SISTEM KENDALI STABILITAS KENDARAAN TERINTEGRASI BERBASIS KECERDASAN BUATAN Disertasi disusun untuk memenuhi salah satu syarat memperoleh gelar Doktor (Dr) di Institut Teknologi Sepuluh Nopember Oleh Muchammad Harly NRP : 2204301601 Tanggal Ujian Periode Wisuda
: :
Disetujui oleh Tim Penguji Disertasi: 1). Prof. Dr.Eng. Ir. Mauridhi H.P, M.Eng NIP : 195809161986011001
(Pembimbing I)
2). Prof. Ir. I.N. Sutantra , M.Sc,Ph.D NIP : 195106051978031002
(Pembimbing II)
3). Prof. Ir. Indra Nurhadi, M.Sc, Ph.D NIP : 130 515 675
(Penguji I)
4). Prof. Ir. Soebagio, MSEE, Ph.D NIP : 194102281967051001
(Penguji II)
5). Prof. Dr. Ing. Ir. I. Made Londen B. , M.Eng NIP : 195811061986011002
(Penguji III)
Direktur Program Pascasarjana
Prof. Ir. Suparno, MSIE, Ph.D NIP : 194807101976031002
i
INTEGRATED VEHICLE STABILITY CONTROL BASED ON ARTIFICIAL INTELLIGENT Student name
: Muchammad Harly
NRP
: 2204301601
1st Promotor
: Prof. Dr.Eng Ir. Mauridhi H P, M.Eng.
2nd Promotor
: Prof. Ir. I N Sutantra , M.Sc, Ph.D. ABSTRACT
Without integrated process among vehicle stability control subsystem like Electronics Stability Program ( ESP), Electronics Four Wheel Steering ( E-4Ws), Active Suspension ( AS) as vehicle in a critical condition each of the system don't work optimally , even harm their performance one with other, so that the important variable output of stability ( state vehicle) such as Yaw-Rate ( YR) , Vehicle Side Slip ( VSS) , Roll Angle ( RA) cannot fulfill control target. In this time there are 4 integration methods that are applied in the vehicle stability. These are feed-forward method, H algorithm, G. Leitmann method and Nonlinear Predictive Control (NLPC). Weakness of 4 methods; the first is less precise plant model and don't adapt to the change of indefinite structure and its parameter at high frequency; the second its work is not based on intelligence. This paper developed a intelligent integration of control design, which exploits combined Multi Dimension Fuzzy C-Mean Clustering (MDFC) and Adaptive Backpropagation Control (ABC). ABC consist of nurel etwork plant (NNP) and neural network controller(NNC). NNP covers uncertainty vehicle dynamics, drivers character and environment (Tree in One Dynamics System (TODS)). Architecture NNP results from fixed order neural network (FONN) and varied order neural network (VONN). NNC was the inverse of NNP. The first step was the training of NNP based on center clustered by MDFC. The second step was the validation and the production process conducted by both virtual and real time to compare approximation and control performance between MDFCABC controller and others. Based on data analyze was concluded that MDFC-ABC performance become better than others on either virtual or real time It is represented by lowest LSE stability variable; result of virtual LSE yaw rate VONN 0.09, while FONN 0.13, Robust 0.17, H-Infinite 0.22, Feed Forward 0.26, NLPC 0,31; result of virtual LSE vehicle side slip VONN 0.16, while FONN 0.22, Robust 0.26, H-Infinite 0.33, Feed Forward 0.42, NLPC 0.47; result of virtual LSE vehicle roll angle VONN 0.02, while FONN 0.03, Robust 0.05, H-Infinite 0.1, Feed Forward 0.2, NLPC 0.22; result of real time maximum error tabulation of yaw rate VONN 7.5 rad/sec, while FONN 8.8 rad/sec Robust 11.0 rad/sec; result of real time maximum error tabulation of vehicle side slip VONN 15.3o, FONN 16.7o, Robust 20.4o; result of real time maximum error tabulation of throttle position VONN 2.8 volt, FONN 2.9 volt, Robust 3.2 volt Key Word: TODS, MDFC-ABC, FOC, FONN, FONN
ii
SISTEM KENDALI STABILITAS KENDARAAN TERINTEGRASI BERBASIS KECERDASAN BUATAN Nama mahasiswa
: Muchammad Harly
NRP
: 2204301601
Pembimbing 1
: Prof. Dr.Eng Ir. Mauridhi H P, M.Eng.
Pembimbing 2
: Prof. Ir. I N Sutantra , M.Sc, Ph.D.
ABSTRAK Tanpa integrasi, diantara subsistem kendali stabilitas kendaraan Electronics Stability Program ( ESP), Electronics Four Wheel Steering (E-4WS) dan Active Suspension ( AS) maka saat kendaraan dalam kondisi kritis masing-masing sistem tidak akan bekerja optimal. Bahkan merugikan kinerja satu subsystem dengan yang lain, sehingga variablevariabel keluaran stabilitas (vehicle state); Yaw-rate (YR s ) , Vehicle side slip (VSS s ) , Roll angle (RA s ) tidak mampu mencapai target yang diinginkan Saat ini sudah ada 4 metoda integrasi yang dipakai yaitu metoda feedforward , algoritma kendali H , metoda G. Leitman dan Nonliear Predictive Control(NLPC). Kelemahan ke 4 metoda yang pertama adalah kurang tepat pemodelan plantnya dan tidak adaptif terhadap perubahan struktur tak tentu dan parameternya pada frekuensi tinggi, yang kedua disebabkan system tidak bekerja secara cerdas. Di dalam disertasi ini dirancang sebuah kendali integrasi yang bekerja secara cerdas dengan menggunakan Multi Dimensi Fuzzy C-Mean ( MDFC) dan skema adaptive backthrough control (ABC). ABC tersusun dari neural network plant (NNP) dan neural network controller (NNC). NNP meliputi kendaraan , pengemudi dan lingkungan (Tree in One Dynamics System (TODS)) Arsitektur NNP tersusun dari fixed order neural network (FONN) dan varied order neural network (VONN), sedang NNC adalah invers dari NNP. Langkah pertama adalah pelatihan dan perbaharuan parameter NNP dengan semua data center hasil pengelompokan MDFC. Langkah kedua adalah memvalidasi dan mengetes kinerja system kendali MDFC-ABC dibanding sytem kendali integrasi yang lain secara virtual maupun real time. Dari analisis data disimpulkan bahwa kinerja kendali stabilitas terintegrasi MDFCABC secara virtual maupun real time terbaik dibanding system kendali stabilitas terintegrasi yang lain. Hal ini direpresentasikan dengan makin rendahnya LSE variabel stabilitas; hasil virtual LSE yaw rate VONN 0.09,untuk FONN 0.13, Robust 0.17, HInfinite 0.22, Feed Forward 0.26, NLPC 0,31; vehicle side slip VONN 0.16, untuk FONN 0.22, Robust 0.26, H-Infinite 0.33, Feed Forward 0.42, NLPC 0.47; vehicle roll angle VONN 0.02, untuk FONN 0.03, Robust 0.05, H-Infinite 0.1, Feed Forward 0.2, NLPC 0.22; hasil real time tabulasi error maksimum yaw rate VONN 7.5 rad/sec, untuk FONN 8.8 rad/sec Robust 11.0 rad/sec; LSE error vehicle side slip VONN 15.3o, FONN 16.7o, Robust 20.4o; LSE throttle position VONN 2.8 volt, FONN 2.9 volt, Robust 3.2 volt
Kata kunci: TODS, MDFC-ABC, FOC, FONN, FONN
iii
DAFTAR ISI Halaman Pengesahan
i
ABSTRAK
ii-iii
Daftar Isi
iv-v
Daftar Gambar
vi-vii
Daftar Tabel
viii
Daftar Simbol
ix-x
Daftar Singkatan
xi-xiii
BAB I : PENDAHULUAN 1.1 Deskripsi Kendaraan
1
1.2 Rumusan Permasalahan
1
1.3 Dampak Permasalahan
3
1.4 Solusi Penelitian Sebelumnya
3
1.5 Solusi Yang Dilakukan di dalam Disertasi
5
1.6 Sasaran Penelitian ( Objective)
6
1.7 Kontribusi Hasil Penelitian
7
BAB II : KAJIAN PUSTAKA (KENDALI INTEGRASI TIDAK CERDAS) 2.1 Kendali Stabilitas Kendaraan Tidak Cerdas
9
2.1.1 Kendali Stabilitas Kendaraan Tidak Cerdas Tanpa Integrasi
9
2.1.2 Kendali Stabilitas Kendaraan Tidak Cerdas Terintegrasi
16
2.2 Kendali Stabilitas Kendaraan Cerdas
31
2.2.1 Kendali Stabilitas Kendaraan CerdasTanpa Integrasi
32
2.2.2 Kendali Stabilitas Kendaraan Cerdas Terintegrasi
48
BAB III : THREE IN ONE DYNAMICS SYSTEM (TODS) SEBAGAI PEMODELAN BERKENDARA ADAPTIF 3.1
Posisi TODS Pada Road Map Penelitian Teknologi Kendali Kendaraan iv
49
3.2
Representasi TODS
51
3.2.1 Uncertainty Vehicle Dynamics
51
3.2.2 Interaksi Pengemudi, Kendaraan dan Lingkungan
58
3.3
Clustering dengan Multi Dimension Fuzzy C-Mean (MDFC)
59
3.4
Model Reference Neural Network Control (MRNNC)
61
3.4.1 Arsitektur Neural Network Plant (NNP)
62
3.4.2 Arsitektur Neural Network Control (NNC)
68
BAB IV : VARIED ORDER NEURAL NETWORK (VONN) SEBAGAI KENDALI CERDAS STABILITAS TODS 4.1
Varied Order of MDFC-ABC Based Integrated Stability Control
77
4.2
Representasi Real Word untuk TODS
77
4.3
Model Real TODS yang Dirancang
80
4.4
Learning Process
84
4.4.1 Inisialisasi Struktur VONN
85
4.4.2 Meng-update Joint Weight
90
4.5
104
Rancangan Varied Order dari ABC Based Integrated Stability Control
4.5.1 Rancangan Neural Network Plant
105
4.5.2 Rancangan Neural Network Controller
105
4.5.3 Variabel Referensi
106
4.5.4 Mean Square Error Control
108
4.5.5 Performance Control Index
111
BAB V : METODA INTEGRASI TODS-VONN 5.1
Road Map TODS-VONN
114
5.2
Prosedur Rancangan Penelitian TODS –VONN
115
5.3
Perangkat Validasi TODS-VONN Real Time
117
5.3.1 Skema Blok Hard Ware
117
5.3.2 Pemilihan Titik Ukur Sensor-sensor
118
5.3.3 Pengukuran Sensor dan Actuator Mesin
118
5.3.4 Pengukuran Sensor dan Actuator ESP
119
5.3.5 Pengukuran Sensor dan Actuator 4 WS
119
5.3.6 Test Drive
120 v
5.4
Pengambilan Data TODS
122
5.4.1 Real Time TODS non-Integrasi
122
5.5
125
Algoritma Sistem Kendali TODS-VONN Virtual
5.5.1 Algoritma dari pencarian Cluster
125
5.5.2 Pelatihan Identifikasi NNP dan Kendali Integrasi NNC
126
5.6 Pengolahan Data TODS –VONN Real Time
128
BAB VI : ANALISIS METODA TODS-VONN 6.1
Analisis Hasil Simulasi
130
6.1.1 Kinerja TODS Fixed Order Intelligent Controller (TODS-FONN)
130
6.1.2 Hasil Kinerja TODS-VONN
133
6.2
139
Analisis Hasil TODS-VONN Real Time
BAB VII : KESIMPULAN 7.1
Keunggulan Pemodelan TODS
141
7.1.1
Pemodelan virtual TODS
141
7.1.2
PrototypeTODS
141
7.2
Kinerja Kendali Stabilitas TODS-VONN
142
7.2.1
Kinerja Virtual Sistem Kendali Integrasi TODS-VONN
142
7.2.2
Kinerja Real Time Sistem Kendali Integrasi TODS-VONN
142
SARAN
143
DAFTAR PUSTAKA
144
LAMPIRAN 1 : hasil-hasil simulasi virtual LAMPIRAN 2 : hasil-hasil test drive real time LAMPIRAN 3 : source code simulasi (virtual) LAMPIRAN 4 : source code real time (prototype) LAMPIRAN 5 : hasil-hasil prototype LAMPIRAN 6 : daftar gambar dan tabel LAMPIRAN 7 : konferensi international (TODS) LAMPIRAN 8 : journal international (VONN)
148 157 164 199 234 240 246 247
vi
DAFTAR GAMBAR Gambar
Halaman
1.1 1.2 1.3 1.4 1.5 1.6 1.7 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
Grafik Perkembangan Kecepatan Maksimal Kendaraan Grafik Penurunan Berat Kendaraan Grafik Perkembangan Jumlah Kendaraan Gandeng-Bertingkat Grafik Batasan Gangguan Akibat Percepatan & frekuensi Grafik Batas Ketahanan Pengemudi Terhadap Percepatan & frekuensi Grafik Batasan Amplitudo Perpindahan Pengemudi & frekuensi Grafik Perkembangan Jumlah dan Panjang Tunel Manuver Sky-hook Feed Forward Manuver J-curve Feed Forward Manuver Sky-hook H-Infinite Manuver J-curve H-Infinite Manuver dan respon NLPC Respon Yaw rate single ABS Respon Yaw rate multi ABS Komparasi Performance ABS standar dan Modifikasi (a) Langkah Pengereman Pumpless (b) Langkah Anti Lock Pumpless (c) Langkah Deliveri Tekanan Rendah Pumpless 3.1 Road Map Penelitian Teknologi Automotive 3.2 Suspensi Standar 3.3 Semi Active Damper Suspension 3.4 Passive Sky-hook Damper Suspension 3.5 Active Sky-hook Damper Suspension 3.6 Blok Plant Susensi Active dan Kontrol 3.7 Skema MRNNC 3.8 Sistem Plant yang akan diintegrasikan (TODS) 3.9 Unsertainty Vehicle Plant Model 3.10 Variation to generate TODS 3.11 Model TODS Plant 3.12 Pemodelan Sebelumnya dan TODS 3.13 Integrated Control based ABC 3.14 Blok Simulasi Kendali MRNNC 3.15 Mapping gFNN optimized 3.16 Mapping gPNN optimized 3.17 Arsitektur Adaptive Network gHFNN 3.18 Arsitektur MLP-NNC 3.19 Neuron dari node NNP dan NNC 4.1 Hidden Layer Neuron 4.2 Output Layer Neuron 4.3 Arsitektur Varied Order Neural Network 4.4 Flow Chart Design Process VONN 4.5 Struktur Koefisien (order) sub-chromosome
vii
4 5 6 6 6 7 16 17 19 19 21 24 26 26 27 28 28 34 35 35 36 36 38 39 40 41 47 48 49 50 50 51 53 57 58 61 71 72 72 75 75
4.6 Koefisien Initian Condition sub-chromosome 4.7 Skema Error Identifikasi 4.8 Skema VONN-ABC 4.9 Arsitektur VONN-NNP 4.10 Arsitektur VONN-NNC 5.1 Road Map Penelitian TODS 5.2 Prosedur TODS 5.3 Skema Hard Ware Kendali TODS 5.4 Distribusi Titik ukur Sensor-actuator Mesin (EFI) 5.5 Distribusi Titik ukur Sensor-actuator Electronic Stability Programm (ESP) 5.6 Distribusi Titik ukur Sensor-actuator 4-WS 5.7 Manuver yang dipakai sebagai standar test drive 6.1 Respon Body 6.2 Respon Roda 6.3 Gaya Peredaman 6.4 Wheel Speed vs Vehicle Speed PID 6.5 Wheel Speed vs Vehicle Speed Fuzzy 6.6 Respon Spin Throttle 700 , 2500 RPM 6.7 Respon Spin Throttle 900 , 3000 RPM 6.8 Respon Putaran Mesin pada Throttle 700 , 2500 RPM 6.9 Respon Putaran Mesin pada Throttle 900 , 3000 RPM 6.10 Respon Rasio CVT pada Throttle 700, 2500 RPM 6.11 Respon Rasio CVT pada Throttle 900, 3000 RPM 6.12 Perbandingan Spin PID-Fuzzy vs NN 6.13 Error RPM PID-Fuzzy vs NN 6.14 Tanggapan Yaw rate dari ke mpat system kendali integrasi 6.15 Yaw rate diperbesar 20 kali (skala)
viii
76 80 93 94 95 103 105 106 108 109 110 112 120 120 121 123 123 124 124 125 125 125 126 127 127 128 129
DAFTAR
TABEL
Tabel 2.1 2.2 4.1 4.2 5.1 5.2 5.3 6.1 6.2 6.3 6.4 6.5 6.6
Halaman
Hasil Data Stopping Distance Komparasi Respon Waktu pompa ABS vs Pumpless Normalisasi Koefisien Bit Contoh Karakteristik Input-output yang diasumsikan Variasi Level Pengemudi 9 Kondisi Jalan Yang dipakai sebagai standar Test Drive Variasi TODS Respon Parameter Sistem Suspensi terhadap Frekuensi Jalan Ranking Kinerja Sistem Integrasi Hasil Training FONN Hasil Training VONN Hasil Testing Kinerja Kendali FOC, FONN, VONN Hasil Test Drive Robust, FONN, VONN
ix
24 28 76 81 112 113 114 122 130 132 132 135 136
DAFTAR SIMBOL
ij
= Bobot dari node i ke j
J
= Jacobian matrik
H
= Hessian matrik
= Laju pela pelatihan
=
Momentum
=
Signal error keluaran
=
Signal error sum
E
=
Error square
X
=
Variable input – output
g
=
Gradient
C
=
Center
U
=
Matrik bobot
d
=
Data posisi
J(M,C)
= Cost function
VSS s
=
YR s
=
Vehicle side slip reference Yaw Rate reference
RA s
=
Roll angle reference
VSS a
=
Vehicle side slip response
YR a
=
Yaw Rate response
RA a
=
Roll angle response
VSS
=
Error vehicle side slip
YR
=
Error yaw rate
x
RA
=
Error Roll angle
C
=
Konstanta stiffnes laterall
M
=
Massa tak berpegas
Ms
= Massa berpegas
h
=
Tinggi titik berat
I
=
Massa inersia
Ts
=
Jarak antar roda kiri ke kanan
=
As depan ke titik berat jarak
b
= As belakang ke titik berat jarak
Tvcd
=
Torsi kendali ESP
Mas
= Moment active suspension
r
= Sudut setir belakang
Kp
=
Konstanta Pegas
K
=
Konstanta Redaman
Vm
= Tegangan Motor Listrik 4 WS
I
m
= Arus Motor Listrik
Vv
= Tegangan Kecepatan Kendaraan
Vα
= Tegangan Percepatan Kendaraan
#E gas
= Posisi Derajat Bukaan Throttle
#F b(i)
= Gaya Rem masing-masing roda
IGN
= Sudut Timing Pengapian
INJ
= Time Delay Bukaan Injector
Fa
= Gaya Dorong Kendaraan
α
= Percepatan Kendaraan
ω
= Kecepatan Kendaraan
OF
= Object Function
xi
FF
= Fitness Function
DAFTAR SINGKATAN
TODS
Three in One Dynamics System
VSS
Vehicle Side Slip
YR
Yaw Rate
RA
Roll Angle
FOC
Fixed Order Control
FONN
Fixed Order Neural Network
VONN
Varied Order Neural Network
MDFC
Multi Dimension Fuzzy Clustering
ABC
Adaptive Backtrough Control
LSE
Least Square Error
NNP
Neural Network Plant
NNC
Neural Network Controller
TPS
Trhottle Position Sensor
NPSO
New Partical Swarm Optimazion
IASTED
International Assosiation Science and Technology for Development
ICP
Imperial College Press
BP
Backpropagation
IJCIA
International Journal Computer Intelligent and Application
EMS
Engine Management System
CMSC
Chassis Management and Safety Comfort
ITS
Intelligent Transportation System
IVN
In Vehicle Networking
AVEC
Advance Vehicle Control xii
ABS
Antilock Brake System
ESP
Electronics Stability Programm
EFI
Electronics Fuel Injection
AS
Active Suspension
4 WS
4 Wheel Steering
CBP
Constructive Back Propagation
CARB
California Assosiation Resource Boaad
xiii
