Constrained control of vehicle steering

In this paper we illustrate the development of ldquoadd-onrdquo control schemes for vehicle applications based on reference governors and specially formulated model predictive controllers. These schemes are normally inactive but become active when necessary to enforce pointwise-in-time constraints on vehicle variables and assist the driver in maintaining control of the vehicle. A specific case study is considered where reference governors and model predictive controllers are applied to modify vehicle steering. The on-line computing effort associated with the reference governor (explicitly solvable scalar optimization problem) is generally less than of the model predictive controller (simple and explicitly solvable parametric quadratic program), while the domain of recoverable states of the model predictive controller (with an appropriately defined cost function and terminal set constraints) can be much larger. Systematic approaches to reduce on-line computing effort based on the use of simply characterized subsets of constraint feasible sets are proposed.