A high-speed generation of goal-oriented scenarios using combination of bidirectional propagation on qualitative and quantitative hybrid model

We address generating efficiently goal-oriented scenarios that are based on initial values and lead to goal values. These scenarios are generated with qualitative and quantitative hybrid simulation, which propagates the effect from initial values by Monte Carlo Simulation. Because the most of scenarios do not often lead to the goal value, we take an approach to combine the scenarios by hybrid simulation (forward simulation) and the scenarios by the inverse simulation that decides initial values from the goal values. The high-speed generation method of goal-oriented scenarios combines the scenarios when the qualitative values on both scenarios are equal. Because the inverse simulation takes longer than the forward simulation, the proposed method keeps the number of trials of the inverse simulation small. As a result of experiments, it is confirmed that the proposed method can decrease the computational time by over 98% compared to the method using only forward simulation.