基于生长神经网络的进化机器人行为研究

Research on Evolutionary Robot Behavior Using Developmental Network

提出了一个基于生长神经网络的进化机器人行为算法 ,新算法的主要特点是 :1、通过自然选择对神经网络进行进化 ,并能自主实现机器人避碰、移动、复制和攻击等行为 ;2、开发了一个自主机器人模拟环境 ,对所提出的算法进行运行测试 .模拟结果证明 ,生长神经网络系统是研究渐增进化的有效工具 ,新算法能够有效地实现机器人创新行为 .

This paper propose an evolutionary robot behavior algorithm which is based on developmental network.The new algorithm use nature selection to evolve neural network and can autonomously avoid obstacles,move,reproduce and attack.A simulation environment is developed to test the new algorithm.The experimental results demonstrate that the developed system is well suited to long term incremental evolution and original behaviors evolution.