A self-adjusting control system with large initial error

A performance measuring, plant-adaptive control system is described. A linear plant having two variable parameters induces a large system error which is reduced by adjustments made in two controller parameters. An adaptive computer, the logical design of which is based on a modification of Newton's method of descent, generates the controller-parameter changes from measurements of the integral-of-error squared (ISE). This method is evaluated with regard to the time interval required to reduce ISE from a large initial value to a small prescribed limit. The average adaptation time for this system is shown to be approximately sixty times the response time of the system when the system error is zero.