Adaptive dead-zone compensation for output-feedback canonical systems (Q4353769)

The following system is under consideration: \[ \dot x=Ax+\varphi_0(x_1)+\sum_{i=1}^qa_i\varphi_i(x_1)+b\sigma(x_1)u, \] where \(x=(x_1,\ldots,x_n)^T\in \mathbb{R}^n\), \(\varphi_i(x_1)\in \mathbb{R}^n\), \(A\) is a \(n\times n\) matrix with elements \(a_{i,i+1}\equiv 1\) and \(a_{i,j}=0\) for any other \((i,j)\), \(b=(0,\ldots,0,b_m,\ldots,b_0)^T\). \(\sigma\) and all the \(\varphi\)'s are known and the \(\varphi_i\)'s are smooth. Only \(x_1\) is observable. All \(a_i\)'s and \(b_j\)'s are unknown, but \(b_ms^m+\cdots+b_1s+b_0\) is a Hurwitz polynomial and \(\text{sign }b_m\) is also known. The problem is to track a given ``reference signal'' \(r(t)\) which is bounded with its \(\rho\) derivatives using the control \(v\) via \(u=DZ(v)\) which has the ``dead-zone characteristics'' of the following structure: \[ DZ(v(t))=m_r(v(t)-b_r)1(v(t)\geq b_r)+m_l(v(t)-b_l)1(v(t)\leq b_l). \] Here all \(m\)'s and (new) \(b\)'s are constants which are known up to certain errors. The adaptive control design is constructed for which the tracking error is bounded as well as some other characteristics. The adaptive estimation technique for unknown parameters is applied. The Lyapunov function approach is used.