论文部分内容阅读
文章研究了一类基于自适应共生进化模型的自主体构造方法，提出了基于自适应共生进化算法（ＡＳＥ）的一种进化强化学习方法，该方法不仅可以高效地进行神经网络决策系统的设计，而且在多自主体组成的自律系统的群体行为进化中可快速学习而收敛，并且通过多样度调节维持了群体多样性，克服了未成熟收敛现象。研究者将该方法用于人工生物中求偶通讯规范的研究，通过在环境中的生存和行为学习，雌雄个体都可有效地学习到成功的求偶通讯规范。
This paper studies a self-body construction method based on adaptive co-evolution model and proposes an evolutionary reinforcement learning method based on adaptive co-evolutionary evolution algorithm (ASE). This method not only can effectively design the neural network decision-making system, But also learns quickly and converges in the evolution of group behavior of autonomous systems composed of multiple autonomies, and maintains the diversity of groups through diversity adjustment and overcomes the immature convergence phenomenon. Researchers apply this method to the study of courtship communication in artificial biology. Through the study of survival and behavior in the environment, both male and female individuals can effectively learn the successful courtship communication specification.