Modification of IPG Driver for Road Robustness Applications Alexander Shawyer (BEng, MSc) Alex Bean (BEng, CEng. IMechE) SCS Analysis & Virtual Tools, Braking Development Jaguar Land Rover
Introduction Presentation Contents 1. Introduction Virtual Engineering at JLR Stability Controls Development 2. Modelling Strategy Hypothesis Vehicle Model Road Model Analysis Methods 3. Results Standard Driver Rally Driver Parameterisation 4. Conclusions & Further Work 2
Introduction Virtual Engineering at JLR JLR is investing significant effort & resource into developing virtual engineering capability across Product Development Efficiency facilitates earlier engineering & decisions Robustness increased design space & test scenario evaluation Cost - reduced prototype fleet & physical testing Stability & ABS functions development have traditionally been very vehicle intensive activities, involving significant overseas tests trips. Chassis Engineering Brakes Design SCS Functions Systems Braking Development Stability Attribute Applications Virtual Tools SCS Analysis 3
Introduction SCS Robustness Testing Static & dynamic vehicle property calibration Braking, traction, yaw, & roll stability tuning The SCS calibration development process High Mu Medium Mu Low Mu Off road (Land Rovers) Calibration robustness & threshold consumption Tests to check for pump duty cycle & false interventions typically public road routes Roll Stability Control function = Covara, Italy Robustness & Validation testing Simulation Models: Vehicle Controller Road Driver 4
Modelling Strategy Hypothesis & Method Hypothesis The IPG Driver model with suitable parameterisation could be applicable to various SCS applications. More specifically; correlating the IPG driver to real driver data would enable virtual Road Robustness testing. Focusing on the general Driver behaviour such as G-G Diagrams, time intensities and peak accelerations we can optimise the Driver model to better represent real driver performance. Sensitivity Analysis A sensitivity analysis is performed to identify the key driver parameters that have the greatest influence on the driver performance. Standard Driver Each Parameter is swept through a range of +/-15% at intervals of 1%. Rally Driver Side Slip & Brake Slip Coefficient: 0.5 6.0 @ 1.0 intervals. 5
Modelling Strategy Vehicle Model: How Confidence is Obtained Correlation (Static) K & C correlation. Bump Steer, Roll Centre Height, Track Change & Wheelbase Change. Correlation (Dynamic). Constant Radius. Understeer Gradient. Frequency Response. Roll, Pitch & Yaw Frequency Other Tyres scaled correctly for Asphalt: LKY: Scaling factor for cornering stiffness. LMUY: Scaling factor for peak lateral friction. 6
Modelling Strategy Static Correlation: Vehicle Model Exp CarMaker Front Mass [kg] 1197 1196 Rear Mass [kg] 799 800 Total Mass [kg] 1996 1996 Average steering ratio 14.8 - Wheelbase [m] 2.660 2.662 Tyres pressure Exp CarMaker Front [bar] 2.48 2.48 Rear [bar] 2.20 2.20 Toe Exp CarMaker Front left [deg] 0.12 0.13 Front right [deg] 0.12 0.13 Rear left [deg] 0.07-0.01 rear right [deg] 0.07-0.01 Camber Exp CarMaker Front left [deg] -0.48-0.41 Front right [deg] -0.48-0.41 Rear left [deg] -1.16-1.13 rear right [deg] -1.16-1.13 7
Modelling Strategy Road Model Generation Using a JLR in-house Road Builder tool; GPS data from the Covara test route was processed and converted into a.road file. 8
Modelling Strategy Segment Selection This segment has been selected for this analysis because it offers a sufficient mix of corners to generate the necessary Acceleration range to effectively optimise the Driver Model. 9
Modelling Strategy Analysis Methods To assess improvements in the driver model, the following plots will be used. These plots are designed to show the Driver Character. The primary concern is developing a driver model that can be utilised in a wider context as opposed to a specific use case. Time History Time Intensity GG diagram (Density Plot) 10
Lateral Acceleration [ms-2] Lateral Acceleration [ms-2] Lateral Acceleration [ms-2] Lateral Acceleration [ms-2] Results Comparison of IPG Driver to Experimental Data Experimental Data Defensive Longitudinal Acceleration [ms-2] Longitudinal Acceleration [ms-2] Normal Aggressive Longitudinal Acceleration [ms-2] Longitudinal Acceleration [ms-2] 11
Lateral Acceleration [ms -2 ] Lateral Acceleration [ms -2 ] Results Sensitivity of IPG Rally Driver to Parameters Side Slip Coeff Brake Slip Coeff Ax [ms-2] Ay [ms-2] Car V [kph] Yaw Rate [degs-1] Time [s] Distance [m] Side Slip Angle [deg] 0.5 0.5-5.943-4.874 130.13-24.102 59.929 571.904-1.7241 6 6-5.944-4.874 130.18-24.102 59.929 571.904-1.724 The initial testing of the Rally driver model showed little difference over the standard Driver model. Reasons; Rally Driver not intended for this scenario (4WD vehicle model and higher friction surface). SS Coeff = 0.5, BS Coeff = 0.5 SS Coeff = 6.0, BS Coeff = 6.0 Long Acceleration [ms -2 ] Long Acceleration [ms -2 ] 12
Analysis Modifications to the Standard Driver 13
Analysis Fine tuning of the Additional Driver Parameters Utilising all the available driver parameters resulted in and unstable behaviour of the driver model. To improve stability of the model just two parameters are chosen to fine tune the driver model: Long.AccuarcyCoef = 0.95 Long.SmootCoef = 0.75 14
Lateral Acceleration [g] Lateral Acceleration [g] Analysis Current State of the Driver Model Correlation G-G Diagram Correlation. Longitudinal Acceleration ms-2] Longitudinal Acceleration [g] Time [s] Distance [m] Min Long Acc [ms-2] Max Long Acc [ms-2] Min Long Acc [ms-2] Max Long Acc [ms-2] 67.31 595.2058-3.4335 2.6487-8.9271 7.848 43.5 478.715-3.1261 2.421-6.2787 7.1264 15
Conclusions Standard driver Aggressive model best represented real test driver data but significant parameter tuning required to obtain best correlation:- Max Longitudinal & Lateral Acceleration, Min Long Acceleration & Corner Cutting Coefficient. Apex Shift Coefficient, Corner Roundness Coefficient, Throttle Accuracy & Smoothness Rally Driver not effective for this Road Robustness scenario Designed for steady state drift scenario Actual usecase was 4WD high mu surface (i.e. no high slip angles) Parameterisation of driver model to produce a similar G-G plot is the biggest challenge 16
Further work Working with IPG to develop standard procedure for parameterising driver model to experimental data Further evaluate Rally Driver for low mu test scenarios Improve road measurement process for road model creation Consider use of steering torque as a feedback loop to improve representation of real driver Ultimate Goal A set of driver models to represent typical US, European & Chinese drivers & SCS Development Engineers! 17
Alexander Shawyer. SCS Analysis, Vehicle Characterisation. ashawyer@jaguarlandrover.com Alex Bean. Technical Specialist, SCS Analysis. & Virtual Tools abean@jaguarlandrover.com