A dynamic programming approach to optimal pollution control under uncertain irreversibility: The Poisson case

We solve a bimodal optimal control problem with a non-concavity and uncertainty through a Poisson process underlying the transition from a mode to another. We use a dynamic programming approach and are able to uncover the global optimal dynamics (including optimal
non-monotonic paths) under a few linear-quadratic assumption, which do not get rid of the non-concavity of the problem. This is in contrast to the related literature on pollution control under irreversibility which usually explores local dynamics along monotonic solution paths to first order
Pontryagin conditions.