ℒp-robustness and exponential convergence of some estimation schemes in adaptive control

We suggest a design procedure for L_p-stable estimation schemes, and use the insights gained from this design to highlight new convergence and robustness properties of well-known estimation schemes, such as leakage-based (switching sigma) and parameter projection algorithms. In particular, we show that new robustness and convergence statements can be obtained if we further restrict the noise to lie in L_p∩L_∞ , 1<p<∞. We also indicate connections with results in robust statistics