Learning control based on pattern recognition applied to vehicle cruise control systems

Modern learning technology has a great deal to offer in the practical application of vehicle cruise control. Using this approach a single controller can be developed for a wide class of vehicle/engine configurations without the need for lengthy individual setting-up and tuning to take account of the many variations that occur in practice between vehicles, engines and actuator geometry. A learning controller is designed using pattern recognition techniques, which are used online for observing patterns of the process output and then applying a set of tuning rules to the recognised patterns in order to set the appropriate controller settings. This process is activated automatically by the learning controller by monitoring the current control performance and deciding if it is necessary to re-tune the controller settings. Alternatively the process is activated manually by the driver using a push-button. A simulation study of a vehicle cruise control system shows the simplicity and effectiveness of the learning control approach