Integral action in robust adaptive control

Conditions are studied under which integral action in robust adaptive control is soundly based. The method of analysis amounts to verifying that the key conditions needed for convergence are satisfied. A simple strategy is described for choosing the integral constant and global convergence is established for the resultant algorithm. This illustrates a proof paradigm which could be similarly applied to other problems. The principal assumptions used are that the plant, when augmented by an integrator, has a known degree of controllability and that an overbounding function is known for the unmodeled system response. The continuous-time case is treated, but the corresponding discrete-time results follow mutatis mutandis by simply replacing p =d/dt by δ=q-1/Δ, L ₂ by l₂, ∫ by Σ, and so on