Optimization of a missile attack with collaborating missiles - A survey

This report gives an overview over a class of techniques and methods associated with the optimal guidance of a missile attack-mission with collaborating missiles. It focuses on methods relying on the conceptual notion of agent and multi-agent systems (MAS) first defined within the field of Artificial intelligence (AI). Machine-learning, a field within AI, describes methods aimed to program an agent behavior by adaption through direct interaction with the environment. Particularly, the Reinforcement Learning expand the ideas existed in optimal control theory, numerical optimization and stochastic approximation theory. The agent behavior is constructed as a solution to an optimization problem with a cost function provided by the designer, which hopefully is a much easier task than specifying the behavior itself. A MAS is defined as a loosely coupled network of autonomous agents working in collaboration to solve a problem which is beyond the agents individual capabilities. Features, advantages and solutions to the problem of collaborated MAS is discussed from an organizational point of view, and how a limited communication may critically affect the global performance. A number of issues related to optimization of problems with Reinforeemnt Learning are presented. Finally a literature list finish the report.