基于SARSA在线规划的软件体系结构自适应

提出基于SARSA算法的在线规划软件体系结构自适应方法,用来解决由于环境内在固有的不确定性、复杂性和不可预见性而产生的离线规划的局限性。在线规划方法指可以根据当前的环境状况自动选择行动的规划方法。结合Robocode的实例详细阐述了实现基于SARSA算法的在线规划方法的三个关键问题和过程策略;为解决自适应的状态和行动表述、适应度和可受理集合关键问题,提出了自适应在线规划的策略。最后用Robocode的坦克战斗实例,证明了基于SARSA在线规划软件体系结构自适应方法的可行性和有效性。

SARSA-based on-line planning software architecture self-adaptation

This paper proposed SARSA-based on-line planning software architecture adaptive method to avoid the limits of off-line plamming made by the environments of the complex software systems with the inherent uncertainty, complexity and unpredictability. On-line planning methods could select the action automatically based on the current state of the environment. This paper described the key elements of on-line planning methods and process strategies. Finally used a Robocode instance named SARSABot as a case to fight with FireBot. The result of the experiment proves the feasibility and effectiveness of the SARSA based on-line approach.