Malcolm Smith, Will Hoult and Panos Brezas
View gallery »
Following its 2017 launch at the Geneva International Motor Show, the McLaren 720S has received adulation and awards, garnering particular praise for its ride quality, handling and driver engagement. The heart of the suspension control system, called Proactive Chassis Control II, was the result of a collaboration between McLaren Automotive and the University of Cambridge.
How does this system deliver a leap forward in ride and handling behavior? Engineers who led its development offered Automotive Engineering insights into their approach and the method they employed for its implementation.
The Proactive Chassis Control II suspension system features hydraulic cross-linking (lead image) and “semi-active” dampers. The dampers’ rates are continuously adjustable by means of a needle valve and solenoid which are controlled by a computer in real time (Fig. 2). Although semi-active systems are not new in themselves, there has been a big advance in performance through use of a new algorithm developed in a PhD project at the University of Cambridge to simultaneously optimize the car's ride and handling response.
The ride behavior is the car's response to undulations in the road, whereas the handling behavior is the response to driver inputs such as steering, accelerating and braking. These can be viewed as two types of exogenous (external) input acting on the vehicle. It is the purpose of the algorithm to control the car under arbitrary and simultaneous excitation from both types of input.
Solving the control algorithm
The first challenge in developing such a system is that the ride and handling inputs have different character and affect the vehicle in different ways. The road inputs were modeled stochastically (i.e., in a statistical manner) taking account of typical road profiles. The driver inputs were treated deterministically. A simplifying modelling assumption was adopted to consider the driver inputs as inertial loads acting on the sprung mass to be estimated in real time.The second challenge was to pose and solve a suitable stochastic optimal control problem. A performance measure was selected consisting of a weighted average of squared quantities (accelerations, velocities, suspension and tire deflections of the vehicle). Minimization of this measure would result in control of the quantities relevant for ride comfort, tire grip and body control for both exogenous inputs at once.
The main difficulty in solving the optimal control algorithm is that the control input (the adjustable damper rate) enters the problem non-linearly. The force provided by the damper obeys a relationship of the form F(t) = c(t)v(t) where c(t) is the adjustable damper rate, which can be varied in real time between a minimum and maximum value, and v(t) is the velocity of the piston. This means that the control input multiplies a system state rather than entering the dynamics additively—which makes the resulting Hamilton-Jacobi-Bellman equation complicated to solve.
Nevertheless, it proved possible to derive a control law in the form of a nonlinear static map (see “A clipped-optimal control algorithm for semi-active vehicle suspensions: theory and experimental evaluation,” by P. Brezas, M.C. Smith and W. Hoult, Automatica, vol. 53, 2015, p.188-194).
The third challenge arose because the algorithm relied on “state feedback,” but not all components of the car's state are directly measurable. A dynamic observer of special type had to be developed to take account of the two types of disturbance on the vehicle. This gave rise to the final control architecture shown in Fig. 3.
Achieving the fusion of comfort and control
It took years of effort working on prototype vehicles and subsequent road testing to get to the stage when the algorithm was finally ready for a production vehicle. The algorithm was first tested in McLaren's driver-in-the-loop vehicle simulator, with positive results. This led to the commissioning and development of a prototype test vehicle, which confirmed the findings from the simulator and paved the way to production approval. The 720S was tested by McLaren in a wide range of motoring conditions including Death Valley, California, the western U.S. state of Colorado and the European Alps.
Fig. 4 (below) shows the algorithm's superior performance (black line) compared to fixed damper settings when negotiating a bumpy roundabout near the McLaren Technology Centre in Woking, U.K. The roundabout provides a combination of demanding handling and an uneven road surface, demonstrating the algorithm’s ability to simultaneously optimize comfort and control.
The integrated nature of the control of ride and handling responses gives the impression of a car which is rather softly sprung, yet delivering tight handling behavior. Such a combination would normally only be found in a stiffly-sprung vehicle with a harsh ride. The system demonstrates how active suspension (albeit semi-active in this case) can reduce or eliminate the trade-off between ride and handling performance that is inevitable in conventional suspensions.
Explained Simon Lacey, head of advanced engineering at McLaren Automotive: “The biggest leap has come in the software that controls the suspension system. It’s an intelligent system that successfully balances the requirements of ride comfort, body control and handling, dealing with road surface changes and challenging corners in a way that we’ve never seen before.”
Malcolm Smith is Professor of Control Engineering and Head of the Control Group in the Department of Engineering at the University of Cambridge.
Will Hoult is a Principal Engineer in the Modelling and Decision Sciences Group at McLaren Applied Technologies. Panos Brezas received a Ph.D. in control theory from the University of Cambridge in 2013, in an industrial collaboration with McLaren Automotive Ltd.
Continue reading »