Decision-theoretic planning of emergency task network

Emergency Response Task Planning (ERTP) is a kind of Random Sequential Decision Making Problems (RSDMP). There have been few relevant researches so far, and the classic Markov Decision Processes (MDP) is not completely suitable for it because of the curse of dimensionality. In order to solve ERTP, this paper proposes a Markov decision method based on Acceptable Risk Level (ARL) and Dynamic Constraint Satisfaction Problem (DCSP). On the one hand, the decision maker´s risk attitude can be reflected by ARL, and on the other hand the new method is able to reduce the total state space of the problem and increase computation efficiency. Experiment results prove that the new method has much higher computation efficiency than classic MDP algorithms and DCSP based MDP algorithms, and it´s more suitable for complex RSDMPs.