Algorithms for general nonlinear bilevel programs

This paper addresses a general nonlinear nonconvex bilevel programming problem (BLPP). Such problems may have proper local solutions, and thus often require special (often laborious) enumeration methods to guarantee global optimality. However, some economic models are known to be free of such difficulties, and the authors investigate methods using gradient information from the lower-level problem that can be used to develop efficient algorithms guaranteed to converge to regular points of a BLPP under appropriate conditions.