Abstract: Combining decision-dependent uncertainty with options evaluation, a least square simulation algorithm was presented to evaluate the real options under the condition of the state variable's stochastic distribution being affected by the actions of decision-makers. It is the core of the algorithm, that the least square was applied repeatedly backward to compute the estimates of termination value, continuation value and the option value on the decision points of each sample path. So it solves the difficult problems to simulate the sample paths of the state variable under the decision-dependent uncertainty due to lack of the optimal stopping rules. The applicability of the algorithm was demonstrated by a numerical example about decision-making of on-orbit upgrading of commercial communication satellites. The method expands the evaluation of options from exogenous uncertainty to endogenous uncertainty, and enriches the numerical methods for options pricing.