New paper on zeroth-order deterministic policy gradient for finite and infinite horizon dynamic programs under preparation - check "publications", preprint out soon.