Introduction Adaptive Cruise Control (ACC) enables automatic following of a vehicle. The relative distance xr is controlled (see Fig. 1). A driver dependent part determines the desired host vehicle acceleration ah,d, while a vehicle dependent part controls the longitudinal dynamics via actuation of the throttle uth and brake system ubr (see Fig. 2). Fig. 1: ACC system setup. Problem statement Focusing on the driver dependent part, nonlinear (situation dependent) driver behaviour generally is accounted for in the controller design via scheduling gains and switching logic, while disregarding stability issues. Furthermore, the lack of appropriately defined performance metrics yields timeconsuming tuning by trial-and-error. Hence, performance metrics as well as a structured control framework for ACC are required. Fig. 2: ACC control structure. ACC performance evaluation On the basis of literature and on-the-road experiments, metrics are determined to enable objective performance evaluation of an ACC system in a qualitative manner. In case of a passenger car, both comfort and desirability have to be considered. Comfort is mainly related to vestibularly detectable variables, whereas desirability is mainly related to visually and auditorily detectable variables. Regarding desirability, xr, vr and the so-called time-tocollision TTC = xr/vr are the most promising metrics, yet some situation dependency seems inevitable. Regarding comfort, acceleration and jerk peak values are appropriate metrics enabling objective performance evaluation. ACC design Besides the control objective regarding xr, the relative velocity vr should be limited based on desirability and the acceleration and jerk should be limited out of comfort reasons. Furthermore, the nonlinear driver behaviour as well as safety considerations yield various (nonlinear) constraints on the control output ah,d. Model Predictive Control (MPC) is adopted as a suitable, structured framework for constrained, MIMO, nonlinear controller design. MPC minimizes a cost function J regarding the control output u over a user-defined prediction horizon; min u J(u, e,R), with e the error with respect to the control objectives and R the performance related requirements. Adopting a closed-loop MPC synthesis enables explicit, offline optimization of the state-dependent controller gains. This yields a hybrid control synthesis, which prevents the need for significant online computational power. Simulations as well as on-the-road experiments have been executed, showing appropriate behaviour of the ACC system. Fig. 3: Screenshot of the simulation environment and the Audi S8 with which the ACC is tested. Future work Current research focusses on further integration of the performance metrics in the tuning process and the possibly automated, driver-specific tuning.
|Title of host publication||Book of Abstracts of the 27th Benelux Meeting on Systems and Control|
|Place of Publication||The Netherlands, Heeze|
|Publication status||Published - 2008|
|Event||27th Benelux Meeting on Systems and Control, March 18-20, 2008, Heeze, The Netherlands - Heeze, Netherlands|
Duration: 18 Mar 2008 → 20 Mar 2008
|Conference||27th Benelux Meeting on Systems and Control, March 18-20, 2008, Heeze, The Netherlands|
|Period||18/03/08 → 20/03/08|