feedback are used. Such controllers also include compensation for nonlinearities and time-delays, such as gain scheduling methods, and learning and adaptive functions. Usually, continuous actuators are seldom used as control devices, and discrete actuators with only a few modes (such as on-off valves) are adopted instead. In

Applying Control Theory to Automobiles From the above considerations, as shown in Table 2, the curriculum was designed to cover many areas of vehicle control technologies without becoming excessively in-depth. 3. EVALUATION OF CONTROL COURSES In both the Basic Course and the Advanced Course, the number of participants is limited to ten respectively, so that each pa~icipant has his/her own training device. Each course was held twice per year. After three years of experience, it has become clear that the Basic Course is very effective in teaching the procedure of designing vehicle conu'ol systems. In the Advanced Course, however, most of the participants complained that the contents cannot be applied to their daily jobs. Especially, the onderstanding of the theories was very poor even though the participants concentrated on learning these theories in the curriculum. Some participants reported that they could not keep up with the curriculum and they had the impression that the control theories are too difficult. In actual applications, MIMO systems are seldom treated, and even in SISO systems, high-order dynamics in the plants are usually ignored. Since the program covers these subjects, it may be unnecessarily complicated. Furthermore, many participants tend to think that since many nonlinearities exist in vehicle systems, basic linear control theory is useless in their daily work. Although controllers and pre/post f'flters should be simultaneously designed, they tend to think that these are completely separate issues. Before applying robust con~ol, they also tend to think that compensation for the variations of the plant should be covered by complex learning and adaptive functions. Thus, this lack of understanding prevents the participants from applying the control theories in their daily work. The IPED might be considered the cause of this problem. Some participants said that training devices like the IPED are too far from what they actually use in their work, therefore making it difficult for them to apply what they had learned.

725

From the above evaluation, it was felt that it would be more effective for both the participants and the instructors to study examples of actual problems to which control thegn/may be applied. 4. M O D I F I C A T I O N OF THE ADVANCED COURSE In order to improve the above si~tion, the program of the advanced course was modified in 1993. Time schedule of the course is shown in Table 3. The ten participants are divided into three groups, and each group has its own university researcher as a tutor for both sessions. Lecture with Practice Basically, the first half of this program is designed to teach how to use the linear control theories, especially LQG methods. The latter half is an attempt to put more emphasis on robust control, nonlinear control and intelfigent control. To solve the problem that the theory is too difficult to absorb in a short period of time, the duration was prolonged from two days to four days, concentrating more on LQG methods. Also, the SIMULINK (Math Works Inc., 1992b) was introduced to reduce the load of model construction. This is planned to solve problems which simulate the actual control objects that the participants face. It is hoped that the project will become mote like cooperative research, performed by the universities and Toyota. Project themes are collected from the participants, and categorized into three group themes. Each group has its own group leader, and he finally decides what the group will do, through meetings with the tutor. The themes in 1993 were; Anti-lock Braking System (ABS), Four Wheel Steering (4WS), Engine Control System for Air-Fuel Ratio.

Table 3 Time schedule of new advanced course. Lecture with Practice May Aug. Trends and Problems of Vehicle Control. CAD for C~trol Systems. Analysis of Conlxol Object. Design of Controller. Analysis of Control System. Sep. One-Year Project Choice of Theme Candidates in the Corporation. Determinating Themes and Goals with University Researchers. Starting Projects with Lectures. (One Hour per Day)

Nov. Fuzzy Control Theory. Robust Control Theory. Adaptive Control Theory. Dec.

Continuing Projects in Each Group. Reporting Previous Results. Modification Acc~din_gto Suggestions by Researchers. Continuing Projects with Lectures. (One Hour per Day)

Continuing Projects in Each Group. Reporting F'mal Results.

726

H. Kousaka et al.

The instructors had to perform extensive preparation to adopt the control theories to these items, while the participants also had to spend much time on these projects in addition to lecture sessions. As this program requires a lot of additional work from both the instructors and the participants, it must be ensured that the skills and knowledge obtained are useful. Therefore, it was decided that at the end of the program, each group would report their results to their managers and the Tutor Professor, in the form of a presentation. 5. RESULTS OF ONE-YEAR PROJECT This section shows the technical results of the OneYear Project as examples. Anti-lock Braking Svstem lABS) The ABS group designed control systems for a wellknown quarter-car model, which involves the dynamics of actual hydraulic actu.a.t0~. Figure 1 shows the block diagram of the control system. Using the pole assignment method, a controller to compensate for slip ratio was designed. From the simulation results shown in Figure 2, it has been confirmed that the slip ratio was controlled to the peak value (0.15) for this road.
Actuating T i m e

Four Wheel Steering (4WS) The 4WS group tried to design a layered controller consisting of a model-following controller in the main loop and an LQG controller in the minor loop, as shown in Figure 3. Figure 4 shows the step responses of the rear wheel angles for the case that the two types of commands (0.25, 1.0 [deg]) are input to the perturbed plant. The remarkable improvement was confh'med in the minor loop of the rear steering actuator. This results indicated that the closed-loop systems satisfied some specifications for performance and robustness.

s ~
Actuator Vehic~

~ s~

ControU~ of MinorLaop

R~ce

Rc41ulatotof M ~ Loop Yaw RIt*

Ot~erve~

-I . .

Vehicle

I Wheel
Controller of Slip Ratio

Angular Velocity

Fig. 3 Block diagram of control system for 4WS (Four-Wheel Steering System).

Controller of [ 4 - - I Hydraulic Pressure | - I

1.2

Command
1

""" Uncontrolled

| Vehicle Model ~

L]

Estimator of MasterPressure
0

Estimator of ~.-S Characteristic

"-d
0 13.4 ~ 0.2

<

0.6

../':' / Uncontrolled

Fig. 1 Block diagram of control system for ABS (Anti-lock Braking System).

i A", Contro|led
Command
0 0.0.$ 0.1 O.IS 0.2 0.25 0.3 0.35 0.4 0.45 0.3

-0.2

t , ~ - . - - - ~ ~

e Velocity ......4 ...........

time (see) Fig. 4 Time responses of rear wheel angles on perturbed plant.

dine Oee) 0.3 , , ,

0.2

0,4

0.6

0.1 (u:,O

1.2

1.4

L6

1.1

Fig. 2 Time responses of longitudinJal velocities and slip ratio due to braking.

Engine Control System for Air-Fuel Ratio Before controller design, the Engine group identified a model with time delay from the cylinder to the air-fuel ratio sensor, using least-square methods. Figure 5 shows the results of system identification. In the model with sensor dynamics, the loss function decreases drastically, compared with the model without sensor dynamics. Therefore, this result suggests that the model using six delays and sensor dynamics is the most suitable for controller design.

Applying

Control

Theory

to Automobiles

727

250

~ o [ ...........~ .............~...............................................~...............~..............~--..- . . ~ . - ' : : ~

.........

210 ............... )- ..........

~-.............. ~ ............... [ ............... ~ ............... . .............. ~-.............. r ...... i ! I ) ~ I l l

i/
] -

,
, .

..............~............. ~.............~...............~ . . . . . ~ ..........


.............. .......... .....

FFI

i
1600 3

i
4

)
$

'
6

i
7 |

'
9 10

IMay ~

oylim~ ~ $mffi

Fig. 5 Results of system identification for Air-Fuel


Ratio Control System. 6. CONCLUSIONS In order to address the problem that it is difficult for the participants to understand completely the procedure for applying the control theories through only the training using the IPED, the advanced course was modified. Through the One-Year Project, the participants have learned how to solve actual problems that they face in their daily work, with the help of universit), researchers. The best way to teach engineers how to utilize control theory is not to teach them some complicated calculations, but to make them understand the concepts and the essence of the control theories, along with the actual problems they would face in applying them. The ultimate goal of Toyota's Control Education Programs is not to develop engineers who can create new control theories, but to develop those who can apply the existing theories to vehicle development. It is the authors' opinion that a~y engineers who have their own experience in other a/eas can be educated to utilize control theory. Based on this belief, it is intended to continue the Programs with further improvement. ACKNOWLEDGMENTS The authors wish to thank Prof. K. Furuta of the Tokyo Institute of Technology and Mr. W. Vesik of the University of Toronto for their invaluable help and encouragement. REFERENCES The Math Works Inc. (1992a). MATLAB : High-Performance Numeric Computation and Visualization Software. User's Guide. The Math Works Inc. (1992b). SIMULINK : A Program for Simulating Dynamic System. User's Guide.

You might also like