A leader-follower partially observed, multiobjective Markov game

The leader-follower partially observed, multi-objective Markov game (LF-POMG) models a sequential decision making situation with two intelligent and adaptive decision makers, a leader and a follower, each of which can choose actions that affect the dynamics of the system and where these actions are selected on the basis of current and past but possibly inaccurate state observations. The decision makers can be cooperative, non-cooperative, or a mixture of both. The leader can consider multiple objectives and the follower considers a single objective in selecting their decision-making policies. The LF-POMG is a new, relatively unexamined combination of a stochastic game and a partially observe Markov decision process (POMDP). We investigate structural properties for this decision-making model and develop a computational procedure for finding good suboptimal solutions. We also show how the leader's observation quality of the follower's state affects the performance of the leader, representing an analysis of the value of information for the LF-POMG. We present the results of a case study where the LF-POMG has been used to provide decision support to the manager of a food processing facility whose objectives are to maximize facility productivity while minimizing the risk of an intentional insertion of a biological or chemical toxin into the facility.