Design and Calibration of the Jaguar XK Adaptive Cruise Control System
Tim Jagger MathWorks International Automotive Conference 2006 JAGUAR XK
Page 3 ADAPTIVE CRUISE CONTROL(ACC) MODEL BASED CALIBRATION
• Adaptive Cruise Control – The Driver’s Perspective • The Adaptive Cruise Control System – Control Aspects • Jaguar XK Adaptive Cruise Control System • Jaguar XK Adaptive Cruise Control Virtual Vehicle Model • Model Correlation • Jaguar XK Adaptive Cruise Control Calibration Development • Queue Assist Development
Page 4 ADAPTIVE CRUISE CONTROL THE DRIVER’S PERSPECTIVE
• Driver Comfort Feature • Allows Host Vehicle to follow a target vehicle • Automatically Resumes to set speed in absence of target vehicle • Driver selects headway • Available Headways – 1.0s, 1.4s, 1.8s, 2.2s • Driver Selects Cruise Speed – Min. 20mph, Max. 112mph
Page 5 ADAPTIVE CRUISE CONTROL CONTROL APSECTS
Page 6 JAGUAR XK ADAPTIVE CRUISE CONTROL
• New Adaptive Cruise Control Algorithms for XK • New Features – Queue Assist Algorithms – Lead Vehicle Acceleration Algorithm (LVA) • Initial Functional Testing – April 2005 • Initial Calibration Testing – May 2005 • Unexpected Results ….
Page 7 APPROACH XK v S-TYPE
COMPARISON OF XK v S-TYPE ACC APPROACH TO LEAD VEHICLE TRAVELLING AT 50mph 2 XK shows 0 discontinuous reduction of range rate with range -2
-4
-6
-8
-10
-12
-14 S-TYPE shows smooth continuous reduction Range Rate m/s -16 of range rate with range
-18
-20 XK Headway 1 90mph XK Headway 1 100mph -22 S-TYPE Headway 1 90mph S-TYPE Headway 1 100mph -24
-26 0 20 40 60 80 100 120 140 Range m
Page 8 TEST & DEVELOPMENT
• Results from 1st ACC Test Session – Harsh Early Brake Intervention – LVA Algorithm gave unpredictable results – Generally unrefined ACC Performance • Expectation was that XK ACC would only need small scale refining – Developed from ACC3
Page 9 TEST & DEVELOPMENT
• In reality major development was required • Second test period was for validation NOT development • Tight timescales & challenging vehicle allocation
Page 10 ADAPTIVE CRUISE CONTROL VIRTUAL VEHICLE MODEL
Page 11 MODEL CORRELATION
X150 ACC-QA Simulink Model Validation 45 Cruising at Set Speed
Auto-Resume to Set Speed 40
35
Approach 1 Approach 3 30 Approach 2
ACC Host Vehicle Speed m/s 25 Following Target Vehicle
Test - SDV1143 Simulation - CADS-X150-CP39-SDB4-trj04h.mdl 20 0 20 40 60 80 100 120 Time s
Page 12 MODEL CORRELATION
X150 ACC-QA Simulink Model Validation 1.5 Residuals approach 1 Residuals approach 2 Residuals approach 3 1
0.5
0
-0.5 Range Rate Error m/s
-1 Test Vehicle - X150 SDV1143 Simulation - CADS-X150-CP39-SDB4-trj04h.mdl -1.5 0 20 40 60 80 100 120 Range m
Page 13 CALIBRATION DEVELOPMENT
• Test Results indicated seven parameters influenced ACC refinement – Lead Vehicle Acceleration Algorithm (LVA) – ACC brake gain – Brake demand slew rate – Time to collision gain – Residual Distance – Coast Deceleration Estimate – Headway Control Maps
Page 14 CALIBRATION DEVELOPMENT
• Test Matrix of 8 possible ACC configurations • 4 days per configuration to test on vehicle = 32 days vehicle testing • Complete test matrix was evaluated by simulation in 10 days
¾ ACC Vehicle Simulation was vital for calibration development
Page 15 SIMULATION PREDICTIONS V TEST RESULTS
Headway 1 Approach Results. Test v Simulation Predictions 5 5 Test - 90mph Test - 80mph Simulation - 90mph Simulation - 80mph 0 0 -5 -5 -10 -10 -15 Range Rate m/s Range Rate m/s
-15 -20 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Range m Range m
5 5 Test - 100mph Test - 112mph 0 Simulation - 100mph 0 Simulation - 112mph -5 -5 -10 -10 -15 -15 -20
Range Rate m/s Rate Range -20 m/s Rate Range -25 -25 -30 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Range m Range m
Page 16 COMPARISON OF SIMULATION PREDICTIONS WITH PRODUCTION CALIBRATIONS
• Two calibrations changed from simulation to production • Changes based on subjective driver assessment
Page 17 QUEUE ASSIST
• Extension of ACC operating range to 0mph • Headway opened significantly at low speeds • QA host vehicle unable to track target vehicle accelerating at a moderate rate • Filtering of ACC Speed demand introduced significant lag
Page 18 QUEUE ASSIST
Jaguar XK QA Simulation. Acceleration from 5mph to 30mph. Jaguar XK QA Simulation. Acceleration from 5mph to 30mph. 25 70
Headway 1 0.1g. Baseline Calibration Lead Car Speed 60 Headway 1 0.1g. Baseline Calibration Follow Speed 20 Headway 1 0.1g. Baseline Calibration Vehicle Speed 50
15 40
Range m 30 10 m/s Speed
20
5 10 Headway 1 0.1g. Baseline Calibration
0 0 10 20 30 40 50 60 70 80 10 20 30 40 50 60 70 80 Time s Time s
Page 19 FILTER CUT-OFF FREQUENCY
•Use frequency response toolbox to re-calculate filter constant
FREQUENCY RESPONSE OF THE FOLLOW SPEED FILTER R44 CALIBRATION. FREQUENCY RESPONSE OF FOLLOW SPEED FILTER - DESIGN 02.
0 0
-1 -1
System: follow_speed_filt_current -2 -2 Frequency (Hz): 0.151 Magnitude (dB): -3 -3 -3 System: follow_speed_filt_new_02 -4
Magnitude (dB) Frequency (Hz): 0.979 Magnitude (dB)Magnitude -4 Magnitude (dB): -3 -5
0 0 -5 System: follow_speed_filt_current -10 -30 Frequency (Hz): 0.15 Phase (deg): -42 -15 -20
Phase (deg) System: follow_speed_filt_new_02 -60 Phase (deg) -25 Frequency (Hz): 0.979 -30 Phase (deg): -29.9 -2 -1 0 -35 10 10 10 -1 0 10 10 FREQUENCY Hz Frequency Hz
Page 20 EFFECT OF NEW FILTER CONSTANT
X150 QA Simulation. Headway 1. Acceleration from 5mph to 30mph. Jaguar XK QA Simulation. Acceleration from 5mph to 30mph Follow Speed Filter -3db @ 1Hz 25 70
Headway 1 0.1g CADS Cal. R44. Follow Speed Filter -3dB @ 1Hz Lead Car Speed Headway 1 0.1g CADS Cal. R44. Follow Speed Filter -3dB @ 1Hz Follow Speed 60 20 Headway 1 0.1g CADS Cal. R44. Follow Speed Filter -3dB @ 1Hz Vehicle Speed
50
15 40
Range m 30 Speed m/s 10
20
5 10 Baseline Calibration filter cut-off @ 0.15Hz Filter cut-off @ 1Hz 0 0 10 20 30 40 50 60 70 10 20 30 40 50 60 70 80 Time s Time s
Page 21 JAGUAR XK MODEL BASED DESIGN AND CALIBRATION
SUMMARY • Jaguar XK ACC system modelled in simulink • ACC Model shows good correlation with test results • Model was used to develop Jaguar XK ACC Calibrations • Development time per calibration loop reduced from 30 days to 10 days (Typically 3 - 4 calibration loops on a vehicle program) • Vehicle test time now used for validation - not experimentation
Page 22