Numerical solutions of index-1 differential algebraic equations can be computed in polynomial time (Q2492797)

It is shown that the cost to solve initial value problems (IVPs) for semi-explicit index-1 differential-algebraic equations (DAEs) is polynomial in the number of bits of accuracy. The results extend that of previous results which showed the cost of solving IVP for ordinary differential equations is polynomial in the number of bits of accuracy. This cost is obtained for an algorithm based on the Taylor series method for solving differential algebraic equations developed by \textit{J. D. Pryce} [ibid. 19, No. 1--4, 195--211 (1998; Zbl 0921.34014); BIT 41, No. 2, 364--394 (2001; Zbl 0989.34005)]. The analysis assumes that the functions defining the DAE are piecewise analytic. This is the key assumption, allowing the use of arbitrary order methods. The results of the standard theory of information-based complexity give exponential cost for solving ordinary differential equations, being based on a different model.