Page 32 - ASME DSCC 2015 Program

Technical Program

Hierarchical Hybrid Predictive Control of an Autonomous Road Vehicle
ConTRIBuTED SESSIon
1-28-2 WM6 Vehicle Dynamics Control 2 Contributed regular paper. DSCC2015-9773
Emerson Burkhart B 1:30pm–3:30pm Qian Wang, Thomas Weiskircher, Beshah Ayalew, Clemson University,
Greenville, SC, United States
Session Chair: Robert D. Gregg, University of Texas at Dallas This paper presents a hierarchical hybrid predictive control framework for an
Session Co-Chair: Punit Tulpule, The Ohio State University autonomously controlled road vehicle. At the top, an assigner module is de-
signed as a finite state machine for decision-making. Based on the current
An Adaptive and fast Control Strategy for Antilock Braking System
information of the controlled vehicle and its environment (obstacles, and
Contributed regular paper. DSCC2015-9779
lane markings, etc), the assigner selects discrete maneuver states through
Sadegh Tajeddin, Mohit Batra, nasser lashgarian Azad, John McPhee, pre-defined switching rules. The several maneuver states are related to
Roydon fraser, University of Waterloo, Waterloo, ON, Canada
different setups for the underlying model predictive trajectory guidance
After more than 30 years since the Antilock Braking System (ABS) was first module. The guidance module uses a reduced-order curvilinear particle
introduced, it has become the most important active safety system used on motion description of the controlled vehicle and obstacle objects as well as
passenger cars. However, it is hard to find a precise description of ABS, its a corresponding description of the reference path, lane and traffic limits. The
stability and performance in the literature. Most of ABS algorithms currently output of the guidance module interfaces with the lower level controller of
used are not adaptive to changes of road friction conditions. The aim of our the continuous vehicle dynamics. The performance of the proposed frame-
work is to provide a new ABS algorithm that is adaptive to changes of road work is demonstrated via simulations of highway-driving scenarios.
conditions. To this end, an online parameter estimator is designed to esti- Dynamic Rear-End Collision Mitigation for a Vehicle About to be Struck
mate the road characteristics and maximum possible deceleration. Then, a Contributed regular paper. DSCC2015-9674
driver demand regulator is proposed to limit the demanded deceleration to
Craig E. Beal, luke Giugliano, Bucknell University, Lewisburg, PA, United
the maximum values. In this new strategy, road characteristics are estimated
States
prior to the braking, not during the braking which makes it fast and adaptive.
The proposed ABS algorithm is simulated on an artificial driving track and Increases in sensing and computing capability in modern vehicles create op-
simulation results have been compared to a simple non-adaptive 6-phase portunities for improving safety through active control of the vehicle motion.
Bosch ABS algorithm as our benchmark that is based on deceleration A system for mitigating rear-end collisions is introduced, focusing on the
thresholds. Results show a better braking performance and more than 30% strategies a vehicle about to be struck could employ to reduce the severity
of reduction in braking distance. of the crash. Due to high differential speeds and tight space constraints, the
timing and precision of control inputs are critical, preventing human drivers
optimal Switching in Anti-lock Brake Systems of Ground Vehicles
from effectively employing these strategies. However, the results of the
Based on Approximate Dynamic Programming
study presented in this paper suggest that there are relatively unobtrusive
Contributed regular paper. DSCC2015-9893
strategies that a single vehicle with some limited autonomy could use in join-
Tohid Sardarmehni, Ali Heydari, South Dakota School of Mines and ing a roadway queue that would improve the safety of the occupants.
Technology, Rapid City, SD, United States
development of a Miniaturized autonomous Vehicle: Modification of a
Approximate dynamic programming, also known as reinforcement learning,
1:18 Scale RC Car for Autonomous operation
is applied for optimal control of Anti-lock Brake Systems (ABS) in ground
Contributed regular paper. DSCC2015-9799
vehicles. As an accurate and control oriented model of the brake system,
Dwarkesh Iyengar, Diane l Peters, Kettering University, Flint, MI, United
quarter vehicle model with hydraulic brake system is selected. Due to the
States
switching nature of hydraulic brake system of ABS, an optimal switching
solution is generated through minimizing a performance index that penalizes Development of a Miniaturized Autonomous Vehicle: Modification of a 1:18
the braking distance and forces the vehicle velocity to go to zero, while pre- Scale RC Car for Autonomous Operation
venting wheel lock-ups. Towards this objective, a value iteration algorithm is
selected for ‘learning’ the infinite horizon solution. Artificial neural networks,
as powerful function approximators, are utilized for approximating the value
function. The training is conducted offline using least squares. Once trained,
the converged neural network is used for determining optimal decisions
for the actuators on the fly. Numerical simulations show that this approach
is very promising while having low real-time computational burden, hence,
outperforms many existing solutions in the literature.

32