Lateral Control of Heavy Vehicles for Automated Systems

This is the final report for MOU 313, "Lateral Control of Heavy Vehicles for Automated Highway Systems". It address the following task items: Analysis and synthesis of control algorithms based on system identification and calibration results (Task 4. in the MPU), Closed loop results at highway speeds (Task 5), limits and persormance of sensors, actuators and control schemes as infered from data collected during closed loop experiments at 55 MPH (task 6) and baseline safety requirements based on closed loop data (task 7) Five controllers were designed based on the model identification study conducted on the previous phase of the project. This report will present the details of eash design methodology: Nonlinear Robust Loop Shaping, Robust controller based on Feedback linearization, Gain -Scheduled H 1 , Robust H 1 and Linear Parameter Varying (LPV) controllers. The sysnthesis of each controller is sufficiently complex to merit a separate Chapter. These Chapters address Task 4 in detail. Closed loop experiments were performed for Robust H 1 , nonlinear Sliding Mode Control (SMC) and LPV contorller. The results of these closed loop experiments will be presented in the Chapters that detail the control design respectively and address Task 5. One of the control strategies, Linear Parameter Varying (LPV) was developed to address issues arising from variation in dynamics with the speed of the vehicle. Data collected from Robust H 1 controller, which is designed for fixed speed, indicated that an LPV control strategy may work better for good steering control over a wide range of speeds. The gain scheduled H 1 controller was also designed to address automated operation over a range of velocity. The closed loop experiment results showed that the sensors (range, sensitivity and noise characteristics) on the experimental vehicle were adequate for effective automated steering at 55 MPH. Lateral deviation measurement sensors were adequate in range and sensitivity. Yaw rate and acceleration sensors were adequate for system identification but inadequate for effec-tive implementation of nonlinear control strategies. Actuator performance was adequate (with bandwidth in excess of 10 Hz) for 400m step curvature change at 55 MPH. Performance of the closed loop controller as characterised by tracking accuracy at high speeds has been excellent. The LPV controller has been tested up to 70MPH. The limited length of track at Crow's Land-ing test site and large inertia of the truck semi-trailer vehicle posed limitations on testing the controllers to the "limit". However, the authors are pleased to report that the LPV controller provided smooth steering action with adequate tracking performance from start of the track un-der maximum possible acceleration to the end of track with aggressive deceleration (occasional brake locking). The curve transitions at Crow's Landing test site are aggressive and unlikely to exist at real highways. In that sense, the range of present sensors and performance of present controller and actuator was tested under limiting conditions and configuration on the test vehicle was deemed adequate. These conclusions addresses Task 6. LPV model based controller emerged to be the right "supervisory controller" which com-bines information from longitudinal sensor (velocity) and imposes restriction on the longitudinal controller in terms of maximum allowed deceleration and maximum allowed aceleration. This addresses Task 7 of MOU 313. A video of the controllers was taken at the Crow's Landing test site and is available at PATH Publications office and on the PATH web site. A copy is included with this report. Report is organized in Chapters. First two Chapters presents Nonlinear Controllers which have sound robustness properties but are sensitive to sensor noise. Subsequent Chapters present Robust H 1 contoller and Gain Scheduled Controller. Finally the last Chapter will propose LPV controller as a method to combine the benefits of converting the vehicle model nonlinear in velocity to a linear time varying model with velocity as the measured parameter and utilizing rich tool set of Linear H 1 based control design. Each Chapter is self sufficient in content. This allows reader to read Chapters in any order or preference. Conclusions presented in the last Chapter tie the lessons from each different controller.