Adaptive Robustification of Uncertain Control Systems

Linear control design uses some "nominal" plant as the object of design. Fixed linear controllers then attempt to maintain some performance in spite of the uncertainty usually existen in real, world. When the plant parameters are expected to vary within large ranges, gain scheduling is used to switch the gain values as a predetermined function of some monitored parameters. In the presence of large and, possibly time-varying, uncertainties, these controllers may result in poor behavior or even in total divergence. Various adaptive techniques have been conceived to cope with uncertainty. However, real-time identification of plant parameters may be too slow to allow for a solution in fast environments such as flight control. Simple adaptive control techniques have been developed for uncertain systems. These techniques utilize the stabilizability property of the plant to be controlled and automatically tune the control gains. The stability and performance are maintained in spite of the large variations in plant parameters. The simple adaptive controller will continuously tune the controller gains as a function of the monitored performance (in particular, the difference between the desired output and the actual plant output, namely the tracking error). Thus, the adaptive gains increase only at those moments when high gains are required to maintain the performance, and may decrease immediately afterwards. The adaptive controller acts to fit the right gains at the right time as a function of the changing need. This way, during regular times the control system is almost unaffected by adaptation, but it is strongly supported during difficult times.