Two-stage waveform relaxation method for the initial value problems with non-constant coefficients (Q2514077)

In this paper, the authors consider a two-stage waveform relaxation (TSWR) method for the initial value problem for a class of linear systems of ordinary differential equations of the form \[ y'+A(t)y=f(t), t\in [0,T],\tag{1} \] and \[ y(t_{0})=y_{0}\tag{2} \] where \(A:[t_{0},T]\rightarrow \mathbb R^{m\times m}\) is a non-singular \(M\)-matrix with continuous entries and \( f:[t_{0},T]\rightarrow \mathbb R^{m}\) is continuous. \(M\)-splitting of the form \( A(t)=C(t)-D(t)\) for the matrix \(A(t)\), \(t\in [t_{0},T]\), is suggested. The waveform relaxation (WR) continuous-time iteration for (1) and (2) is given by the \[ {y'}^{k+1}(t)+C(t)y^{k+1}(t)=D(t)y^{k}(t)+f(t)\leqno{(3)} \] and the \[ y^{k+1}(t_{0})=y_{0},\leqno{(4)} \] \(k=0,1,2,\dots\). Iterative method (3) and (4) is called the outer iteration. An inner iteration is then generated by using the splitting \(C(t)=M(t)-N(t)\). Thus the composite \(M\)-splitting of \(A(t)\) on \([t_{0},T]\) is given by \(A(t)=M(t)-N(t)-D(t)\). In this process, continuous-time TSWR iterations are generated. The Euler method is used for discretizing the continuous TSWR. A numerical example is given to illustrate the effectiveness of the method.