control system
Vinod Cherian, The MathWorks - July 03, 2008
The ongoing trend in automotive electronics to incorporate active safety systems has led automobile
manufacturers to incorporate antirollover capabilities into vehicle chassis control systems, such as
those for antilock brakes and traction control. This trend is driven by the National Highway Traffic
Safety Administration (NHTSA), which mandates that all 2011 model year vehicles and beyond have
antirollover controllers.
This requirement is based on NHTSA analysis of accident data from rollover crashes. For example,
in 2001, according to NHTSA's National Center for Statistics and Analysis, 10,138 people died in
rollover crashes, representing 32% of occupant fatalities for the year. Implementing active safety
mechanisms to lower the risk of vehicle rollover could potentially reduce fatalities.
One method of lowering the risk of rollover is to implement an Electronic Stability Control (ESC)
that applies differential braking based on measured and estimated vehicle states. This article
highlights the use of model-based design to develop and automatically optimize an ESC for sportutility vehicles (SUVs), those with the greatest rollover risk.
Vehicle and controller model
A central concept in model-based design is the executable specification, or model, that describes the
dynamic behavior of the system. A validated model of the automobile, in this case a high-fidelity
model of an SUV, can be leveraged to achieve a significant savings in the controller development
cost and time. Numerical simulation of the model can be used to study the vehicle response to
various steering maneuvers and these tests can be repeated easily while varying parameters like
road surfaces, tire models, and vehicle properties. Additionally, the models can be used in the
development and verification of the embedded control system.
The vehicle used in this article is representative of a midsize SUV. The vehicle model is available in
CarSim, a commercial off-the-shelf vehicle dynamics simulation tool. The vehicle model
performance has been verified against test data and is suitable for simulating vehicle response
under significant roll motions. The model has dual independent front suspension and a solid rear
axle that supports the sprung mass. It is a nonlinear mathematical model with degrees of freedom
for the sprung mass, each axle, each wheel, steering, and braking system, and can be customized
using different vehicle parameters as well as road and environmental conditions.
View a full-size image Signals (left) fed to the signal constraint blocks vary the roll and sliprate signals (right) as the optimization progresses. The yellow area represents the
inadmissible range of signal values.Each signal constraint defines piecewise linear upper and
lower bounds on the signal. During optimization the controller gains are adjusted and the simulation
is rerun in an iterative loop until the simulated signals satisfy the specified bounds or the
optimization routine fails to solve the problem. The above figure shows the variation of the roll
signal and the slip-rate signal as the optimization algorithm iterates to a solution. In solving this
feasibility problem, the optimization algorithm computes the maximum signed distance of the signal
being constrained to each edge of the piecewise linear bound. Typically, a negative value indicates
that the constraint is satisfied. The optimization algorithm uses the signed distance to each edge to
update the controller parameters. The optimization algorithm constructs the optimization problem
independently of the numerical solver used to compute the states of the system. Gradient- or
nongradient-based methods, such as genetic algorithms, can be used. In this case, given the
switching nature of the controller and consequent nonsmooth behavior, gradient-based solvers are
less likely to find a global solution. As a result, a pattern search algorithm is used. In practice, it is
smart to switch among a few different types of optimization methods to ensure that the optimization
algorithm finds a global extremum and to rule out convergence to local minima of the cost function.
Controller verification and performance validation
The figure below shows a visual representation of the performance of the optimized ESC eliminating
the rollover in the vehicle. The red vehicle rolling over has no controller; the blue vehicle has an
optimized controller. Through this simulation, we have demonstrated controller design that
eliminates SUV rollover, thereby greatly reducing the amount of on-track tuning and complete
reliance on testing with a physical vehicle.