Rollover Warning/Control for Sports Utility Vehicles Bo-Chiuan Chen, GSRA Huei Peng, Associate Prof. Dept. of Mechanical Engineering and Applied Mechanics University of Michigan 1
Outline Introduction TTR Warning/Anti-Rollover Control TTR Robustness Study Human-in-the-Loop Evaluation of the TTR Based Anti- Rollover Control Conclusion 2
Rollover Fatality Rates All Vehicles Utility Vehicle 47 98 Small Pickup 93 Standard Pickup 59 Small Car 47 Minivan 4 Standard Van 34 Medium Car 28 Large Car 19 Source: 1991-94 Average/Annual deaths per million registered vehicles provided by 3
Why SUVs Roll Over SUV High center of gravity: Sideways force can cause rollover. SEDAN Low center of gravity: Resists rollover. Similar width between wheels 4
Outline Introduction TTR Warning/Anti-Rollover Control TTR Robustness Study Human-in-the-Loop Evaluation of the TTR Based Anti- Rollover Control Conclusion 5
Time To Rollover (TTR) TTR: Index defined to assess rollover threat in this research. Steering angle Vehicle Vehicle Model Model Future Roll angle Roll Roll angle angle exceeds threshold? Yes TTR now Yes No Time Time < X secs? secs? No TTR = X secs 6
Desired TTR For rollover cases, a straight line with slope= -1 is created starting from the rollover point backwards in time. It gives a uniform countdown toward rollover threats and thus serves well as the basis of warning/control. TTR=.5 The vehicle will rollover.5 sec later. 7
3.5 3 Neural Network (NN) Architecture 2.5 2 1.5 1.5 3. 5 3 -.5 2 4 6 8 1 12 time (seconds) 2. 5 TTR TTRfrom Simplified Model Model Roll Roll Angle Angle -.5 2 1. 5 1. 5 2 4 6 8 1 12 output tim e (se cond s ) NN NN TTR TTR Desired TTR - + Change of of Roll Roll Angle Angle inputs Layer 1 Layer 2 Train NN to approach the Desired TTR 8
TTR Based Anti-Rollover Control Controller < Reference TTR Steering Input (Disturbance) K Differential TruckSim Braking Cherokee Braking Moment TTR Calculation Lateral Acceleration TTR 9
Previous Studies ARC 1998 ARC 1999 Rollover Warning of An Articulated Heavy Truck Rollover Warning and Control of Sport Utility Vehicles 1
Ramp steering ARC 1998 Entering ramp Obstacle avoidance TTR (sec) 3 2 1 5 1 Time (sec) 3 TTR (sec) NN TTR Model TTR 2 1 Worst-case 2 1 1 1 time (sec) 2 2 time (sec) 4 TTR (sec)3 2 TTR (sec)3 Without Worse Case Training TTR (sec) 3 2 1.5 1 1.5 2 2.5 time (sec).5 1 1.5 2 2.5 time (sec) 11
Newton 1 8 6 4 Fz of the left front tire ARC 1999 Newton 1 8 6 4 Fz of the right front tire No control P-control 2 2 2 4 6 8 sec 2 4 6 8 sec 1 Fz of the left rear tire 1 Fz of the right rear tire 8 8 Newton 6 4 Newton 6 4 2 2 2 4 6 8 sec 2 4 6 8 sec 12
Outline Introduction TTR Warning/Anti-Rollover Control TTR Robustness Study Human-in-the-Loop Evaluation of the TTR Based Anti- Rollover Control Conclusion 13
TTR Robustness Study Ramp radius & Superelevation angle: (, o ) vs. (1m, 1 o ) Vehicle load & C.G. Height: Empty vs. Gross (45% more of the empty weight) Tire pressure: 33 psi vs. 16.5 psi Road friction:.85 vs..3 Speed: 3 mph vs. 6 mph Sensor noise: - Steering, longitudinal speed, lateral accel., roll angle - Noisy vs. Clean measurement 14
Superelevation Speed = 5 mph Radius = 1 m Sup. = 1 o Lack of the superelevation info. 15
Gross Load Condition.5.45 Misdetection (False negative reading) Simple TTR TruckSim TTR.4.35.3 Simple TTR TTR (sec).25.2.15 TruckSim TTR.1.5.5.1.15.2.25.3.35.4.45.5 time (sec) 16
Tire Pressure vs. Tire Stiffness Longitudinal Force Lateral Force Longitudinal Force, F x (N) 1 9 8 7 6 5 4 3 2 1 Lateral Force, F y (N) 8 7 6 5 4 3 2 1 Aligning Moment, M z (N-m).5.1.15.2.25.3.35.4.45.5 Slip ratio, λ 3 25 2 15 1 5 Aligning Moment 33 psi 16.5 psi stiffness (N/mm) 5 1 15 Slip angle, α (deg) 25 2 15 1 5 Vertical Stiffness 5 1 15 Slip angle, α (deg) 5 1 15 2 25 3 35 tire pressure (psi) 17
Tire Pressure Effect Nominal pressure: 33 psi Low pressure:16.5 psi.5.5.45.45.4.35 TruckSim TTR.4.35.3.3.25.25.2.2.15.15.1.5 Simple TTR.5.1.15.2.25.3.35.4.45.5.1.5.5.1.15.2.25.3.35.4.45.5 18
Road Friction µ =.85 µ =. 3.5.5.45.4.35 TruckSim TTR.45.3.4.25.2.35.15.1.5 Simple TTR.5.1.15.2.25.3.35.4.45.5.3.25.5.1.15.2.25.3.35.4.45.5 False Alarm (False positive reading) 19
Speed 3 mph 6 mph.5.45 Simple TTR TruckSim TTR.5.45 Simple TTR TruckSim TTR.4.35 TruckSim TTR.4.35.3.3 TTR (sec).25.2 TTR (sec).25.2.15.15.1.5 Simple TTR.1.5.5.1.15.2.25.3.35.4.45.5 time (sec).5.1.15.2.25.3.35.4.45.5 time (sec) Gain-scheduling works well for the test cases (step steering). 2
Sensor Noise No Sensor Noise With Sensor Noise.5.45 Simple TTR TruckSim TTR.5.45 Simple TTR TruckSim TTR.4.35 TruckSim TTR.4.35.3.3 TTR (sec).25.2 TTR (sec).25.2.15.15.1.5 Simple TTR.1.5.5.1.15.2.25.3.35.4.45.5 time (sec).5.1.15.2.25.3.35.4.45.5 time (sec) 21
Robustness Solution Low tire pressure and speed are not significant for the test cases (step steering). Sensor noise can be handled by low-pass filters. Superelevation, road friction, and gross load condition can be handled by the adaptive TTR algorithm. 22
Low-Pass Filter Without Filter With Filter.6 Simple TTR TruckSim TTR.6 Simple TTR TruckSim TTR.5.5.4.4 TTR (sec).3.2 TTR (sec).3.2 Simple TTR.1.1 -.1.1.2.3.4.5.6.7.8.9 1 time (sec) -.1.1.2.3.4.5.6.7.8.9 1 time (sec) TruckSim TTR T delay Depends on the signal offset and the filter cut-off frequency. 23
Adaptive TTR Algorithm : Gain-scheduling for gross load condition Steering angle Yaw Model 1 Lateral acceleration + + + - ˆ) K(µ Roll Model Roll angle Roll rate Braking torque Yaw Model 2 Coupling Gain 1 Yaw acceleration Coupling Function â y Coupling Gain 2 (for initial conditions) µˆ Adaptive TTR algorithm aˆ y _superlelvation a y _sensor 24
Friction Adaptation µ =.3 µ =. 5 1.5 Simple TTR TruckSim TTR.5.45 Simple TTR TruckSim TTR.4 1.35 Simple TTR.3 TTR (sec).5 TTR (sec).25.2.15.1.5 TruckSim TTR -.5.5.1.15.2.25.3.35.4.45.5 time (sec).5.1.15.2.25.3.35.4.45.5 time (sec) 25
Superelevation 15 8 Steering (deg) 1 5 Roll (deg) 6 4 2-5 4.5 5 5.5 6 6.5-2 4.5 5 5.5 6 6.5 Time (sec) 1.5 Lateral accel. (g).5 TTR (sec).4.3.2.1 Simple TTR TruckSim TTR -.5 4.5 5 5.5 6 6.5 4.5 5 5.5 6 6.5 Time (sec) 26
Gross Load Condition Without Gain-Scheduling With Gain-Scheduling.5.5 Simple TTR Simple TTR.45 TruckSim TTR.45 TruckSim TTR.4.4.35.35 TruckSim TTR.3.3 TTR (sec).25.2 TTR (sec).25.2.15.15.1.1.5.5.1.15.2.25.3.35.4.45.5 time (sec).5 Simple TTR.5.1.15.2.25.3.35.4.45.5 time (sec) 27
Outline Introduction TTR Warning/Anti-Rollover Control TTR Robustness Study Human-in in-the-loop Evaluation of the TTR Based Anti- Rollover Control Conclusion 28
SGI Onyx2 Graphics Supercomputer VVSS Graphics Driving Simulator Setup Dell Inspiron 5 PIII 6Mhz Notebook TruckSim Dynamics Microsoft Force Feedback Steering Wheel 29
Calspan - MOTS Driving Course 1 4 2 3 3
1 Large Radius Arc Inner Radius = 2 ft Outer Radius = 212 ft Lane width = 12 ft 31
2 Avoidance Maneuver Unit: ft Obstacles 32
3 Small Radius Arc 152 14 Inner Radius = 14 ft Outer Radius = 152 ft Lane width = 12 ft 33
4 Fishhook Maneuver 75 15 15 1 st Radius = 75 ft 2 nd Radius = 15 ft Lane width = 12 ft 34
Human-in in-the-loop Evaluation Each test subject is required to drive 3 vehicles. - No control. - Control with differential braking. - Control with just braking on the front axle. Each vehicle is driven through 4 MOTS scenarios. 18 people participated the evaluation. The types of vehicles are not revealed until the end of the experiment. Because of the drivability concern drivability concern, the braking torque is limited at 25 N-m for differential braking and 125 N-m for both front wheels of the front braking. 35
Drivability Drivers use lateral velocity and yaw rate for the steering decision. 36
1 Differential Braking Evaluation Results Larger Fz_min Front-axle Braking 5 success rate (%)* 1 5 1 1 2 Smaller Roll_max 1 2 Smaller Ay_max Large Radius Arc Avoidance Maneuver Small Radius Arc 5 1 2 Fishhook Maneuver * success rate (%) = # of success 18 tests 37
1 Differential Braking Verification Results Larger Fz_min Front-axle Braking 5 success rate (%)* 1 5 1 1 2 Smaller Roll_max 1 2 Smaller Ay_max Large Radius Arc Avoidance Maneuver Small Radius Arc 5 1 2 # of success * success rate (%) = 18 tests Fishhook Maneuver 38
Human-in in-the-loop The controller was designed and verified under the given step steering and fishhook steering. BUT not with human- in-the the-loop. 12 1 8 6 4 2 1 2 3 4 5 6 7 8 sec 6 4 2-2 -4-6 -8-1 -12 1 2 3 4 5 6 7 8 sec Human close the loop of the vehicle and the environment. Drivers might feel the vehicle equipped with the antirollover control difficult to drive and thus degrade its performance. Driving Simulator can be used to verify the control performance and collecting driving patterns for controller design. 2 15 1 5-5 -1-15 5 1 15 2 25 3 35 39
Outline Introduction TTR Warning/Anti-Rollover Control TTR Robustness Study Human-in-the-Loop Evaluation of the TTR Based Anti- Rollover Control Conclusion 4
Conclusion An Adaptive TTR algorithm was proposed to enhance the TTR robustness against the variation of the superelevation, road friction, and the gross load condition. Anti-rollover control did not perform well with human-in in- the-loop evaluation. We will design a more complicated or better tuned controller. Active safety device should be designed and evaluated designed and evaluated with human-in-the-loop to guarantee the performance. 41